PDA

View Full Version : Rating computation suggestion



Pages : [1] 2

pax
02-09-2004, 02:01 PM
I have a suggestion that I may have mentioned before.

Is it possible to implement the Glicko calculations such that ratings are updated after every game? That is, instead of using the last published rating for every rating calculation in a period, you instead keep track of updated ratings and use the "current rating" for each calculation.

I believe this is the way Glicko is intended to be calculated, and indeed the way it is calculated on FICS.

This method of calculation is one way of avoiding the large swings in rating that can occur due to the time between lists. E.g if you have a number of poor performances throughout a rating list, you will lose points based on your initial high rating for each game you play. If ratings are recalculated after each game, then losses later in the list will be lower due to the reduction in rating that has already occurred (argument also holds for increases in rating due to prolonged good performance).

In order for this to work, you ideally need some "time" information in order to place the games in order for processing. The best result would be if games are sequenced individually, but I realize it may not be possible to extract separate date info for each game. Sequencing can be done on a tournament by tournament basis - it may lead to slightly inaccurate ordering (for example if people are playing weekly in one tournament and also in a weekender), but this would still be better than the present batch processing.

Processing the games this way would definitely take longer, but should still be feasible.

The main downside is that it becomes more difficult for people to estimate their rating change through a period, since they don't know what rating will be used for their opponent in a particular game. This isn't a big deal at the moment however, since any calculation is approximate unless you know everybody's RD and volatility precisely.

Any thoughts on this approach in principle?

Pax


p.s

Come to think of it, the pronciple difficulty is probably getting tournament organisers to submit their results punctually. When tournaments are submitted months late, this would obviously mess with the system. Still, it messes with the system as it stands in any case.

Bill Gletsos
02-09-2004, 02:29 PM
I have a suggestion that I may have mentioned before.

Is it possible to implement the Glicko calculations such that ratings are updated after every game? That is, instead of using the last published rating for every rating calculation in a period, you instead keep track of updated ratings and use the "current rating" for each calculation.

I believe this is the way Glicko is intended to be calculated, and indeed the way it is calculated on FICS.
Glicko's documents make it clear he expects a group of results to be batched and not calculated one game at a time.


This method of calculation is one way of avoiding the large swings in rating that can occur due to the time between lists.
Actually it doesnt.
I looked at this years ago and it makes very little difference for the majority of players. It has an adverse effect on rapidly improving players.


E.g if you have a number of poor performances throughout a rating list, you will lose points based on your initial high rating for each game you play. If ratings are recalculated after each game, then losses later in the list will be lower due to the reduction in rating that has already occurred (argument also holds for increases in rating due to prolonged good performance).
This isnt correct.


In order for this to work, you ideally need some "time" information in order to place the games in order for processing. The best result would be if games are sequenced individually, but I realize it may not be possible to extract separate date info for each game. Sequencing can be done on a tournament by tournament basis - it may lead to slightly inaccurate ordering (for example if people are playing weekly in one tournament and also in a weekender), but this would still be better than the present batch processing.
This is nigh impossible to organise.


Processing the games this way would definitely take longer, but should still be feasible.
There is no actual advantage in doing this.


The main downside is that it becomes more difficult for people to estimate their rating change through a period, since they don't know what rating will be used for their opponent in a particular game. This isn't a big deal at the moment however, since any calculation is approximate unless you know everybody's RD and volatility precisely.
The main downside is that your approach has significant impact on rapidly improving players.



Any thoughts on this approach in principle?
Yes, it isnt going to happen. ;)

Alan Shore
02-09-2004, 04:32 PM
Looks like someone's been reading up on their Edwards, eh Bill? ;)

I'm guessing it's the suggestions to change the rating system that's given you ants in your pants, right? :rolleyes:

bobby1972
02-09-2004, 04:46 PM
gliko is ok leave it alone ,me and 2 of my mates have been over 2000 many times thanks to gliko .we love it

Bill Gletsos
02-09-2004, 05:28 PM
Looks like someone's been reading up on their Edwards, eh Bill? ;)

I'm guessing it's the suggestions to change the rating system that's given you ants in your pants, right? :rolleyes:
Nope.
I'd looked at this issue ages ago.
Originally back before we implemented Glikco and again before implementing Glicko2.

pax
03-09-2004, 12:07 AM
Glicko's documents make it clear he expects a group of results to be batched and not calculated one game at a time.


Theoretically, it is far better to calculate one game at a time. Can you reference the quote? It certainly is true that the calculations can still take place in a batch, it's just that intermediate ratings are calculated at each step, not just rating changes which are summed at the end.

It's quite obvious that the best rating change for a particular game will be calculated if the most up to date information is used. The most up to date information should incorporate the information for all games prior.



Actually it doesnt.

I looked at this years ago and it makes very little difference for the majority of players. It has an adverse effect on rapidly improving players.


The only adverse effect it would have is to prevent the calculated rating from overshooting the true rating. This can clearly happen in the current system.




This isnt correct.


Great. That's a really helpful comment. Why isn't it correct? Barry Cox's Glicko estimator certainly seems to suggest this is the case.

Example: Player A 1700! plays a series of games against 1500!! players. For every win , 11 points are gained, for every loss 35 points are lost. This number is the same regardless of the number of games played in the period. After three losses, player A is effectively only 1600, but continues to lose 35 points for every loss rather than the 28 he would lose if his rating was actually 1600. There will be similar effects if RD is not updated until after all the games in the period are processed.



This is nigh impossible to organise.

Individual game dating would be difficult, I agree. Tournament dating is surely standard information. Then sequence tournaments by completion date, and games by round number. It's not perfect, but it's ok and not at all hard to implement.



There is no actual advantage in doing this.


There is an advantage. Namely, that it results in calculations that has one extra layer of approximation removed. Glickman's formulas have layer upon layer of approximation as it is, and this is one layer that is easily fixed in the calculation process.



The main downside is that your approach has significant impact on rapidly improving players.


Like I said, the only impact is to avoid artificial overshoot.

e.g player B 1300? plays a series against 1700!! players. Each win earns 150 points, and each loss loses 15 points. These numbers stay the same even after a few wins (until the point when rating equals performance rating).

Note that the problem that causes ratings to potentially increase above the performance rating for a period (or drop below performance rating) does not occur if you calculate game by game. This is the problem that you have admitted has had to be fixed artificially. It is symptomatic of a more general problem with batch processing the ratings.




Yes, it isnt going to happen. ;)

Yes, that much is perfectly clear. You don't have to be so supercilious about it though. The suggestion is meant entirely in good spirit.

Pax

pax
03-09-2004, 12:20 AM
Glicko's documents make it clear he expects a group of results to be batched and not calculated one game at a time.

Ok, I found it. From Glickman's web site:

"To apply the rating algorithm, we treat a collection of games within a ``rating period'' to have occurred simultaneously. A rating period could be as long as several months, or could be as short as one minute. In the former case, players would have ratings and RD's at the beginning of the rating period, game outcomes would be observed, and then updated ratings and RD's would be computed at the end of the rating period (which would then be used as the pre-period ratings and RD's for the subsequent rating period). In the latter case, ratings and RD's would be updated on a game-by-game basis (this is currently the system used by FICS). The Glicko system works best when the number of games in a rating period is moderate, say an average of 5-10 games per player in a rating period. The length of time for a rating period is at the discretion of the administrator."

Ok, games in a 'period' are treated simultaneously. But he does say, that there's nothing stopping a 'period' from being defined as a single game. Glickman says that the system works best with 'moderate' periods (5-10 games). That would suggest treating each tournament as a period for the purposes of the rating computations. I still believe that the separate game option is the better.

Note that the last list has one player with 78 rated games, several with over 50 games and many more over 30. Most active players in the period played more than 10 games.

Pax

Bill Gletsos
03-09-2004, 01:12 AM
Theoretically, it is far better to calculate one game at a time.
You would have an impossible time supporting this by any statement by Glickman where he says this in his papers.


Can you reference the quote?
I never said it was a quote.


It certainly is true that the calculations can still take place in a batch, it's just that intermediate ratings are calculated at each step, not just rating changes which are summed at the end.
So what it makes virtually no difference and certainly does not warrant the extra effort involved in attempting to rate the games individually.


It's quite obvious that the best rating change for a particular game will be calculated if the most up to date information is used. The most up to date information should incorporate the information for all games prior.
Check out his formula. It all revolves around summing game results in a period where games are from 1...n.



The only adverse effect it would have is to prevent the calculated rating from overshooting the true rating.
Incorrect. The rating for a rapidly improving player does not increase as dramatically if you rate it game by game.


This can clearly happen in the current system.
Irrelevant to the point at hand.



Great. That's a really helpful comment. Why isn't it correct? Barry Cox's Glicko estimator certainly seems to suggest this is the case.
Easy.
Barry Cox's calculator is wrong under these circumstances because it does not allow for the different c squared value required to age the rating between games if you rate the games one game at a time.



Example: Player A 1700! plays a series of games against 1500!! players. For every win , 11 points are gained, for every loss 35 points are lost. This number is the same regardless of the number of games played in the period.
If you are going to make statements like this as though they are fact then be damn sure you know what your talking about.
Unfortunately for you, you dont.

Your figures are Incorrect.
You clearly have not done the actual calculations.
Each win does not gain 11 points. Successive wins give rise to reduced increases per win. The same is true for losses.

In fact its clear you didnt even use Barry Cox's calculator, or if you did you did so incorrectly.
Using his web page calculator it is obvious that a 1700! beating a 1500! in just one game gives rise to a change of 13 points. However if the same 1700! beats the 1500!! 5 times the change then averages 11 points per game for a total increase of 55 points. If he had happened to win 10 games the average increase drops to 8.9 points per game for a total increase of 89 points. thats a far cry from you supposed 110 points.



After three losses, player A is effectively only 1600, but continues to lose 35 points for every loss rather than the 28 he would lose if his rating was actually 1600. There will be similar effects if RD is not updated until after all the games in the period are processed.
Incorrect.


Individual game dating would be difficult, I agree. Tournament dating is surely standard information. Then sequence tournaments by completion date, and games by round number. It's not perfect, but it's ok and not at all hard to implement.
There is no significant advantage in doing this.



There is an advantage. Namely, that it results in calculations that has one extra layer of approximation removed. Glickman's formulas have layer upon layer of approximation as it is, and this is one layer that is easily fixed in the calculation process.
It isnt warranted.



Like I said, the only impact is to avoid artificial overshoot.
Like I said that isnt the impact.


e.g player B 1300? plays a series against 1700!! players. Each win earns 150 points, and each loss loses 15 points. These numbers stay the same even after a few wins (until the point when rating equals performance rating).
Another apparent statement of fact that is totally false.
Same reasons as above.


Note that the problem that causes ratings to potentially increase above the performance rating for a period (or drop below performance rating) does not occur if you calculate game by game. This is the problem that you have admitted has had to be fixed artificially. It is symptomatic of a more general problem with batch processing the ratings.
Incorrect.
You clearly have not done any actual calculations.
This problem can occur in game by game or batch pricessing. It even happens in the ELo system irrespectve of whether you do game by game or batch.


Yes, that much is perfectly clear. You don't have to be so supercilious about it though. The suggestion is meant entirely in good spirit.
Not true.
If I was being supercilious or even sarcastic I would have not put a ;) at the end of the sentence.

Bill Gletsos
03-09-2004, 01:13 AM
Ok, I found it. From Glickman's web site:

"To apply the rating algorithm, we treat a collection of games within a ``rating period'' to have occurred simultaneously. A rating period could be as long as several months, or could be as short as one minute. In the former case, players would have ratings and RD's at the beginning of the rating period, game outcomes would be observed, and then updated ratings and RD's would be computed at the end of the rating period (which would then be used as the pre-period ratings and RD's for the subsequent rating period). In the latter case, ratings and RD's would be updated on a game-by-game basis (this is currently the system used by FICS). The Glicko system works best when the number of games in a rating period is moderate, say an average of 5-10 games per player in a rating period. The length of time for a rating period is at the discretion of the administrator."

Ok, games in a 'period' are treated simultaneously. But he does say, that there's nothing stopping a 'period' from being defined as a single game. Glickman says that the system works best with 'moderate' periods (5-10 games). That would suggest treating each tournament as a period for the purposes of the rating computations.
Actually it doesnt.
As Glickman notes with regards the Glicko2 system he states "The Glicko-2 system works best when the number of games in a rating period is moderate to large, say an average of at least 10-15 games per player in a rating period".
First it was 5-10 now 10-15. Based on your logic that makes it two tournaments.


I still believe that the separate game option is the better.
Believe what you like it does not make it true.


Note that the last list has one player with 78 rated games, several with over 50 games and many more over 30. Most active players in the period played more than 10 games.
The number of games that it works best for is dependant on a number of factors. As Glickman said in an email you have to experiment with the various parameters to determine what works best in your circumstances.
This is especially true with the Glicko2 system.

PHAT
03-09-2004, 06:28 AM
As Glickman said in an email you have to experiment with the various parameters to determine what works best in your circumstances.

:eek: FMD. Isn't Glicko capable of "handling" normal circumstances such as pooling. :doh:

BTW, could you name these parameters, and tell us how they effect the "handling" of pooling. :owned:

Bill Gletsos
03-09-2004, 10:47 AM
:eek: FMD. Isn't Glicko capable of "handling" normal circumstances such as pooling. :doh:
Who mentioned pooling.


BTW, could you name these parameters, and tell us how they effect the "handling" of pooling. :owned:
I'm not your nor anyone else's lecturer/research assistant.

Again I ask, who mentioned pooling.
If referring to processing a number of games in a rating period, then I'll leave that as an exercise for the interested reader. Of course to anyone who reads Glickmans Glikco and Glicko2 papers and understands them then the answers should be obvious.

pax
03-09-2004, 11:20 AM
Your figures are Incorrect.
You clearly have not done the actual calculations.
Each win does not gain 11 points. Successive wins give rise to reduced increases per win. The same is true for losses.

In fact its clear you didnt even use Barry Cox's calculator, or if you did you did so incorrectly.
Using his web page calculator it is obvious that a 1700! beating a 1500! in just one game gives rise to a change of 13 points. However if the same 1700! beats the 1500!! 5 times the change then averages 11 points per game for a total increase of 55 points. If he had happened to win 10 games the average increase drops to 8.9 points per game for a total increase of 89 points. thats a far cry from you supposed 110 points.

Ok, my example was not accurate, but the fundamental point I'm making is still true.

The reduced increases per game arise from an effective adjustment in RD, not an adjustment to the rating. Note that changes in both directions are reduced, and the contribution of games at the start of the period is the same as games at the end of the period.

Here is a better example:

Player A 1500! plays ten games against players 1500!!. if the sequence is 1010101010, the change is 16 points per game (up or down depending on win and loss). If the sequence is 1111111111, the gain is still 16 points per game.

What I'm saying, is that after 9 wins against 1500 players you have a pretty good idea that this player is correctly rated higher than 1500, and therefore the gain for the tenth win should be lower than the gain for the first win. If you calculate game by game, then by the time the tenth game is calculated, the information from the first nine games is encoded in the ratings used for the tenth game.

Pax

Rincewind
03-09-2004, 11:38 AM
Ok, my example was not accurate, but the fundamental point I'm making is still true.

The reduced increases per game arise from an effective adjustment in RD, not an adjustment to the rating. Note that changes in both directions are reduced, and the contribution of games at the start of the period is the same as games at the end of the period.

Here is a better example:

Player A 1500! plays ten games against players 1500!!. if the sequence is 1010101010, the change is 16 points per game (up or down depending on win and loss). If the sequence is 1111111111, the gain is still 16 points per game.

What I'm saying, is that after 9 wins against 1500 players you have a pretty good idea that this player is correctly rated higher than 1500, and therefore the gain for the tenth win should be lower than the gain for the first win. If you calculate game by game, then by the time the tenth game is calculated, the information from the first nine games is encoded in the ratings used for the tenth game.

When you reduce the batch size down to 1 game you are not guaranteed to get a more accurate rating, just a more volatile one. Your argument would have more weight if it could be shown that the resulting system did lead to greater accuracy in ratings and also if you could suggest a way of overcoming the logisitical problems of getting games rated within 7 days of being played. Otherwise I could be playing against someone who's current rating is 1500 but won 10 games last weekend which were not rated.

Yes this happens now but to a lesser degree as there is a conscious effort to round up results and include them when ratings are about to be published. If this was just an ongoing chore I could see lots of results falling through the cracks for weeks, months and even forgotten about altogether.

pax
03-09-2004, 11:50 AM
This problem can occur in game by game or batch pricessing. It even happens in the ELo system irrespectve of whether you do game by game or batch.

Really? Can you give an example where such overshoots happen in either Glicko or ELO when calculated game by game? It is very easy to construct such examples in batch calculations, but it's certainly less obvious in the game by game scenario.

Pax

pax
03-09-2004, 12:44 PM
When you reduce the batch size down to 1 game you are not guaranteed to get a more accurate rating, just a more volatile one. Your argument would have more weight if it could be shown that the resulting system did lead to greater accuracy in ratings and also if you could suggest a way of overcoming the logisitical problems of getting games rated within 7 days of being played. Otherwise I could be playing against someone who's current rating is 1500 but won 10 games last weekend which were not rated.

You misunderstand my point.

I'm not suggesting that the ratings be actually calculated in real time. What you do is collect all the results for a period (say three months), you place the games in order and process them one by one. As I say, the ordering will be approximate when tournaments overlap.

I would have thought my suggestion would make ratings less volatile rather than more. This is because when a rating is going up (or down), the calculations are always based on all of the available information up to that point in time (i.e current calculated ratings).

Pax

Rincewind
03-09-2004, 01:17 PM
I would have thought my suggestion would make ratings less volatile rather than more. This is because when a rating is going up (or down), the calculations are always based on all of the available information up to that point in time (i.e current calculated ratings).

OK. That's not as bad as I thought then from a logistical point of view. The only thing you would need is the date played which is not currently collected reliably I believe.

However, not sure if the system would lead to ratings which more reliably predict future performance though. Might just be a lot of stuffing around for a negligible improvement in predictiveness.

Bill Gletsos
03-09-2004, 02:05 PM
Ok, my example was not accurate, but the fundamental point I'm making is still true.
I disagree.


The reduced increases per game arise from an effective adjustment in RD, not an adjustment to the rating.
This isnt correct.
There is no effective reduction in the RD. The number of games that are batched effects the d squared value. The RD value of the player remains constant throught the period.


Note that changes in both directions are reduced, and the contribution of games at the start of the period is the same as games at the end of the period.
Irrelevant.



Here is a better example:

Player A 1500! plays ten games against players 1500!!. if the sequence is 1010101010, the change is 16 points per game (up or down depending on win and loss). If the sequence is 1111111111, the gain is still 16 points per game.
Your conclusions are incorrect and your example isnt a good one.
In the case of two equally rated players the E(s|r,rj,RDj) is always 0.5 irrespective of the RD.
Also Barry's delta value is misleading.

In your above example after 8 games of the win/loss sequence the players rating remains 1500 and the 9th game win adds around 16.8 rating points for a rating of 1517. By adding a 10th game the players rating drops back to 1500, hence the addition of that game effective subtracts 16.8 points.
If there was an 11th game win that would only contribute 15.3 points. A 12th game loss would contribute a 15.3 point loss. As for your 10 win sequence see below.
However what if your example was a 1500! playing a 1600!! with the same scenarios.
After 8 games of the win loss sequence the players rating is 1540. The 9th win adds 20 points for a rating of 1560. The 10th game loss is an effective loss of 15 points and the players rating is 1545. An 11th game win would contribute 19 points and the players rating is now 1564. A 12th game loss would result in a 14 point loss for a rating of 1550.


What I'm saying, is that after 9 wins against 1500 players you have a pretty good idea that this player is correctly rated higher than 1500, and therefore the gain for the tenth win should be lower than the gain for the first win. If you calculate game by game, then by the time the tenth game is calculated, the information from the first nine games is encoded in the ratings used for the tenth game.
This isnt correct.
The gain from the tenth win is less than the gain for the first 9.
If the player only played 9 games and won them all his rating gain is 151 points for an average gain of 16.8. If he played 10 and won them all then his rating gain is 160 points for an average of 16.


Bottom line is this if your 10 game win loss sequence was played at the rate of one game per week and the rating period is one week then the players rating after 10 weeks would be approx 1498.7 where as if the 10 games are batched the players rating would be exactly 1500.

Bill Gletsos
03-09-2004, 02:05 PM
Really? Can you give an example where such overshoots happen in either Glicko or ELO when calculated game by game? It is very easy to construct such examples in batch calculations, but it's certainly less obvious in the game by game scenario.
Firstly it should be pointed out that the occurence of overshoot or undershoot occurs very infequently in Glicko or Elo.
In fact for all the rating periods since Dec 2000 to the present there has been 63 occurences from 20425 rating updates.

The simplest example under Elo is as follows.
A player has a rating of 1500.
He plays with 10 different players all rated 1500.
His results are 1010101010.
The players performance rating over the 10 games is 1500.
However after rating the 10 results individual the players rating at the end of the 10 games is 1499.12. This is an undershoot.
If the results had been in the order 0101010101 then his rating would have ended up at 1500.75 which represents an overshoot.

Bill Gletsos
03-09-2004, 02:29 PM
I should point out that even in Elo's book when he talks about doing game by game calculations he is still batching them.

pax
03-09-2004, 02:38 PM
This isnt correct.
There is no effective reduction in the RD. The number of games that are batched effects the d squared value. The RD value of the player remains constant throught the period.


Look, I didn't say that RD actually changes. I said "effective" change, because the d^2 value enters the equation and scales the results in the same way as RD^2. (Yes, RD enters elsewhere, but this is the principal effect).



Your conclusions are incorrect and your example isnt a good one.
In the case of two equally rated players the E(s|r,rj,RDj) is always 0.5 irrespective of the RD.


The value of E(s|r,rj,RDj) is not important in my example. The point is that after a sequence of results you have more information about the rating, and that there is no reason not to use that information prior to computing the result of the next game.



In your above example after 8 games of the win/loss sequence the players rating remains 1500 and the 9th game win adds around 16.8 rating points for a rating of 1517. By adding a 10th game the players rating drops back to 1500, hence the addition of that game effective subtracts 16.8 points.
If there was an 11th game win that would only contribute 15.3 points. A 12th game loss would contribute a 15.3 point loss. As for your 10 win sequence see below.


So? I'm not talking about 10 games compared to 9, or 11 compared to 10. Within any sequence of n games (against the same player) processed in one period, the contribution of a win is the same, independently of other results.




However what if your example was a 1500! playing a 1600!! with the same scenarios.
After 8 games of the win loss sequence the players rating is 1540. The 9th win adds 20 points for a rating of 1560. The 10th game loss is an effective loss of 15 points and the players rating is 1545. An 11th game win would contribute 19 points and the players rating is now 1564. A 12th game loss would result in a 14 point loss for a rating of 1550.


And your point is?



The gain from the tenth win is less than the gain for the first 9.
If the player only played 9 games and won them all his rating gain is 151 points for an average gain of 16.8. If he played 10 and won them all then his rating gain is 160 points for an average of 16.


You are deliberately skewing my examples. Fix the number of games played in the period, then the contribution of each game is the same.



Bottom line is this if your 10 game win loss sequence was played at the rate of one game per week and the rating period is one week then the players rating after 10 weeks would be approx 1498.7 where as if the 10 games are batched the players rating would be exactly 1500.

Yes! I would put it to you that 1498.7 is a more accurate representation that 1500, since the most recent result in the sequence was a loss. If the sequence continues, then the rating will be slightly above 1500 the next week, slightly below the week after and so on.

Pax

pax
03-09-2004, 02:50 PM
Firstly it should be pointed out that the occurence of overshoot or undershoot occurs very infequently in Glicko or Elo.
In fact for all the rating periods since Dec 2000 to the present there has been 63 occurences from 20425 rating updates.

The simplest example under Elo is as follows.
A player has a rating of 1500.
He plays with 10 different players all rated 1500.
His results are 1010101010.
The players performance rating over the 10 games is 1500.
However after rating the 10 results individual the players rating at the end of the 10 games is 1499.12. This is an undershoot.
If the results had been in the order 0101010101 then his rating would have ended up at 1500.75 which represents an overshoot.

Ok that's true, but the undershoot/overshoot in this case has a valid interpretation. The point of the game by game calculations is that the sequence of results is significant. If your most recent result was a win you might reasonably expect your rating to be slightly above 1500, and vice versa. Note that if the sequence was 0000011111, the rating at the end should be higher still.

By contrast the Glicko overshoot where a 1200? player scores 50% against 1600!! and ends up with a rating of 1650 has no valid interpretation.

Pax

Bill Gletsos
03-09-2004, 02:53 PM
Look, I didn't say that RD actually changes. I said "effective" change, because the d^2 value enters the equation and scales the results in the same way as RD^2. (Yes, RD enters elsewhere, but this is the principal effect).
Yes but d^2 is based on the number of games.



The value of E(s|r,rj,RDj) is not important in my example.
I disagree.


The point is that after a sequence of results you have more information about the rating, and that there is no reason not to use that information prior to computing the result of the next game.
I dont agree.




So? I'm not talking about 10 games compared to 9, or 11 compared to 10. Within any sequence of n games (against the same player) processed in one period, the contribution of a win is the same, independently of other results.

Every game added alters the effect based on the previous results, because the algorithm camn gain more information because of the added game result it effects the total contribution.



And your point is?
That the effect on the rating and the contribution of a game changes as games are added to the pool of games rated in the period.




You are deliberately skewing my examples. Fix the number of games played in the period, then the contribution of each game is the same.
No i'm not skewing them at all.
I'm showing that the effect of adding a game to a group of games effects the rating by a different amount from the previous results.



Yes! I would put it to you that 1498.7 is a more accurate representation that 1500, since the most recent result in the sequence was a loss. If the sequence continues, then the rating will be slightly above 1500 the next week, slightly below the week after and so on.
I disagree.

pax
03-09-2004, 02:54 PM
I should point out that even in Elo's book when he talks about doing game by game calculations he is still batching them.

When I talk about game by game calculations, I am batching them. Your point?

pax
03-09-2004, 03:04 PM
I disagree.

What if the games were one a year, your rating period is ten years and the results were 0000011111 compared to 1111100000? What about one a month and ten months? One a week and ten weeks?

The length of the rating period is a subjective matter and can have significant effects on the ratings calculations. I'm simply pointing out that from a theoretical point of view, shorter is better.

The whole premise of the Glicko and Elo systems depends on the assumption that more recent games are more significant than old games. What is variable is the classification of what you call recent. Under the current system, what counts as recent depends on how close you are to the beginning of the current rating period.

Pax

PHAT
03-09-2004, 03:49 PM
Who mentioned pooling.

You did, dog's breath. You said "normal circumstances". Since "pooling" is a normal phenominae in natural/complex systems, you have included "pooling" as being handlable by Glicko. But, it doesn't does it. :P

PHAT
03-09-2004, 03:58 PM
What if the games were one a year, your rating period is ten years and the results were 0000011111 compared to 1111100000? What about one a month and ten months? One a week and ten weeks?

The length of the rating period is a subjective matter and can have significant effects on the ratings calculations. I'm simply pointing out that from a theoretical point of view, shorter is better.

The whole premise of the Glicko and Elo systems depends on the assumption that more recent games are more significant than old games. What is variable is the classification of what you call recent. Under the current system, what counts as recent depends on how close you are to the beginning of the current rating period.

Pax

Pax, the quote above say it all. You are right. BG is just being a stuborn Nigel.

Bill Gletsos
03-09-2004, 04:59 PM
When I talk about game by game calculations, I am batching them. Your point?
No you are not.
You have been suggesting that on a game by game basis the rating at the start of each game is the calculated rating from the previous game calculation.
When Elo batches them on a game by game basis the rating at the start of the batch is constant for the batch.

Bill Gletsos
03-09-2004, 05:02 PM
You did, dog's breath. You said "normal circumstances". Since "pooling" is a normal phenominae in natural/complex systems, you have included "pooling" as being handlable by Glicko. But, it doesn't does it. :P
You are a total moron.
You were the only one to use the word normal.
I never mentioned the word normal in any post.

Bill Gletsos
03-09-2004, 05:12 PM
What if the games were one a year, your rating period is ten years and the results were 0000011111 compared to 1111100000? What about one a month and ten months? One a week and ten weeks?
In these cases your c^2 value would be totally differnt for each case.
Hence if you were rating on a 10 week period and batched the 10 results or rating one game per week on a game by game basis the results are not significantly different. Ceratinly not significant enough to warrant the trouble of being concerned about the order of the games within the rating period.


The length of the rating period is a subjective matter and can have significant effects on the ratings calculations.
Is that a fact. Provide an actual worked example.


I'm simply pointing out that from a theoretical point of view, shorter is better.
Not according to Glickman. He clearly recommends batching them.


The whole premise of the Glicko and Elo systems depends on the assumption that more recent games are more significant than old games. What is variable is the classification of what you call recent. Under the current system, what counts as recent depends on how close you are to the beginning of the current rating period.
Immaterial.
The difference in rating with a single game per period or on a periodic basis with multiple games is not significantly different.
As I said it certainly doesnt warrant the extra effort involved.

PHAT
03-09-2004, 05:37 PM
You are a total moron.
You were the only one to use the word normal.
I never mentioned the word normal in any post.

"...you have to experiment with the various parameters to determine what works best in your circumstances." BG.

Are you suggesting that Australian chess is not an example of normal circumstances?

If yes, then Glicko is not the right system for us.
If no, then "normal circumstances" includes pooling - and now tell us which parameters relate to modelling the pooling phenominae?

Bill Gletsos
03-09-2004, 06:07 PM
"...you have to experiment with the various parameters to determine what works best in your circumstances." BG.

Are you suggesting that Australian chess is not an example of normal circumstances?
The word normal is irrelevant in the context.


If yes, then Glicko is not the right system for us.
If no, then "normal circumstances" includes pooling - and now tell us which parameters relate to modelling the pooling phenominae?
You are a fool and a waste of time just like DR.

pax
03-09-2004, 06:17 PM
[Pax said rating period length can have a significant effect]



Is that a fact. Provide an actual worked example.

Come on, don't pretend you don't know this is true.

Ok this example is calculated on Barry Cox's calculator, so the numbers are approximate at best. It is also based on Glicko-1, but this effect is visible in Elo Glicko-1 or Glicko-2. It was also carried out rather quickly since I'm not going to spend hours doing this stuff when it's falling on deaf ears anyway.

Player A with rating 1256 (we will experiment with RDs of !!, !,_ ,? and ??) plays 30 games against a player rated 1892! at the rate of one game per week. For the purposes of the experiment, player B has a constant rating (consider him to be 30 different players rated 1892 if you must).

(Pop quiz: who are these players? ratings officers not allowed to answer)

Player A scores 50% uniformly across the period.

Calculations have shown what happens without the artificial hack which checks against PR to prevent overshoot. It is interesting to note how often the hack is required when batch size is large.

If processed as a 30 game batch (30 week rating period):
1256!!--> 1595
1256!--> 1887
1256_--> 2304 (and must be corrected via the hack)
1256?--> 2831 (ditto)
1256??--> 3347 (ditto)


two 15 week periods:
1256!!--> 1436->1585
1256!--> 1609->1774
1256_--> 1892->1892
1256?--> 2327->1732 (hack required)
1256??--> 2869-> -247 (I kid you not - of course it only happens without the hack)

three 10 week periods:
1256!!-->1585
1256!--> 1772
1256_--> 1857
1256?--> 1899 (hack required in intermediate weeks)
1256??-->1893 (hack required in intermediate weeks)

six 5 week periods:
1256!!-->1589
1256!--> 1778
1256_--> 1871
1256?--> 1880
1256??-->1893

For 3 week periods and one week periods, the results are similar to the 5 week example, with nice smooth convergence towards the performance rating.

Pax

Cat
03-09-2004, 06:39 PM
[Pax said rating period length can have a significant effect]



Come on, don't pretend you don't know this is true.

This guy has no capacity to evaluate information objectively, rather he would take every opportunity to discredit what he can't understand. You know the flaws, I know the flaws, we all know the flaws, but Bill won't see it even when it's at the end of his nose. However, don't loose heart - suggestions have an uncanny way of resurfacing in some form or other months later when all the tension has dissipated.

Rincewind
03-09-2004, 06:41 PM
Ok this example is calculated on Barry Cox's calculator, so the numbers are approximate at best.

Thanks for the vote of confidence. ;)

Cat
03-09-2004, 06:43 PM
Thanks for the vote of confidence. ;)

Yours' is the best BJC, you know he was only referring to a scenario

Bill Gletsos
03-09-2004, 08:01 PM
[Pax said rating period length can have a significant effect]



Come on, don't pretend you don't know this is true.

Ok this example is calculated on Barry Cox's calculator, so the numbers are approximate at best. It is also based on Glicko-1, but this effect is visible in Elo Glicko-1 or Glicko-2. It was also carried out rather quickly since I'm not going to spend hours doing this stuff when it's falling on deaf ears anyway.

Player A with rating 1256 (we will experiment with RDs of !!, !,_ ,? and ??) plays 30 games against a player rated 1892! at the rate of one game per week. For the purposes of the experiment, player B has a constant rating (consider him to be 30 different players rated 1892 if you must).

(Pop quiz: who are these players? ratings officers not allowed to answer)

Player A scores 50% uniformly across the period.

Calculations have shown what happens without the artificial hack which checks against PR to prevent overshoot. It is interesting to note how often the hack is required when batch size is large.

If processed as a 30 game batch (30 week rating period):
1256!!--> 1595
1256!--> 1887
1256_--> 2304 (and must be corrected via the hack)
1256?--> 2831 (ditto)
1256??--> 3347 (ditto)


two 15 week periods:
1256!!--> 1436->1585
1256!--> 1609->1774
1256_--> 1892->1892
1256?--> 2327->1732 (hack required)
1256??--> 2869-> -247 (I kid you not - of course it only happens without the hack)

three 10 week periods:
1256!!-->1585
1256!--> 1772
1256_--> 1857
1256?--> 1899 (hack required in intermediate weeks)
1256??-->1893 (hack required in intermediate weeks)

six 5 week periods:
1256!!-->1589
1256!--> 1778
1256_--> 1871
1256?--> 1880
1256??-->1893

For 3 week periods and one week periods, the results are similar to the 5 week example, with nice smooth convergence towards the performance rating.

Pax
You cannot use Barry calculator because his c^2 value is fixed and is based on a 3 month rating period.
For a weekly rating period the c^2 would be significantly different and much smaller.

Also where you say hack required you appear to have errors.
e.g 1256?--> 2327->1732 (hack required)
this should be 1256?--> 2327 which is reduced by hack to 1892 ->1892

On top of that you have screwed up the following example:
1256??--> 2869-> -247 (I kid you not - of course it only happens without the hack)

When presented by barry's calculator with the value 2869 the new RD symbol was ? down from the original ??
However you failed change it from ?? to ? when you did the next calculation that why you got -247. If you had changed the symbol correctly you would have bben given a rating of 1252. Ofc ourse you should have applied the hack as you call it before using the the 2869 for the second 15 and replaced it by 1892 (the performance rating of the first 15 games period).

I cant be bothered checking all the other results to see if you made this mistake elsewhere but I suspect you did.

The results are pretty meaningless anyway due to the incorrect c^2 values used. Your example of picking such a highly unlikely set of results doesnt help your cause.

Bill Gletsos
03-09-2004, 08:02 PM
This guy has no capacity to evaluate information objectively, rather he would take every opportunity to discredit what he can't understand. You know the flaws, I know the flaws, we all know the flaws, but Bill won't see it even when it's at the end of his nose. However, don't loose heart - suggestions have an uncanny way of resurfacing in some form or other months later when all the tension has dissipated.
Stick with medicine DR.
You are a complete joke when it comes to ratings.

Rincewind
03-09-2004, 08:41 PM
You cannot use Barry calculator because his c^2 value is fixed and is based on a 3 month rating period.
For a weekly rating period the c^2 would be significantly different and much smaller.

This is true. I use a value of 1350 which is close to what the ACF used to use in the Glocko-1 days. This number is arrived at by examination of the data and predictiveness of the ratings in general. Basically this is the aging factor for ratings and for the same set of data should be decreased if you decrease the duration of the rating period.

Bill Gletsos
03-09-2004, 08:46 PM
This is true. I use a value of 1350 which is close to what the ACF used to use in the Glocko-1 days. This number is arrived at by examination of the data and predictiveness of the ratings in general. Basically this is the aging factor for ratings and for the same set of data should be decreased if you decrease the duration of the rating period.
Or increased if you have a period greater than the ACF 3mth period.

firegoat7
04-09-2004, 09:54 AM
I have some simple questions to offer, and some simple suggestions of the practical nature for OZ chess.

1) Is time relevent to ratings? Common sense would suggest yes, I will leave the mathematical gurus to decide if this is the case
2) If time is relevent then a rating after each game would be best, yes or no?
3) If time is not relevent then Bill has no need to defend his position, yes or no?
4) Let us presume that time is relevent. How ought the games be rated?
One possible idea would be for the ACF to adopt a standard piece of rating software, something like swiss perfect and distribute it to all the clubs. Ideally this software would have an internet updating sight controlled by the ACF.

5) Would it be easy for an arbiter to record the date and time he processed every result with a piece of autoimatic software? yes,no?
6) Would it be easy for an arbiter to upload every round to a pre-determined internet location? yes,no
7)Would this reduce the workload for the tournament,state and national rating organisers? yes,no?
8) should the acf ensure that every functioning tournament organiser has the ability to access at least one location in each state for uploading these results? yes,no?
9) Could a tournament organiser take out the fee of going to an internet cafe and uploading a set of game results to the acf? yes,no?
10) since most internet cafes are about $2 per hour would a cost of $14 be reasonable for th acf to ensure that its rating system was up to date with the latest information?
11) Is it such a bad idea to have an acf website that displays ratings that are as up to date as they could practically possibly be? yes,no?
12) While I am at it, would it be a good idea for the ACF website developers to display rating based on sperate states? (I find it annoying to look for victoria on the master list) yes, no?

Cheers FG7

Bill Gletsos
04-09-2004, 07:26 PM
I have some simple questions to offer, and some simple suggestions of the practical nature for OZ chess.

1) Is time relevent to ratings? Common sense would suggest yes, I will leave the mathematical gurus to decide if this is the case
Answer yes, but depends on a number of factors.

2) If time is relevent then a rating after each game would be best, yes or no?
No.


3) If time is not relevent then Bill has no need to defend his position, yes or no?
Time is relevant but in the ACF case is irrelevant because doing it on a more regular basis e.g. monthly, weekly, daily or game by game does not have any significant benefits. In fact they introduce significant problems.



4) Let us presume that time is relevent. How ought the games be rated?
One possible idea would be for the ACF to adopt a standard piece of rating software, something like swiss perfect and distribute it to all the clubs. Ideally this software would have an internet updating sight controlled by the ACF.

5) Would it be easy for an arbiter to record the date and time he processed every result with a piece of autoimatic software? yes,no?
6) Would it be easy for an arbiter to upload every round to a pre-determined internet location? yes,no
7)Would this reduce the workload for the tournament,state and national rating organisers? yes,no?
8) should the acf ensure that every functioning tournament organiser has the ability to access at least one location in each state for uploading these results? yes,no?
9) Could a tournament organiser take out the fee of going to an internet cafe and uploading a set of game results to the acf? yes,no?
10) since most internet cafes are about $2 per hour would a cost of $14 be reasonable for th acf to ensure that its rating system was up to date with the latest information?
11) Is it such a bad idea to have an acf website that displays ratings that are as up to date as they could practically possibly be? yes,no?
All of this is irrelevant.


12) While I am at it, would it be a good idea for the ACF website developers to display rating based on sperate states? (I find it annoying to look for victoria on the master list) yes, no?
No. After all the find function is your friend. ;)

pax
04-09-2004, 07:28 PM
Thanks for the vote of confidence. ;)

Nothing personal Barry. I think it's a great tool, and fine for the purpose for which it was intended. Since the calculator only deals with approximate RD (!!,! etc) and not actual numbers, the numbers will be approximate. This is especially true with an experiment which attempts to calculate changes over several periods, as it is hard to model the change in RD over those periods.

Bill Gletsos
04-09-2004, 07:39 PM
Nothing personal Barry. I think it's a great tool, and fine for the purpose for which it was intended. Since the calculator only deals with approximate RD (!!,! etc) and not actual numbers, the numbers will be approximate. This is especially true with an experiment which attempts to calculate changes over several periods, as it is hard to model the change in RD over those periods.
No the fact the calculator only deals with approximate RD's and not actual numbers has absolutely nothing to do with the problem or your attempt to use it.

Barry's calculator is great for what it was designed for, which is estimating ratings based on the Glicko1 implmentation of the ACF system.

It is useless if you want to experiment with changing things like the size of the rating period as you cannot manipulate the c^2 value.

Bill Gletsos
04-09-2004, 07:42 PM
Ok, here as far as I am concerned is the final word on the subject.

As far as batching or not batching is concerned Glickman makes it clear that you should batch.

In fact everyone batches ratings with the exception of the online servers.
For online serevrs where players can play an excessive number of games in a 24 hr period doing it on a game by game basis is ok, although it would still be better if they batched them so that ratings were updated every hour or just once a day.

Rating game by game as a single period has a number of problems, amongst them being:
1) It doesnt handle unrated/new players in anything like a satisfactory manner whatsoever.
2) Rapidly improving/declining players lag.

There is no significant difference for the majority of players (those not rapidly improving or declining) whatsoever between the ACf 3 mth period and a shorter period.

As an example a 1500!! RD=60 player who plays 24 different 1600! RD=95 players and gets a uniform 50% score. (I picked 24 games because it divides easily in a 3 month, 1 month and 1 week rating period).

Using Glicko1 but with the the current ACF 3 mth period (24 games in the period) the 1500!! ends up with a 1548 rating and a new RD of 52.
Under a monthly period (8 games per period) the rating would be 1546 and a RD of 54.
Under a weekly rating system (2 games per period) the rating would be 1545 and a RD of 54.
Under a half weekly rating system (1 games per period) the rating would be 1544 and a RD of 55.

However lets look at the situation where you have monthly rating periods but the player plays 12 games in the first and 12 in the second and none in the 3rd.
How does this compare to the situations above.
Well the new rating would be 1542 and a RD of 55.
What if the monthly games had been played in the sequence none,12,12.
The new rating would have been 1546 and a RD of 53.
What if the monthly games had been 12,6,6.
Then the new rating would have been 1545 and a RD of 54.

What if the rating period was 4mths but still only 24 games were played.
The new rating would be 1550 and a RD of 53.
What if the rating period was 4mths but the games played were 32 maintaining the rate of 2 games per week.
The new rating would be 1557 and a RD of 50.

What all this means is obvious.

The most accurate rating at the end of 3 mths irrespective of the size of the rating period is the one closest to 1600 since that is clearly the players consistent new strength.

For a given number of games (in this case 24) in a specific time frame (3 mths) the best system for rating is a 3mth period.
However if games are played at a specific constant rate (e.g 2 games a week every week) then the longer the period the better.

What I find annoying is that when people ask questions I give them answers and they dont believe me.
Perhaps they think that because I usually answer the question very quickly I'm guessing.
If thats the case they are pure and simply wrong.
The reason I can answer so quickly is because in virtually all cases we (Graham and I) have already tested/tried/experimented with your idea.

However what pisses me right off with you clowns is that the majority of you do no actual calculations whatsoever. :mad:
Those that make an attempt virtually always do it incorrectly and when this is pointed out still want to argue the case or even worse just ignore the facts and expound their "belief".
In the world of ratings "beliefs" mean SFA.

Of course Matt and DR are just fools so I expect them to act like it.
The rest of you (with the exception of Kevin and Barry) I expect more from.
Try not to disappoint in future.


So endeth the lesson. :hand: :hand:

pax
04-09-2004, 07:46 PM
You cannot use Barry calculator because his c^2 value is fixed and is based on a 3 month rating period.
For a weekly rating period the c^2 would be significantly different and much smaller.

Also where you say hack required you appear to have errors.
e.g 1256?--> 2327->1732 (hack required)
this should be 1256?--> 2327 which is reduced by hack to 1892 ->1892


If you actually READ my post, you would notice that i stated that I did NOT apply the hack.

I was evaluating the system, not the hack.

Clearly, every time the hack is required we simply end up with the performance rating, which isn't useful information as it doesn't tell you how badly the Glicko result is. The number of times you need to use the hack is reflection of how good the underlying system is (not the only reflection, but one).



The results are pretty meaningless anyway due to the incorrect c^2 values used. Your example of picking such a highly unlikely set of results doesnt help your cause.

The incorrect c^2 values would make a difference. How much, I couldn't say without coding up Glicko II to do a proper experiment.

As for the results I used, you well know that proper evidence for more 'likely' results would require hundreds of computations which I don't have the facility to do at present. A slightly extreme example is require to show the flaws with relatively few calculations.

I have certainly showed enough to prove the point that you queried, namely that the performance of the system is dependent on the length of the rating period. It is also clear that the performance deteriorates the longer the period.

Incidentally, on the "batching" of results suggested by Elo, this has little relevance since at the time the book was written the computations for any reasonable size system would have been totally impractical on a game by game basis. "batching" was necessary simply to make calculations possible by hand or on the computers of the time.

Bill Gletsos
04-09-2004, 08:03 PM
If you actually READ my post, you would notice that i stated that I did NOT apply the hack.
No you didnt.
You said hack required.
This could mean either the hack needs to be applied or the hack has been applied in this case.
Although its true that if you had applied the hack the rating would have equalled the performance rating in the intermediate and final steps, the fact you didnt show this could as far as I was concerned mean you didnt know what you were doing.


I was evaluating the system, not the hack.
Well your attempt was pretty poor then, given I had explained back in post #8 that Barry's calculator could not handle the necessary different c^2 values required, it was damn obvious to anyone with any understanding of the Glicko system that using Barry's calculator was a complete waste of time.
Even so you still used it after this.


If Clearly, every time the hack is required we simply end up with the performance rating, which isn't useful information as it doesn't tell you how badly the Glicko result is. The number of times you need to use the hack is reflection of how good the underlying system is (not the only reflection, but one).
Clearly you should learn to read an comprehend.
I said much earlier in this thead:
"In fact for all the rating periods since Dec 2000 to the present there has been 63 occurences from 20425 rating updates."


If The incorrect c^2 values would make a difference. How much, I couldn't say without coding up Glicko II to do a proper experiment.
I cant tell if your reference to Glicko2 is a typing error or just ignorance on your part because only Glicko1 has a c^2 value. Glicko2 does not have a c^2 value.


As for the results I used, you well know that proper evidence for more 'likely' results would require hundreds of computations which I don't have the facility to do at present. A slightly extreme example is require to show the flaws with relatively few calculations.
Incorrect.
Your inability to do an accurate test is not my problem or concern.
You want to make unsupported claims then back them up with accurate data.
Near enough definately isnt good enough.


I have certainly showed enough to prove the point that you queried, namely that the performance of the system is dependent on the length of the rating period. It is also clear that the performance deteriorates the longer the period.
Rubbish, you have shown nothing.
Your means of calculation were flawed hence your calculations were flawed.



Incidentally, on the "batching" of results suggested by Elo, this has little relevance since at the time the book was written the computations for any reasonable size system would have been totally impractical on a game by game basis. "batching" was necessary simply to make calculations possible by hand or on the computers of the time.
Impracticality had nothing to do with it.
Computers that could handle game by game have been around for many years now but no national federation does game by game calculation.
As I said above Glickman recommends batching.

pax
04-09-2004, 08:20 PM
I am likewise getting tired of this thread, so I am making my final reply and then will go beat my head against some other brick wall. I have engaged my trusty Billbot(tm) to help. ;-)



Ok, here as far as I am concerned is the final word on the subject.

<billbot>
No it's not.
</billbot>



In fact everyone batches ratings with the exception of the online servers.
For online serevrs where players can play an excessive number of games in a 24 hr period doing it on a game by game basis is ok, although it would still be better if they batched them so that ratings were updated every hour or just once a day.

<billbot>
Really? Prove it.
</billbot>


Rating game by game as a single period has a number of problems, amongst them being:
1) It doesnt handle unrated/new players in anything like a satisfactory manner whatsoever.

<billbot>
Irrelevant.
</billbot>
Any system needs a way of initializing ratings into the system. This is independent of the way you calculate ratings for establiched players.




2) Rapidly improving/declining players lag.

<billbot>
No they don't.
</billbot>


There is no significant difference for the majority of players (those not rapidly improving or declining) whatsoever between the ACf 3 mth period and a shorter period.
<example deleted>


Your example does not prove your thesis. You make a statement about the "majority" of players, and then show the results for one player. Meaningless.



What all this means is obvious.

The most accurate rating at the end of 3 mths irrespective of the size of the rating period is the one closest to 1600 since that is clearly the players consistent new strength.

<billbot>
False.
</billbot>

If this were true, then most recent performance rating would be the best rating system.



For a given number of games (in this case 24) in a specific time frame (3 mths) the best system for rating is a 3mth period.


By this logic, for 100 games in a 5 year time frame you would use a 5 year period.



However if games are played at a specific constant rate (e.g 2 games a week every week) then the longer the period the better.


<billbot>
incorrect.
</billbot>
This is only true if you assume that true strength does not change with time. It does, so you cannot.



What I find annoying is that when people ask questions I give them answers and they dont believe me.

My apologies for not worshipping the ground you walk on. I mean no disrespect to you, you clearly have much experience and knowledge of Glicko and Glicko II. This does not make you automatically right. Neither is Glickman automatically right, despite the PhD.



However what pisses me right off with you clowns is that the majority of you do no actual calculations whatsoever. :mad:


It is easy for you to claim the high ground on calculations when you are the only one with the facility to do precise calculations. The rest of us plebs don't even get to see the real RD and volatility parameters so actual calculations of actual ratings are impossible even if we went to the trouble of writing software to do the calculations. And why would we go to that much trouble when you refuse to even acknowledge the merit in a theoretical argument.



Of course Matt and DR are just fools so I expect them to act like it.
The rest of you (with the exception of Kevin and Barry) I expect more from.
Try not to disappoint in future.

So endeth the lesson. :hand: :hand:

<sweeneybot>
You're the fool. Wingnut.
</sweeneybot>
Oops, how did that slip in there *restrains Sweeneybot(tm) from further comment*.

Pax

Bill Gletsos
04-09-2004, 08:40 PM
I am likewise getting tired of this thread, so I am making my final reply and then will go beat my head against some other brick wall. I have engaged my trusty Billbot(tm) to help. ;-)
Lets hope this really is your last post on the subject.


<billbot>
No it's not.
</billbot>
I said as far as I'm concerned.
I'm no longer interested in what you believe.



<<billbot>
Really? Prove it.
</billbot>
I have no intention wasting any more of my time on this with you.


<<billbot>
Irrelevant.
</billbot>

Useless rubbish.


Any system needs a way of initializing ratings into the system. This is independent of the way you calculate ratings for establiched players.
Not necessarily true.
But I cant be bothered wasting my time with you any longer.



<billbot>
No they don't.
</billbot>
More useless rubbish.


Your example does not prove your thesis. You make a statement about the "majority" of players, and then show the results for one player. Meaningless.
Try not to demonstrate stupidity.
I gave a simple example.
It demonstrates the points at hand.



<billbot>
False.
</billbot>

Irrelevant crap.


If this were true, then most recent performance rating would be the best rating system.
.
Dont be a complete idiot. I never said the rating had to equal 1600.
It is clear the rating that is closest to 1600 is the most accurate.




By this logic, for 100 games in a 5 year time frame you would use a 5 year period.
I didnt say nor imply that. If you interpreted it that way then thats your mistake,



<billbot>
incorrect.
</billbot>
This is only true if you assume that true strength does not change with time. It does, so you cannot.
Not true at all.
In fact you are making claims based on no actual calculations.
Unlike you I have done the calculations.



My apologies for not worshipping the ground you walk on. I mean no disrespect to you, you clearly have much experience and knowledge of Glicko and Glicko II. This does not make you automatically right. Neither is Glickman automatically right, despite the PhD.
Well when it comes to ratings and especially the workings of the Glicko/Glicko2 systems my belief in whether Glickman or some bb poster is correct is a no contest.



It is easy for you to claim the high ground on calculations when you are the only one with the facility to do precise calculations. The rest of us plebs don't even get to see the real RD and volatility parameters so actual calculations of actual ratings are impossible even if we went to the trouble of writing software to do the calculations. And why would we go to that much trouble when you refuse to even acknowledge the merit in a theoretical argument.
You asked a question.
I gave the answer based on actual previous testing/calculations.
The onus is on you to do the actual calculations.
You dont need to know the actual RD values.
For example with my 1500 RD=60, 1600 RD=95 I just picked those ratings and RD's as representative examples. You could easily have done the same.
In fact there was nothing stopping you from asking me to suggest some ratings/rd values. You didnt.

pax
04-09-2004, 09:11 PM
Useless rubbish.


And just as useful as your comments in the same manner.



Try not to demonstrate stupidity.
I gave a simple example.
It demonstrates the points at hand.

It proves one thing about one player. It says nothing about the majority of players.




Dont be a complete idiot. I never said the rating had to equal 1600.
It is clear the rating that is closest to 1600 is the most accurate.

And the rating that is closest to 1600 is the performance rating. Who's the idiot?

The rating closest to 1600 is blatantly NOT necessarily the most accurate, since the player has a history of performing at 1500. Is that history meaningless? Certainly not. There is an expectation that the performance will at some point drop back below 1600 based on history. Therefore the correct rating is somewhere between 1500 and 1600. Exactly where is impossible to know in the real case. To test this sort of thing properly, you would have to do a proper simulation with simulated players based on a statistical process - in that situation you can designate a (known) true strength and then evaluate your system for how fast the rating converges to the true strength. I bet not even you have done that test.



In fact you are making claims based on no actual calculations.
Unlike you I have done the calculations.


It's not my faullt you are incapable of discussing a theoretical point. If I were to start making serious propositions to the ACF I would of course do comprehensive calculations. Excuse me for not wanting to spend days coding up the rating system for the sake of proving you wrong. For all you cry of 'calculations' your own calculations haven't proved anything.

Pax

Bill Gletsos
04-09-2004, 09:31 PM
It proves one thing about one player. It says nothing about the majority of players.
I used it as an example. I'm clearly not going to post 2000 examples.



And the rating that is closest to 1600 is the performance rating. Who's the idiot?
Perhaps I should have been clearer. As you say the see below.


The rating closest to 1600 is blatantly NOT necessarily the most accurate, since the player has a history of performing at 1500. Is that history meaningless? Certainly not.
No one said the history was meaningless.

There is an expectation that the performance will at some point drop back below 1600 based on history.
Incorrect.
You have no way of determining this.
In fact given the consistent perfomance of 1600 over 24 games it is more likley to remain than decline.


Therefore the correct rating is somewhere between 1500 and 1600. Exactly where is impossible to know in the real case.
Actually the ACF system effectively puts the players rating between 1496-1600. The monthly period between 1492-1600. The weekly period between 1491-1599. The half weekly period between 1489-1599.

Now you claim the rating should be between 1500-1600 and I agree.
For the 24 games in the 3mth time frame it is blatantly clear that the 1496-1600 is the best estimate of the players rating, which is the 3mth period.


To test this sort of thing properly, you would have to do a proper simulation with simulated players based on a statistical process - in that situation you can designate a (known) true strength and then evaluate your system for how fast the rating converges to the true strength. I bet not even you have done that test.
Then you would lose that bet.
One of the very first tests Graham and I did was doing exactly this sort of test on a mainframe computer (because it was vastly superior to any PC) we had access to (and still do).


It's not my faullt you are incapable of discussing a theoretical point.
Your theoretical point is a waste of time. There is no significant difference in doing game by game as opposed to 3 mthly batches.


If I were to start making serious propositions to the ACF I would of course do comprehensive calculations. Excuse me for not wanting to spend days coding up the rating system for the sake of proving you wrong. For all you cry of 'calculations' your own calculations haven't proved anything.

You proved nothing with your flawed calculations.

Cat
04-09-2004, 11:07 PM
The most accurate rating at the end of 3 mths irrespective of the size of the rating period is the one closest to 1600 since that is clearly the players consistent new strength.


Its interesting you should look at it like that, because that leads to another Glicko anomalie, which is clear using Glicko 1, but of course we're all in the dark with G2;

Joe is a determined junior with a very reliable rating of 1000. His Dad, Dave, joins the club and over a 3 month period, playing individuals similarly with very reliable ratings, they both demonstrate a rating performance for that period of 1600.

In fact they play just over 20 games and Dave gets his first ACF rating 1600!!, very reliable.

3 questions for you Bill;

1. What will young Joe's rating be?
2. What rating performance would have delivered the same rating as his Dad?
3. If he continues to average 1600, when will his rating reach 1600 if he plays an average 30 games a year, if it ever does?

Bill Gletsos
05-09-2004, 12:36 AM
Its interesting you should look at it like that, because that leads to another Glicko anomalie, which is clear using Glicko 1, but of course we're all in the dark with G2;

Joe is a determined junior with a very reliable rating of 1000. His Dad, Dave, joins the club and over a 3 month period, playing individuals similarly with very reliable ratings, they both demonstrate a rating performance for that period of 1600.

In fact they play just over 20 games and Dave gets his first ACF rating 1600!!, very reliable.

3 questions for you Bill;

1. What will young Joe's rating be?
2. What rating performance would have delivered the same rating as his Dad?
3. If he continues to average 1600, when will his rating reach 1600 if he plays an average 30 games a year, if it ever does?
You are a fool.
In my example that you quoted the perfomace increase was 100 points above the players rating.
No player is going to have an actual improvement of 600 points in a 3mth period.

Garvinator
05-09-2004, 01:00 AM
No player is going to have an actual improvement of 600 points in a 3mth period.
thats a big call ;)

pax
05-09-2004, 02:43 PM
thats a big call ;)

Even if there are players with 600 point improvements in three months, I don't want a rating system that reflects it immediately. The same rating system would have wildly fluctuating ratings for normal players who have a good tournament here, a bad tournament there.

Pax

Alan Shore
05-09-2004, 02:49 PM
Even if there are players with 600 point improvements in three months, I don't want a rating system that reflects it immediately. The same rating system would have wildly fluctuating ratings for normal players who have a good tournament here, a bad tournament there.

Pax

At least it reflects the most recent play. It depends upon the definition of what you want 'rating' to constitute, an indication of recent performance strength or a more dynamic indication of how you've performed in the past. I don't mind the system currently in employ yet having a 'quick update' system of rating wouldn't necessarily be such a shocking thing.

Cat
05-09-2004, 04:29 PM
In my example that you quoted the perfomace increase was 100 points above the players rating.
No player is going to have an actual improvement of 600 points in a 3mth period.


Oh no, what about Moulten Ly? He certainly came pretty close! I remember when my rating was 1240 I performed at over 1600 in the Gold Coast Open and it's not fallen much short of that since. Anyway, the example I gave was theoretical to illustrate a point.

I'll let you know that according to Barry's brilliant calculator Joe's rating moved to 1236, and of course Dave's rating 1600. After that they both describe identical performances, each demonstrating a 50 point increase per year over the subsequent years, each playing very reliably rated player and each playing 20 games / yr. For convenience I batched the results. The rating changes are as follows;

-------------------Joe-------------------------------------Dave

Rating------------1236-------------------------------------1600

Yr 1 (perf 1650)---1418-------------------------------------1622

Yr2 (perf 1700)----1545-------------------------------------1656

Yr3 (perf 1750)----1637-------------------------------------1698


So it takes young Joe 3 years to reach a rating of 1600+, even though he had been playing consistently above it for the preceding 3yrs.

The point of the example is to raise the question, when is history of value, when should it be disgarded? Clearly Dave played enough games to have his rating described as very reliable from the initial 20 games he played. Joe however, was bogged down by his history even though he was playing enough games for his rating performance to be described as very reliable, should he have been a virgin.

And what about the fairness of all this? Joe, a stalwart at the club penalised for his dedication, while his Dad, off down the pub, gets a free hand. The trouble is, Joe was about 10 at the time his Dad joined the club, and though he could see he was doing as well as his Dad, felt a certain amount of injustice about the situation. In fact, yr 3 never happened, he joined up to Auskick and never looked back. AFL loved him!!

Bill Gletsos
05-09-2004, 06:28 PM
Even if there are players with 600 point improvements in three months, I don't want a rating system that reflects it immediately. The same rating system would have wildly fluctuating ratings for normal players who have a good tournament here, a bad tournament there.
FWIW I agree with you.

Bill Gletsos
05-09-2004, 06:30 PM
At least it reflects the most recent play.
Although the performance rating does indicate the players most recent play, it is unlikely to be an indication of "true strength".
Most players who have a significant increase in performance dont follow that up with a similar performance in the next rating period.
Therefore giving an established player a rating equal to their performance rating is in most cases going to result in overshoot or for those with bad results undershoot.
In fact testing shows that giving established players a new rating at the end of a period equal to their performance rating leads to worse predictive accuracy for the following period.



It depends upon the definition of what you want 'rating' to constitute, an indication of recent performance strength or a more dynamic indication of how you've performed in the past.
One of the functions of a rating is as a predictor of future results.

Cat
05-09-2004, 09:52 PM
Although the performance rating does indicate the players most recent play, it is unlikely to be an indication of "true strength".
Most players who have a significant increase in performance dont follow that up with a similar performance in the next rating period.

Not necessarily at all. Juniors that are improving generally continue to improve. What information does the performance of a junior 3 years carry about a junior's performance today? It is likely to reduce the accuracy of the rating, not improve it.


Therefore giving an established player a rating equal to their performance rating is in most cases going to result in overshoot or for those with bad results undershoot.

For stable, established players, maybe, but many of the 'established' in the ranks are juniors and the historical information from past performance carries little useful information about current performance. If 20 games is enough to give the virgin the grand title of being 'very reliable', why isn't that amount of information adequate to give reasonable certainty of a juniors performance, freed from the shackles of his historical immaturity?


In fact testing shows that giving established players a new rating at the end of a period equal to their performance rating leads to worse predictive accuracy for the following period.


OK out of interest, maybe you'd publish the ratings of (say) Ptachara, Toshi and Matt Sonter only including the last 3 years performances, so we can fairly judge the difference. Or maybe Moultun performance for the last 12/12?

Garvinator
05-09-2004, 10:02 PM
OK out of interest, maybe you'd publish the ratings of (say) Ptachara, Toshi and Matt Sonter only including the last 3 years performances, so we can fairly judge the difference. Or maybe Moultun performance for the last 12/12?
that would be only four or so players out of the whole rating system. You need more than just four or so players to make any general statements that affect the system as a whole.

Cat
05-09-2004, 10:11 PM
that would be only four or so players out of the whole rating system. You need more than just four or so players to make any general statements that affect the system as a whole.

What, you think we should re-vamp the entire system? Now that's just what I've been saying. Good on-yer Garvin, always ready to speak your mind!

PHAT
05-09-2004, 10:12 PM
The same rating system would have wildly fluctuating ratings for normal players who have a good tournament here, a bad tournament there.


But seriously, the term "wildly" is subjective. If players' rating did rise/fall much faster, and, the ratings were calculated monthly, would it realy matter. I think that it would be more accurite. (At this point BG says "prove it." I say to BG, it should have been tested by now, in which case he could say "tried it - it don't work." But if it ain't been tested, then in advance, I say, FO patzer.)

Bill Gletsos
05-09-2004, 10:26 PM
One man's wildly fluctuating rating is another's mildly fucuaching mating.

But seriously, the term "wildly" is subjective. If players' rating did rise/fall much faster, and, the ratings were calculated monthly, would it realy matter. I think that it would be more accurite. (At this point BG says "prove it." I say to BG, it should have been tested by now, in which case he could say "tried it - it don't work." But if it ain't been tested, then in advance, I say, FO patzer.)
I've already answered that question you fool.
Testing shows doing it monthly or weekly offers no benefit.
So follow your own advice patzer.

Bill Gletsos
05-09-2004, 10:28 PM
Oh no, what about Moulten Ly? He certainly came pretty close!
What about Ly.
As his rating history shows he never performed anywhere near 600 above his rating.


Results For Player Moulthun Ly

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Dec 2002 1165 1170 5.5 12
Mar 2003 1171 1144 8.0 16
Jun 2003 1282 1495 9.0 13
Sep 2003 1461 1742 7.5 13
Dec 2003 1647 1769 19.5 34
Mar 2004 1811 1906 17.5 28
Jun 2004 1856 1891 19.5 32
Sep 2004 1894 1980 14.5 22


I remember when my rating was 1240 I performed at over 1600 in the Gold Coast Open and it's not fallen much short of that since.

Ha ha ha.
Your deluding yourself.
Then again that is par for the course.


Results For Player David Richards

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Aug 2000 Unr 4.5 8
Dec 2000 1243 1477 3.0 4
Apr 2001 1372 1428 9.5 16
Aug 2001 1471 1584 9.0 14
Dec 2001 1471 1431 4.0 7
Aug 2002 1508 1728 3.5 4
Dec 2002 1556 1637 8.0 11
Jun 2003 1590 1670 3.0 9
Nov 2003 1687 0.0 0
Mar 2004 1775 2.0 2
Sep 2004 1792 1862 8.5 10


Anyway, the example I gave was theoretical to illustrate a point.
It was a pretty useless unrealistic example.
Anyway 20 games a year is an inadequate amount of games, especially if you are trying to justify a player is rapidly improving.
Even for a mildly improving player 20 would be inadequate.


I'll let you know that according to Barry's brilliant calculator Joe's rating moved to 1236, and of course Dave's rating 1600. After that they both describe identical performances, each demonstrating a 50 point increase per year over the subsequent years, each playing very reliably rated player and each playing 20 games / yr. For convenience I batched the results. The rating changes are as follows;

-------------------Joe-------------------------------------Dave

Rating------------1236-------------------------------------1600

Yr 1 (perf 1650)---1418-------------------------------------1622

Yr2 (perf 1700)----1545-------------------------------------1656

Yr3 (perf 1750)----1637-------------------------------------1698


So it takes young Joe 3 years to reach a rating of 1600+, even though he had been playing consistently above it for the preceding 3yrs.
So what?
A 600 point improvement in a period is totally unrealistic.


The point of the example is to raise the question, when is history of value, when should it be disgarded? Clearly Dave played enough games to have his rating described as very reliable from the initial 20 games he played. Joe however, was bogged down by his history even though he was playing enough games for his rating performance to be described as very reliable, should he have been a virgin.
Actually the point of your example is to show a totally unrealistic scenario in an attempt to justify your view.
You are a waste of time as usual.

PHAT
05-09-2004, 10:30 PM
Testing shows doing it monthly or weekly offers no benefit.

Quantify it. Zero or not zero. What are the predictability figures for 1, 2, ... monthly processing.

I know you know the figures. Spit 'em out.

Bill Gletsos
05-09-2004, 10:40 PM
Quantify it. Zero or not zero. What are the predictability figures for 1, 2, ... monthly processing.

I know you know the figures. Spit 'em out.
I gave an example in a previous post.

I'm not your slave or your servant.
You can whistle dixie for anything else.

PHAT
05-09-2004, 11:05 PM
I'm not your slave or your servant.


But you aaare my servant. You are my NSWCA pres and my ACF ratings officer. You are serving me. And since you ain't been paid, I reckon you are my slave as well, Mr Van Winkle.

Bill Gletsos
05-09-2004, 11:08 PM
But you aaare my servant. You are my NSWCA pres and my ACF ratings officer. You are serving me. And since you ain't been paid, I reckon you are my slave as well, Mr Van Winkle.
No schit for brains.
I'm not your servant.
I'm a volunteer.
I'm most certainly not volunteering to waste my time on fools like you and DR.

PHAT
05-09-2004, 11:17 PM
No schit for brains.
I'm not your servant.
I'm a volunteer.
I'm most certainly not volunteering to waste my time on fools like you and DR.

You are volunteeringly here, and volunteeringly wasting time replying to us. Either you are the "schit for brains" or you are our servant. The president of any public not-for-profit organisation is a servant, servant. Oi, peel me a grape.

Bill Gletsos
05-09-2004, 11:24 PM
You are volunteeringly here, and volunteeringly wasting time replying to us. Either you are the "schit for brains" or you are our servant. The president of any public not-for-profit organisation is a servant, servant. Oi, peel me a grape.
You know where you can shove your grape.

Garvinator
05-09-2004, 11:33 PM
i am getting an image of kevin grabbing his eraser real soon :lol: :whistle:

Bill Gletsos
05-09-2004, 11:48 PM
Clearly Dave played enough games to have his rating described as very reliable from the initial 20 games he played.
Just how exactly did you determine this.
Barry's calculator requires you to enter a rating for Dave but Dave was unrated. Therefore you cannot get Barry's calculator to generate a rating and know what the symbol will be.

Bill Gletsos
06-09-2004, 12:36 AM
What, you think we should re-vamp the entire system? Now that's just what I've been saying. Good on-yer Garvin, always ready to speak your mind!
Garvin never suggested that at all.
You really are a fool.

Cat
06-09-2004, 12:43 AM
[QUOTE=Bill Gletsos]What about Ly.

Results For Player Moulthun Ly

Players Historical Rating Performance Normal
Period Rating Perf Score Games

Sep 2004 1894 1980 14.5 22


So what?

The point is that 22 games with a rating performance of 1980 gives an enormous amount of information about Moulthun's current rating strength. Is it justified to use historical information which may not be of any real value in predicting his performance in the future?

In other words, which rating is more likely to depict his current playing performance, his ACF rating of 1894, or his period performance with 22 games of 1980? This is not a mathematical question Bill, it's a philosophical one. Moulthun has changed, his biochemistry is different, he is the new Moulthun!

Oh and I take your point, Moulthun took 6 months to increase his performance by 600pts, not 3 months, I stand corrected.

Cat
06-09-2004, 12:52 AM
Ha ha ha.
Your deluding yourself.
Then again that is par for the course.


Results For Player David Richards

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Aug 2000 Unr 4.5 8
Dec 2000 1243 1477 3.0 4
Apr 2001 1372 1428 9.5 16
Aug 2001 1471 1584 9.0 14
Dec 2001 1471 1431 4.0 7
Aug 2002 1508 1728 3.5 4
Dec 2002 1556 1637 8.0 11
Jun 2003 1590 1670 3.0 9
Nov 2003 1687 0.0 0
Mar 2004 1775 2.0 2
Sep 2004 1792 1862 8.5 10



Your figures don't make sense Bill. How can my rating go from 1372 to 1471 with a performance of 1428, or stay at 1471 with a performance of 1584, or then go to 1508 with a performance of 1431? Wierd this Glicko, eh?

Bill Gletsos
06-09-2004, 01:19 AM
Your figures don't make sense Bill. How can my rating go from 1372 to 1471 with a performance of 1428, or stay at 1471 with a performance of 1584, or then go to 1508 with a performance of 1431? Wierd this Glicko, eh?
Nothing weird at all, you dope.

Your Dec 2000 rating of 1243 went to 1372 in Apr 2001 based on a performance rating in Apr 2001 of 1428 based on a score of 9.5/16.

Your Apr 2001 rating of 1372 went to 1471 in Aug 2001 based on a performance rating in Aug 2001 of 1584 based on a score of 9/14.

Your Aug 2001 rating of 1471 stayed at 1471 in Dec 2001 based on a performance rating in Dec 2001 of 1431 based on a score of 4/7.

Your Dec 2001 rating of 1471 went to 1508 in Aug 2002 based on a performance rating in Aug 2002 of 1728 based on a score of 3.5/4.

Surely you can understand the rest yourself without me having to spell it out.


And you wonder why I think you are a fool. :hmm:

Bill Gletsos
06-09-2004, 01:23 AM
The point is that 22 games with a rating performance of 1980 gives an enormous amount of information about Moulthun's current rating strength. Is it justified to use historical information which may not be of any real value in predicting his performance in the future?
Of course it is.
There is no reason to believe that he will maintain that level in the next period, let alone improve.


In other words, which rating is more likely to depict his current playing performance, his ACF rating of 1894, or his period performance with 22 games of 1980? This is not a mathematical question Bill, it's a philosophical one. Moulthun has changed, his biochemistry is different, he is the new Moulthun!
No its a mathematical one.
His rating of 1894 is a better estimate of his strength than his performance rating.


Oh and I take your point, Moulthun took 6 months to increase his performance by 600pts, not 3 months, I stand corrected.
Dont be a complete fool.
He didnt go up 600 points in 6 mths.
And before you say his perfomance rating in Sept 2003 was 600 above his march 2003 thats irrelevant.
13 games in the June 2003 period and 13 in the next period is nowhere near enough games in a period to justify increasing his rating at a faster rate.

Now here are the rating histories of the other juniors you mentioned.

Here is Sonters rating history.


Results For Player Matthew Sonter

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Aug 2000 1689 8.5 14
Apr 2001 1716 6.0 6
Aug 2001 1794 1893 31.5 41
Dec 2001 1767 1556 4.0 7
Aug 2002 1735 1673 13.5 21
Dec 2002 1759 1922 5.5 7
Mar 2003 1751 1715 4.5 6
Jun 2003 1773 1796 9.0 16
Sep 2003 1760 1707 7.5 13
Nov 2003 1790 0.0 0
Dec 2003 1811 1825 12.5 21
Mar 2004 1966 2132 9.0 12
Jun 2004 1943 1886 11.0 16
Sep 2004 1953 1.0 1

He has improved one period then gone backwards in perfomance the next.

Here is Phachara's rating history.

Results For Player Phachara Wongwichit

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Aug 2000 1490 5.0 7
Dec 2000 1499 1603 4.0 7
Apr 2001 1542 1729 7.5 9
Aug 2001 1563 1579 16.5 27
Dec 2001 1561 1524 9.0 14
Apr 2002 1545 1415 4.0 6
Aug 2002 1623 1716 24.0 29
Dec 2002 1731 2111 16.5 20
Mar 2003 1694 1623 16.5 27
Jun 2003 1790 1997 21.5 29
Sep 2003 1765 1709 17.0 29
Nov 2003 1774 0.0 0
Dec 2003 1786 1798 11.5 19
Mar 2004 1844 1746 7.0 11
Jun 2004 1831 1473 6.5 7
Sep 2004 1924 2049 21.5 29

He also exhibits the same scenario as Sonter.

Here is Toshi's rating history.

Results For Player Toshi Kimura Id number = 4165245
Current Normal Rating = 1864

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Aug 2000 1345 12.0 24
Dec 2000 1364 1500 6.0 11
Apr 2001 1437 1566 13.5 24
Aug 2001 1552 1664 27.5 43
Dec 2001 1577 1630 16.0 25
Apr 2002 1662 1759 20.5 34
Aug 2002 1722 1772 31.0 49
Dec 2002 1715 1687 14.0 20
Mar 2003 1719 1704 12.0 22
Jun 2003 1712 1689 3.5 9
Sep 2003 1755 1805 18.5 29
Nov 2003 1770 0.0 0
Dec 2003 1741 1675 10.5 19
Mar 2004 1760 1609 5.5 11
Sep 2004 1864 2019 15.5 21

He like the two above shows the same scenario.


Consider this the last time I post any juniors results.

It isnt fair for you to bandy about junior names for your own ends.

Bill Gletsos
06-09-2004, 01:23 AM
Not necessarily at all. Juniors that are improving generally continue to improve.
Incorrect.
Then again you have made this statement in the past without anything to back it up.
In fact Kerry Stead said "As Bill says, just because someone is a junior does not mean that they will automatically improve, but the opposite also applies.". He also said "I don't think there can be a blanket rule for all kids - some improve, others don't; rates of improvement are different. As Bill says, you can really only measure something based on the information that you have (in much the same way as you determine ratings for adults), so a degree of 'lag' in the ratings of some players is inevitable.

You of course rabbitted on with your puberty arguments.

Improving juniors more often than not plateau for a number of periods before they again improve followed by another period of plateauing. They may even go backwards.




What information does the performance of a junior 3 years carry about a junior's performance today? It is likely to reduce the accuracy of the rating, not improve it.
Dont demonstrate you complete lack of understanding.
The historical information of 3yrs back has no real impact on a player who has been sufficiently active over most of the intervening periods.



For stable, established players, maybe, but many of the 'established' in the ranks are juniors and the historical information from past performance carries little useful information about current performance.

As usual you have no clue what you are saying.


If 20 games is enough to give the virgin the grand title of being 'very reliable', why isn't that amount of information adequate to give reasonable certainty of a juniors performance, freed from the shackles of his historical immaturity?
Simple. The new player has no information other than his current results to provide a rating with. His rating may be reliable but volatile however until you get further results in the next period you wont know if this is true or not.

In the case of a player with an existing rating you already have information about the players strength up until that point and that data should be used.

Your 600 point example is just plain unrealistic crap.



OK out of interest, maybe you'd publish the ratings of (say) Ptachara, Toshi and Matt Sonter only including the last 3 years performances, so we can fairly judge the difference.
Or maybe Moultun performance for the last 12/12?
They have been posted.

Kevin Bonham
06-09-2004, 04:33 AM
In other words, which rating is more likely to depict his current playing performance, his ACF rating of 1894, or his period performance with 22 games of 1980? This is not a mathematical question Bill, it's a philosophical one.

It is a mathematical question - the reliability of each could be empirically investigated, using a sufficiently large number of players. Indeed, the weighting of variables has been investigated by Bill to see which settings work best as predictors in general.

Using Moulthon as an example is pointless unless you can establish general principles that apply to detecting Moulthon-like players and rating them better than the current system. It would be nice if you can do this using existing data without risking major overshoots or breaching any anti-discrimination laws in the process.

Also, you haven't answered my question from a previous thread yet, so I'll put it again (though the numbers may have changed) - a player performs at 1000 over 10 games and gets a rating of 1000. In the next period he plays 10 games and performs at 1200. What should be this player's new rating and why?

As for your example with the 1000 rated junior and the 1600 rated adult, firstly, you contradict yourself - across the nation as a whole the player more likely to have a lot of past form on their record dragging them down if they suddenly improve would be an adult. The only reason this is not often seen is that adults don't so often suddenly improve.

Your calculations (if correct) show that after 80 games over three years plus one rating period, there is a gap of 61 points. Well, so there should be. If players are only playing at 20 games per year then the fact that one of them was 750 points weaker than their current performance a mere 3.25 years ago seems clearly worth 61 ratings points' difference to me.

pax
06-09-2004, 09:27 AM
At least it reflects the most recent play. It depends upon the definition of what you want 'rating' to constitute, an indication of recent performance strength or a more dynamic indication of how you've performed in the past. I don't mind the system currently in employ yet having a 'quick update' system of rating wouldn't necessarily be such a shocking thing.

The rating system as devised by Elo was always intended to be a predictor of results, not a reflector of recent form. This is also true of the Glicko systems. If the system is too dynamic, it will reflect the most recent performance rating at the expense of useful information contained in the previous rating. This can damage the predictive ability of the system. To make any attempt at predicting the trend in ratings you must use the historical information, not just the most recent performance.

It is notable with some of the improving juniors mentioned, that their performance rating fluctuates pretty dramatically in both directions. The trend is up, and that is what you want to track. Not the performance rating, which can be 2100 one period, and 1600 the next!

Note that the volatility factor of GlickoII goes some way towards speeding up the increase of an improving junior's rating.

It can be argued (and has been), that the rating system should not be simply a predictive system but also a reward system or something else. I don't fancy that path, having seen the bridge Master Point system where ancient club hackers can make GM by playing in thousands of tournaments.

Gee I can't figure this BB out. Here I was having a nice little argument with Bill, next minute I'm agreeing with him and even defending him. Sheesh ;)

Pax

Cat
06-09-2004, 11:33 AM
[QUOTE=Kevin Bonham]It is a mathematical question - the reliability of each could be empirically investigated, using a sufficiently large number of players. Indeed, the weighting of variables has been investigated by Bill to see which settings work best as predictors in general.

Yes, KB, but not all the variables affecting performance are known or modelled in the system. In fact there is signficant over-weighting of the one variable (RD) relative to it's importance, simply as a mechanism for engineering dynamic change. As such it is a limited model, it cannot account for all the compounding variables and the mathematics is extremely impure to say the least.


Using Moulthon as an example is pointless unless you can establish general principles that apply to detecting Moulthon-like players and rating them better than the current system. It would be nice if you can do this using existing data without risking major overshoots or breaching any anti-discrimination laws in the process.

As you know this can indeed be done. And as you also know, the anti-discrimination laws do not apply.


Also, you haven't answered my question from a previous thread yet, so I'll put it again (though the numbers may have changed) - a player performs at 1000 over 10 games and gets a rating of 1000. In the next period he plays 10 games and performs at 1200. What should be this player's new rating and why?

Seeing as you are being so polite, I might engage you on this my dear KB! Give me some time.


As for your example with the 1000 rated junior and the 1600 rated adult, firstly, you contradict yourself - across the nation as a whole the player more likely to have a lot of past form on their record dragging them down if they suddenly improve would be an adult. The only reason this is not often seen is that adults don't so often suddenly improve.

Yes this is true KB, but as you say adults as a population group are stable or decline slowly. I'm not saying its possible to model for every scenario, just possible to do better. As I've said in the past, adults should be treated as a stable population, generally over time the dynamic is one of slow decline (certainly after 30), so rapidly reacting rating systems aren't generally required here.


Your calculations (if correct) show that after 80 games over three years plus one rating period, there is a gap of 61 points. Well, so there should be. If players are only playing at 20 games per year then the fact that one of them was 750 points weaker than their current performance a mere 3.25 years ago seems clearly worth 61 ratings points' difference to me.

Yes, but 3 years may not seem much in your extended lifespan, but to a child making choices about his or her future, it may be the difference between continuing and quiting.

Cat
06-09-2004, 11:42 AM
The rating system as devised by Elo was always intended to be a predictor of results, not a reflector of recent form. This is also true of the Glicko systems. If the system is too dynamic, it will reflect the most recent performance rating at the expense of useful information contained in the previous rating. This can damage the predictive ability of the system. To make any attempt at predicting the trend in ratings you must use the historical information, not just the most recent performance.

It is notable with some of the improving juniors mentioned, that their performance rating fluctuates pretty dramatically in both directions. The trend is up, and that is what you want to track. Not the performance rating, which can be 2100 one period, and 1600 the next!

Note that the volatility factor of GlickoII goes some way towards speeding up the increase of an improving junior's rating.

It can be argued (and has been), that the rating system should not be simply a predictive system but also a reward system or something else. I don't fancy that path, having seen the bridge Master Point system where ancient club hackers can make GM by playing in thousands of tournaments.

Gee I can't figure this BB out. Here I was having a nice little argument with Bill, next minute I'm agreeing with him and even defending him. Sheesh ;)

Pax

Yes, all fair points. However, there is a big difference when rating established players 1500+ or 2000+, and juniors rated around 1000. In the prior instance, obviously stability is important. But in the latter case, what do the rating really reflect? The fluidity is so great that the value as a predictive system is diminished. In fact, the ratings are really only serving as a means to organise tournament, and a measure for improvement in the child's eye. There is very little accuracy, the capacity for change in the child is enormous. I think we really have to think about what the ratings are actually there for, in that group at least. Churning out ratings for the sake of ratings, without any concern about what we're trying to achieve i.e. keep kids in chess, is simply not wise.

pax
06-09-2004, 11:54 AM
Yes, but 3 years may not seem much in your extended lifespan, but to a child making choices about his or her future, it may be the difference between continuing and quiting.

I think your estimation of the effect of underrating a junior is overstated.

In my years as an underrated junior, I took extra relish in beating players rated four or five hundred points above me.

In my experience, juniors will look at their results to decide if they have been successful, not at their rating. That said, ratings for many juniors are much better under the current system than they used to be.

Pax

pax
06-09-2004, 12:02 PM
Yes, all fair points. However, there is a big difference when rating established players 1500+ or 2000+, and juniors rated around 1000. In the prior instance, obviously stability is important. But in the latter case, what do the rating really reflect? The fluidity is so great that the value as a predictive system is diminished. In fact, the ratings are really only serving as a means to organise tournament, and a measure for improvement in the child's eye. There is very little accuracy, the capacity for change in the child is enormous. I think we really have to think about what the ratings are actually there for, in that group at least. Churning out ratings for the sake of ratings, without any concern about what we're trying to achieve i.e. keep kids in chess, is simply not wise.

Sure, a child's ability is generally more volatile than an adult, but there are still identifiable trends in the performances of even the most rapidly improving juniors. I think there would have to be a substantial weight of evidence before making changes that would treat juniors differently to adults. I know many adults who have improved hugely over the space of a few years. Conversely, there are many juniors who have had very stable (or gradually improving) performances over a number of years. I'm sure many of Australia's top players would balk at seeing Zhao, Chow or Rej treated differently by the rating system.

Pax

Cat
06-09-2004, 12:28 PM
[QUOTE=pax]Sure, a child's ability is generally more volatile than an adult, but there are still identifiable trends in the performances of even the most rapidly improving juniors.

Yes, that's right. Changes in junior performance are more dynamic than adults, so why not represent it so in the modelling?




I think there would have to be a substantial weight of evidence before making changes that would treat juniors differently to adults. I know many adults who have improved hugely over the space of a few years

Pax


The evidence is there in the record, it needs the will and the desire to analyse the information. Yes adults can change quickly too, but as a population they are stable. A group of 10 year old, as a cohort, behaves differently, the population mean of the group changes over time. What I say is study the population, then derive the model.

Bill Gletsos
06-09-2004, 03:00 PM
Yes, that's right. Changes in junior performance are more dynamic than adults, so why not represent it so in the modelling?
Glicko and more so Glicko2 already handle dynamic performance changes.




The evidence is there in the record, it needs the will and the desire to analyse the information. Yes adults can change quickly too, but as a population they are stable. A group of 10 year old, as a cohort, behaves differently, the population mean of the group changes over time. What I say is study the population, then derive the model.
Where is it stated that the glicko system is based on adult populations.

Cat
06-09-2004, 03:10 PM
Glicko and more so Glicko2 already handle dynamic performance changes.

Where is it stated that the glicko system is based on adult populations.

Glicko is a fine generic system. Look, enough said, I just thought your discussion of the value of rating history deserved discussion

PHAT
06-09-2004, 03:19 PM
...so a degree of 'lag' in the ratings of some players is inevitable.

Defeatist.

PHAT
06-09-2004, 03:29 PM
[A] player performs at 1000 over 10 games and gets a rating of 1000. In the next period he plays 10 games and performs at 1200. What should be this player's new rating and why?


Without a time component being given, there is only one answer 1250.

If the subject was known to be static, then the new rating would be (1000+1200)/2 = 1100

If the subject was known to be dynamic, then the new rating would be 1400.

Since we do not know if the subject is static of dynamic, we should take a compromise course. (1100+1400)/2 = 1250

PHAT
06-09-2004, 03:33 PM
Churning out ratings for the sake of ratings, without any concern about what we're trying to achieve i.e. keep kids in chess, is simply not wise.

I agree.

pax
06-09-2004, 03:36 PM
Without a time component being given, there is only one answer 1250.

If the subject was known to be static, then the new rating would be (1000+1200)/2 = 1100

If the subject was known to be dynamic, then the new rating would be 1400.

Since we do not know if the subject is static of dynamic, we should take a compromise course. (1100+1400)/2 = 1250

Ah Matt, you're just baiting Bill now aren't you?

Bill Gletsos
06-09-2004, 03:38 PM
Without a time component being given, there is only one answer 1250.

If the subject was known to be static, then the new rating would be (1000+1200)/2 = 1100

If the subject was known to be dynamic, then the new rating would be 1400.

Since we do not know if the subject is static of dynamic, we should take a compromise course. (1100+1400)/2 = 1250
What a load of rubbish.
How do you know that the child who has just performed at 1200 for the period will improve or not in the next period.
You dont.
To therefore assume that if he is dynamic he will improve to 1400 is just plain stupid.
Of course your previous attempt at predicting future ratings was an abject failure.

Bill Gletsos
06-09-2004, 03:39 PM
Ah Matt, you're just baiting Bill now aren't you?
Possibly, but he is also stupid enough to believe what he types.

Bill Gletsos
06-09-2004, 03:49 PM
Defeatist.
Firstly that was Kerry's quote, not mine.
Of course Kerry reply to you last year "If your 'online stirrer' persona had any respect in my book, " would seem to sum things up nicely.

Your previous attempts at predicting players future ratings were a resounding failure.

PHAT
06-09-2004, 05:27 PM
How do you know that the child who has just performed at 1200 for the period will improve or not in the next period.
You dont. I don't know if he will or won't. I do know that he might or he might not.


To therefore assume that if he is dynamic he will improve to 1400 is just plain stupid.

I assumed that he could, not that he would. Assumption of a "could" is not stupid.



Of course your previous attempt at predicting future ratings was an abject failure.

In a relative way. But like mammals had to wait for the dinosaurs to die out, I will have to wait until you stop copying, before I rule the world.

Cat
06-09-2004, 06:46 PM
I think your estimation of the effect of underrating a junior is overstated.

In my years as an underrated junior, I took extra relish in beating players rated four or five hundred points above me.

In my experience, juniors will look at their results to decide if they have been successful, not at their rating. That said, ratings for many juniors are much better under the current system than they used to be.

Pax

Good points again, what impact does ratings have on the psychology of the child? Of course, direct measurement has never been done, but one can gain insight from general priniciples.

One interesting experiment was done on infants around 3/12's of age. They were given a string which when pulled changed a picture in front of them. Interestingly, boys and girls both pulled the string with the same frequency, but when the string was cut, the girls quickly moved onto something else, while the boys kept pulling the string harder and harder.

Perhaps this tells us a little about why so few girls persevere with their chess?

Chess provides a series of rewards, both immediately on the board, but also through the rating system. I'm sure early progress in the rating system is a powerful attraction to many juniors. However, the danger with creating such a powerful system of rewards is that it can lead to disillusionment when the reward is no longer forthcoming.

I appreciate what you say about your own perceptions, but you were also presumably reasonably successful. What happens is success is postponed for what appears to be an eternity? One could argue that this is the kind character building which breeds champions. However, the evidence suggests that champions perform best when roaming in fertile pastures.

Bill Gletsos
06-09-2004, 07:48 PM
I don't know if he will or won't. I do know that he might or he might not.
It doesnt matter whether its might or might not. The point is you have no reason to believe its one or the other therefore its stupid to base any decsion on it.


FO, ya rectovaginal fistula. I assumed that he could, not that he would. Assumption of a "could" is not stupid.
It doesnt matter if he could or couldnt you fool, you have no reason to believe its one or the other therefore its stupid to base any decsion on it.


In a relative way. But like mammals had to wait for the dinosaurs to die out, I will have to wait until you stop copying, before I rule the world.
You will never rule the world, because the rest would never allow it.

PHAT
06-09-2004, 08:02 PM
You will never rule the world, because the rest would never allow it.

And I s'pose being president by default born of members' apathy, makes you a working class hero.

Bill Gletsos
06-09-2004, 08:11 PM
And I s'pose being president by default born of members' apathy, makes you a working class hero.
I and the other members of the NSWCA Council all contribute, unlike you who contributed nothing during your time on Council.

Kevin Bonham
06-09-2004, 10:05 PM
Yes, KB, but not all the variables affecting performance are known or modelled in the system. In fact there is signficant over-weighting of the one variable (RD) relative to it's importance, simply as a mechanism for engineering dynamic change. As such it is a limited model, it cannot account for all the compounding variables and the mathematics is extremely impure to say the least.

Totally irrelevant - the issue was whether the system does or doesn't predict performance and the above makes no contribution to that debate, especially as you provide no evidence and no alternative.


As you know this can indeed be done.

No evidence and no alternative.


And as you also know, the anti-discrimination laws do not apply.

No evidence and no I don't know that.


As I've said in the past, adults should be treated as a stable population, generally over time the dynamic is one of slow decline (certainly after 30), so rapidly reacting rating systems aren't generally required here.

Extreme generalisation. Some adults improve rapidly, some decline rapidly, the slow decline for over-45s is a statistical average only.


Yes, but 3 years may not seem much in your extended lifespan, but to a child making choices about his or her future, it may be the difference between continuing and quiting.

No evidence that juniors drop out over ratings. IIRC your previous claim was that juniors like to be given ratings rewards in little nibbles of steady increase (rather than going up a lot then maybe down - surely overshooting then not being able to reach that level again would be more disappointing than being underrated?), but you don't seem to realise that doing that involves some juniors being underrated.

PHAT
06-09-2004, 10:13 PM
I and the other members of the NSWCA Council all contribute, unlike you who contributed nothing during your time on Council.

:boohoo:

Garvinator
06-09-2004, 10:26 PM
from reading the many, many posts on ratings ;) a point i dont think has been made or said in straight forward terms.

David, you would like juniors to be given ratings points based on their age because they are young and most juniors improve. Ok lets do that, then after we do that, their performance is shown to not match their adjusted inflated ratings and they lose more than their rating suggests they should.

This would then lead to more players 'earning' rating points that they didnt deserve.

But also David, you havent addressed as far as i can see, that on average, girls mature physically earlier than boys, so are you then saying that this factor should also be taken into account and add more rating points to girls earlier than boys as they grow up.

It seems the more you look at, the more the system would have to be set up differently for each and every player based on:

1) Age
2) Sex
3) Maturation rate
and any other factors i might have forgotten.

Im sure bill will be pleased with yours ideas :doh:

Kevin Bonham
06-09-2004, 11:11 PM
Without a time component being given, there is only one answer 1250.

You don't get a time component. All you know is that the player played the first ten games in one rating period, then the next ten in the next one. This is all you know about that player ... as is so often the case when the system has to rate players on limited information.

I don't agree with your answer and have my own, but I don't want to reveal either my answer or my reasons until David has had a shot at it.

PHAT
06-09-2004, 11:33 PM
You don't get a time component. All you know is that the player played the first ten games in one rating period, then the next ten in the next one. This is all you know about that player ... as is so often the case when the system has to rate players on limited information.

I don't agree with your answer and have my own, but I don't want to reveal either my answer or my reasons until David has had a shot at it.

Look, I have all most got a skinn full and that mkes some concepts abit slippery :) Neverthe less, you just DID give me a time componet! - the two performances were in consecutive ratting periods. That makes a diifference bacuase they coulda bin concecutive weekenders or consecutive years.

Consective weakenders would make me favor the idea that the 1000:1200 change was mor e like ly to be statistical niose .

Consective years would make me favor noise less. And logically that means that "trand" would be of a greater concideration.

I hope you aren't gonna keep moving the goal posts, gamester.

Kevin Bonham
07-09-2004, 02:38 AM
I hope you aren't gonna keep moving the goal posts, gamester.

Never moved any goal posts. My first post that you replied to said "In the next period he plays 10 games and performs at 1200." I'd assume your idea that I moved the goal posts was down to being drunk but it looks like you missed this stipulation the first time you quoted, too.

Cat
07-09-2004, 08:41 AM
Totally irrelevant - the issue was whether the system does or doesn't predict performance and the above makes no contribution to that debate, especially as you provide no evidence and no alternative.



No evidence and no alternative.



No evidence and no I don't know that.



Extreme generalisation. Some adults improve rapidly, some decline rapidly, the slow decline for over-45s is a statistical average only.



No evidence that juniors drop out over ratings. IIRC your previous claim was that juniors like to be given ratings rewards in little nibbles of steady increase (rather than going up a lot then maybe down - surely overshooting then not being able to reach that level again would be more disappointing than being underrated?), but you don't seem to realise that doing that involves some juniors being underrated.

No evidence? Well why don't you try opening your eyes?

Cat
07-09-2004, 09:13 AM
[QUOTE=ggrayggray]from reading the many, many posts on ratings ;) a point i dont think has been made or said in straight forward terms.

David, you would like juniors to be given ratings points based on their age because they are young and most juniors improve. Ok lets do that, then after we do that, their performance is shown to not match their adjusted inflated ratings and they lose more than their rating suggests they should.

This would then lead to more players 'earning' rating points that they didnt deserve.

No Garvin, what I have suggested repeatedly is skewing the variance for juniors. I've posted many times on this.


But also David, you havent addressed as far as i can see, that on average, girls mature physically earlier than boys, so are you then saying that this factor should also be taken into account and add more rating points to girls earlier than boys as they grow up.

By skewing variance Garvin, juniors only get rating improvement for improved performance. You're simply making their ratings directional and more dynamic. In essence it's not dissimilar to volatility, but it's directional and more widely applied.


It seems the more you look at, the more the system would have to be set up differently for each and every player based on:

1) Age
2) Sex
3) Maturation rate
and any other factors i might have forgotten.

Not if you read what I've said. Age yes, sex would be good to have records, but not related to anything I've suggested. It normally good governance to keep these kinds of demographics on your players.


Im sure bill will be pleased with yours ideas :doh:

I hope so!

Bill Gletsos
07-09-2004, 11:12 AM
Glicko is a fine generic system. Look, enough said, I just thought your discussion of the value of rating history deserved discussion
We havent been discussing rating history at all.
You have repeatedly said that the Glicko system is based on adult populations. I'm just asking you to back this up with an actual reference by Glickman in his Glicko paper.

Cat
07-09-2004, 12:49 PM
We havent been discussing rating history at all.
You have repeatedly said that the Glicko system is based on adult populations. I'm just asking you to back this up with an actual reference by Glickman in his Glicko paper.

What I said was that Glicko better describes what is happeing in the adult population. The system is a generic rating system, to be used in any suitable sporting requiring a ratings system. How it is applied to that sport depends always on the charateristics of the sport. The demographics and peculiarities of the sport and the country need to be taken into account. As I remember you've had e-mails yourself from Glickman saying similar things, that the particular dynamics of the Australian chess environment needs to be considered when applying the system.

Like any generic system, there will be peculiarities not anticipated, such as geographic isolation and idosyncratic demographics. Sensibly modifications should be made to account for this. This may require a little field research to identify these idiosyncratic features. For example, the greater preponderance of juniors in isoslated pockets, such as Canberra, Box Hill and The Gold Coast. Nothing in the Glicko system could anticipate these kinds of distributions. Proper application of the Glicko system requires therefore, modifications to accomodate these unanticipated peculiarities. There's nothing fundamentally wrong with the Glicko system, it's the application that's not being ideally managed. It's limitations, which I'm sure Professor Glickman would be aware of, are not being astutely scrutinised, it's simply being dumped on a population that doesn't fit neatly into the Glicko suitcase, and your bouncing your bum up and down to get the lid to close.

Bill Gletsos
07-09-2004, 01:28 PM
Now you are just changing your tune.
You had previously said:

Professor Glickman in his original paper on the Glicko system describes how the 'bedrock adult population' was essential for the proper operation of the Glciko system

This claim by you was shown to be completely false. Glickman never mentioned anything like this with regards the Glicko system.

Having failed miserably with that claim you are now trying to switch to the generic argument.
You are a waste of time.


What I said was that Glicko better describes what is happeing in the adult population.
You may have said it but as always you have offered no evidence to back it up.


The system is a generic rating system, to be used in any suitable sporting requiring a ratings system. How it is applied to that sport depends always on the charateristics of the sport. The demographics and peculiarities of the sport and the country need to be taken into account.
Rubbish.
You have no evidence of this.


As I remember you've had e-mails yourself from Glickman saying similar things, that the particular dynamics of the Australian chess environment needs to be considered when applying the system.
Dont try and draw conclusions based on things you know nothing about.


Like any generic system, there will be peculiarities not anticipated, such as geographic isolation and idosyncratic demographics. Sensibly modifications should be made to account for this. This may require a little field research to identify these idiosyncratic features. For example, the greater preponderance of juniors in isoslated pockets, such as Canberra, Box Hill and The Gold Coast.
Players ratings should be based on their results.
The Glicko system does this.


Nothing in the Glicko system could anticipate these kinds of distributions. Proper application of the Glicko system requires therefore, modifications to accomodate these unanticipated peculiarities.
Irrelevant.
A players rating should be based on their results.


There's nothing fundamentally wrong with the Glicko system, it's the application that's not being ideally managed.
As usual you have no clue.



It's limitations, which I'm sure Professor Glickman would be aware of, are not being astutely scrutinised, it's simply being dumped on a population that doesn't fit neatly into the Glicko suitcase, and your bouncing your bum up and down to get the lid to close.
The only fool bouncing his bum anywhere is you.
Juniors should be treated like any other player.
Their ratings should be based on their results.

Cat
07-09-2004, 02:05 PM
Now you are just changing your tune.
You had previously said:


This claim by you was shown to be completely false. Glickman never mentioned anything like this with regards the Glicko system.

Having failed miserably with that claim you are now trying to switch to the generic argument.
You are a waste of time.


You may have said it but as always you have offered no evidence to back it up.


Rubbish.
You have no evidence of this.


Dont try and draw conclusions based on things you know nothing about.


Players ratings should be based on their results.
The Glicko system does this.


Irrelevant.
A players rating should be based on their results.


As usual you have no clue.



The only fool bouncing his bum anywhere is you.
Juniors should be treated like any other player.
Their ratings should be based on their results.

I offered you the Olive branch and you didn't take it - you're the one stoking the fire, you asked for my opinion & I gave it. I already know you don't like my opinion, but I'd suggest if you don't like it, don't ask for it.

Bill Gletsos
07-09-2004, 02:18 PM
I offered you the Olive branch and you didn't take it - you're the one stoking the fire, you asked for my opinion & I gave it. I already know you don't like my opinion, but I'd suggest if you don't like it, don't ask for it.
I never asked for your opinion.
I asked you to back up your claim that Glickman said the Glicko system was based on adult populations.

Cat
07-09-2004, 04:10 PM
I never asked for your opinion.
I asked you to back up your claim that Glickman said the Glicko system was based on adult populations.

Oh no you didn't!

Garvinator
07-09-2004, 04:20 PM
We havent been discussing rating history at all.
You have repeatedly said that the Glicko system is based on adult populations. I'm just asking you to back this up with an actual reference by Glickman in his Glicko paper.
this is the post in question.

Bill Gletsos
07-09-2004, 04:28 PM
this is the post in question.
Correct.

Bill Gletsos
07-09-2004, 04:30 PM
Also, you haven't answered my question from a previous thread yet, so I'll put it again (though the numbers may have changed) - a player performs at 1000 over 10 games and gets a rating of 1000. In the next period he plays 10 games and performs at 1200. What should be this player's new rating and why?
I notice DR still hasnt attempted to answer this question.

bobby1972
07-09-2004, 04:31 PM
glicko is the best love it ,dont ever change it ever ,its so good,in no other system can you so quickly redeem bad results by playing a above yourself for a few games ,and i say this from experince i loose a lot of games(say 3 a rating list) to players 300 or more below me ,but if i just beat a couple of better rated players it works out so sweetly.

PHAT
07-09-2004, 05:04 PM
Like any generic system, there will be peculiarities not anticipated, such as geographic isolation and idosyncratic demographics. Sensibly modifications should be made to account for this. This may require a little field research to identify these idiosyncratic features. For example, the greater preponderance of juniors in isoslated pockets, such as Canberra, Box Hill and The Gold Coast. Nothing in the Glicko system could anticipate these kinds of distributions. Proper application of the Glicko system requires therefore, modifications to accomodate these unanticipated peculiarities. There's nothing fundamentally wrong with the Glicko system, it's the application that's not being ideally managed. It's limitations, which I'm sure Professor Glickman would be aware of, are not being astutely scrutinised, it's simply being dumped on a population that doesn't fit neatly into the Glicko suitcase, and your bouncing your bum up and down to get the lid to close.

This is 100% true. BG will not ever concede it. KB wants to publically agree, but is playing politics - the didums doesn't wan't to annoy his neighbour.

Kevin Bonham
07-09-2004, 09:06 PM
No evidence? Well why don't you try opening your eyes?

Why don't you try learning some quoting netiquette? :wall: :wall: :wall:

I will take your failure to answer my questions or to respond to my points as an admission that your posts on ratings are indeed an incoherent movable beat-up that now has no purpose other than to attempt to annoy Bill.


No Garvin, what I have suggested repeatedly is skewing the variance for juniors. I've posted many times on this.

This would cause those who had very good tournaments to jump way over their correct rating then backslide. If the psych claims you used before are relevant that would make them more likely to drop out - this is what I meant about your rubbish being incoherent.


I offered you the Olive branch and you didn't take it

Virtually every "olive branch" you have ever offered goes something like this: "Let's just agree to compromise and settle on [insert reworded statement of typical DR humbug here]". You tend to "offer" these when your position in the debate has been exposed as completely untenable.


This is 100% true. BG will not ever concede it. KB wants to publically agree, but is playing politics - the didums doesn't wan't to annoy his neighbour.

The only politics here is the wannabe trolls backing each other up even when they know that their fellow troll is wrong, just so they can pretend to have friends on the board.

If you've got methodically sound evidence that the system is failing to cope with improving juniors in the areas you have mentioned, then bring it on. However it's a bit hard to expect you to agree with a refutation when you seem to think it is correct for a player to be given a rating higher than any performance they have ever acheived. I assume that following your logic, if a junior performs at 1000 in their new tournament then 800 in the next, their rating should be 750? I don't think that David would like that at all. :rolleyes:

PHAT
07-09-2004, 11:06 PM
However it's a bit hard to expect you to agree with a refutation when you seem to think it is correct for a player to be given a rating higher than any performance they have ever acheived.

In one circumstance, yes.


I assume that following your logic, if a junior performs at 1000 in their new tournament then 800 in the next, their rating should be 750?

No, I think that Under12s with <1200 should not drop in rating points.

Ratings aren't just about mathematical purity. Pursuit of perfect ratings assumes that the rating number is more important than retaining junior players.

Given the choice between a mathematically pure rating system, and keeping kiddies happy in there chess life, I pick real flesh and blood people. To do otherwise, is to be a zealot for the primacy of the inanomate.

Garvinator
07-09-2004, 11:36 PM
No, I think that Under12s with <1200 should not drop in rating points.
a junior is more likely to stop playing if they are getting beaten regularly by players who are way too good for the junior. If you have every junior rated at least 1200, some of those juniors will end up playing in the top half against 1900. that would be a waste of time for both players and would have the junior not wanting to play quicker than any loss of rating points.

Never mind how much this would affect the rating system as a whole.

PHAT
07-09-2004, 11:45 PM
If you have every junior rated at least 1200, ...

No, that is not what I said. I said that those who are <1200 should not drop.

Kevin Bonham
07-09-2004, 11:55 PM
No, I think that Under12s with <1200 should not drop in rating points.

Ratings aren't just about mathematical purity. Pursuit of perfect ratings assumes that the rating number is more important than retaining junior players.

We're still waiting for evidence that the rating system affects that anyway.

Any less-than-utterly-hopeless junior can get lucky against a decent-strength adult now and then. Possibly a sympathy draw, or maybe just a freak queen blunder. Suppose a junior gets a start rating of 1180 from a win in their first tournament then scores zip in their next ten tournaments in a row, are you saying that their rating should stay at 1180? Could take some explaining to stronger juniors with lower ratings who can beat this junior nine games out of ten. :hmm:


Given the choice between a mathematically pure rating system, and keeping kiddies happy in there chess life, I pick real flesh and blood people.

I'd accept minor inaccuracies if they could be proven to make a serious difference to retention rates.

Garvinator
08-09-2004, 12:13 AM
No, that is not what I said. I said that those who are <1200 should not drop.
its still the same point though, they would be seeded higher than they should be and end up getting upfloated first round into a section they should not be in and face a 600 point mismatch.

Bill Gletsos
08-09-2004, 12:25 AM
I think jenni summed it up well on the ACF BB last year when she said:

The advise I have been giving them is not to get too focused on the ratings. They really only count when you are in contention for a place in the Australian Junior team. I have been encouraging the children to look more at their performance than the number that pops out every 3 months.

PHAT
08-09-2004, 12:28 AM
I'd accept minor inaccuracies if they could be proven to make a serious difference to retention rates.

Good. The only, but difficult thing to do nwo is, to "prove" that retention rates rise with a no-drop-policy. So how do we do that without running a NDP?

Somewhat of an impass when BG exercises his power to ignore the suggestion. :hmm:

PHAT
08-09-2004, 12:35 AM
I think jenni summed it up well on the ACF BB last year when she said: "The advise I have been giving [juniors] is not to get too focused on the ratings."

Why don't you ask her to "prove" that this advice is taken by juniors? Silly me, I forgot that you never question something that supports your case., ya prion.

Bill Gletsos
08-09-2004, 12:42 AM
FWIW there are only 114 players on the ACF rating list who are U12.
Of those 90 are U1200.

Bill Gletsos
08-09-2004, 12:43 AM
Why don't you ask her to "prove" that this advice is taken by juniors? Silly me, I forgot that you never question something that supports your case., ya prion.
When it comes to juniors, I'm always going to listen to jenni as opposed to a moron like you.

Cat
08-09-2004, 10:01 AM
Floorings may not be ideal, but I agree with Matt that they're certainly better than what we have at the moment. It would certainly serve to anchor the system. As for accuracy, that's an illusion, certainly for most juniors less than 1500 rating. Their performance varies so quickly that accurate rating is worthless. What we have at the moment is a rating that often reflects their performance 6-12 months earlier.

So in other words, all efforts are being put in to producing a fool's gold without any attention at all to creating something that is of genuine benefit. I 'd like to distinguish here between the higher ratings, which serve some value, and those less than (say) 1500.

Where is the proof, you say, that ratings are affecting participation? Surely the question should be, what can we do to improve participation? When the BB is full of lie bemoaning lack of involvement, falling attendance and the poor relation status of chess, when is anyone going to ask why is this happening, what can we do to improve structure?

In countries from Russia to Indonesia the streets are full of cafes with people playing chess. Why can't we do that in Australia? Well there are competing interests I here you say, it'll never happen. Well tha may be true, but we'll never know unless we start competing. Keeping the status quo is fine when things are working, I wonder what it will take for the ACF to realise change is a necessity, not a luxury.

Bill Gletsos
08-09-2004, 10:31 AM
Floorings may not be ideal, but I agree with Matt that they're certainly better than what we have at the moment.
As usual useless speculation on your part based on no evidence.


It would certainly serve to anchor the system. As for accuracy, that's an illusion, certainly for most juniors less than 1500 rating.
This is just complete rubbish. Just like your foolish tag team partner, you have absolutely no idea what the impact would be.


Their performance varies so quickly that accurate rating is worthless. What we have at the moment is a rating that often reflects their performance 6-12 months earlier.
More useless speculation based on no evidence.
If your GC juniors are lagging the answer is simple get them to play more normal rated games.



So in other words, all efforts are being put in to producing a fool's gold without any attention at all to creating something that is of genuine benefit.
The only fool around here is you.


I 'd like to distinguish here between the higher ratings, which serve some value, and those less than (say) 1500.
As usual you have no clue what you are babbling about.


Where is the proof, you say, that ratings are affecting participation? Surely the question should be, what can we do to improve participation? When the BB is full of lie bemoaning lack of involvement, falling attendance and the poor relation status of chess, when is anyone going to ask why is this happening, what can we do to improve structure?

In countries from Russia to Indonesia the streets are full of cafes with people playing chess. Why can't we do that in Australia? Well there are competing interests I here you say, it'll never happen. Well tha may be true, but we'll never know unless we start competing. Keeping the status quo is fine when things are working, I wonder what it will take for the ACF to realise change is a necessity, not a luxury.
This has absolutely nothing to do with ratings.

Kevin Bonham
09-09-2004, 05:37 AM
Floorings may not be ideal, but I agree with Matt that they're certainly better than what we have at the moment. It would certainly serve to anchor the system.

Actually the US and FIDE experiences have shown that ratings floors often have the reverse effect to anchoring - namely severe and snowballing ratings inflation.

The rest was evidence-free except for this and below which was also relevance-free:


Where is the proof, you say, that ratings are affecting participation? Surely the question should be, what can we do to improve participation?

:wall: :wall: :wall: :wall: :wall:

I've dealt with this attempt to evade the point by you before.

Unless ratings can be shown to affect participation then participation is an issue for threads other than ratings threads. I have suggested before that you start a new thread if you want to raise ideas for increasing participation that are unrelated to ratings ... instead of just paying such ideas lip service because a pretence of interest in them appears to be convenient to your ratings beat-up.

pax
09-09-2004, 09:12 AM
No, I think that Under12s with <1200 should not drop in rating points.

Matt,

That's a pretty daft idea. Ok, retaining juniors is important, but far better done by teaching them not to place too much importance on whether their rating goes up or down. I don't think it's wise to screw up the rating system for the sake of a move that may or may not keep junior in the system.

Pax

pax
09-09-2004, 09:16 AM
In countries from Russia to Indonesia the streets are full of cafes with people playing chess. Why can't we do that in Australia? Well there are competing interests I here you say, it'll never happen. Well tha may be true, but we'll never know unless we start competing. Keeping the status quo is fine when things are working, I wonder what it will take for the ACF to realise change is a necessity, not a luxury.

Ok, pop quiz.

-Does Russia have a domestic rating system?
-If so, what is it?
-What evidence is there that their rating system has anything to do with participation levels?

If Russia does have a rating system I would bet money that it is ELO based, and that juniors are given no special treatment whatsoever.

Pax

Bill Gletsos
09-09-2004, 02:04 PM
Matt,

That's a pretty daft idea.
Daft ideas are his forte. ;)

PHAT
09-09-2004, 03:16 PM
Matt,

That's a pretty daft idea.

Not daft, ambit. A totally one way trip is (i admit) daft, as is the specifying of an age limit.
Ok, retaining juniors is important, but far better done by teaching them not to place too much importance on whether their rating goes up or down. Have you tried teach that to a junior. They (and we) say that ratings aren't important, but many of us "know" they are.[qoute] I don't think it's wise to screw up the rating system for the sake of a move that may or may not keep junior in the system.[/quote] There is no doubt that it would cause some distortion. But is that the same a screwing it up? Personally, I can live with a little distortion for the chance of retaining, say, 1% of juniors. Why? Because we retain less than 1% now.

pax
09-09-2004, 03:43 PM
There is no doubt that it would cause some distortion. But is that the same a screwing it up? Personally, I can live with a little distortion for the chance of retaining, say, 1% of juniors. Why? Because we retain less than 1% now.

Fine, but there's no evidence that it would assist in retaining even one junior.

Hell, it might even put off juniors who see how crazy it is. E.g player A is consistently better than player B, but B had a good first tourney and an initial rating of 1100. Player A had a bad first tourney and an initial rating of 600. Thereafter player B stays at 1100 depsite playing uniformly badly, and A takes ages to catch up, eventually giving up in disgust. Maybe player B will give up after being taunted by the other kids for his massively inflated rating.

Juniors (especially chess players) are not stupid. They will value a system that makes sense over one with obvious flaws.

Pax

Cat
09-09-2004, 05:36 PM
Ok, pop quiz.

-Does Russia have a domestic rating system?
-If so, what is it?
-What evidence is there that their rating system has anything to do with participation levels?

If Russia does have a rating system I would bet money that it is ELO based, and that juniors are given no special treatment whatsoever.

Pax

I think for a long time there were no ratings in Russia, interesting eh? However, even today they really don't have the same problems of recruitment, competition of sporting interest, they don't need to make their sport grow. Sign of the times, first time I've ever seen a suggestion that we should model ourselves on some aspect of Russian society.

Cat
09-09-2004, 05:46 PM
Unless ratings can be shown to affect participation then participation is an issue for threads other than ratings threads. I have suggested before that you start a new thread if you want to raise ideas for increasing participation that are unrelated to ratings ... instead of just paying such ideas lip service because a pretence of interest in them appears to be convenient to your ratings beat-up.

It's simple behavioural conditioning isn't it? Offer a reward and a response is forthcoming. The more the reward is re-inforced, the more powerful the response. Remove the reward for a short time, the response continues. Remove the reward for longer, gradually the response will diminish. It's the basis of all marketing, advertising, commerical enterprise. Thats why we get freebies at shopping centres, that's why Target has a cheapie hour, Fly Buys - it's what makes the world go around.

Generally generating increased participation requires local committment, mostly it can only be done at a local level without sophisticated infra-structure such as we see with the AFL. That requires money of course. What is important at a national level is structure.

Cat
09-09-2004, 05:48 PM
Juniors (especially chess players) are not stupid. They will value a system that makes sense over one with obvious flaws.

Pax

I agree. I think Matt would too.

Bill Gletsos
09-09-2004, 06:54 PM
I think for a long time there were no ratings in Russia, interesting eh? However, even today they really don't have the same problems of recruitment, competition of sporting interest, they don't need to make their sport grow. Sign of the times, first time I've ever seen a suggestion that we should model ourselves on some aspect of Russian society.
Stop distorting peoples posts you clown.

Nowhere in pax's post did he suggest we model ourselves on any aspect of Russia.
He simply asked a few of questions and made a speculation.

Kevin Bonham
09-09-2004, 06:59 PM
It's simple behavioural conditioning isn't it? Offer a reward and a response is forthcoming. The more the reward is re-inforced, the more powerful the response. Remove the reward for a short time, the response continues. Remove the reward for longer, gradually the response will diminish. It's the basis of all marketing, advertising, commerical enterprise. Thats why we get freebies at shopping centres, that's why Target has a cheapie hour, Fly Buys - it's what makes the world go around.

This is irrelevant because you can only give a player so much reward before their rating massively exceeds their performance, unless you allow the ratings to gradually inflate (in which case people will realise the rewards are hollow). Many of your ideas would cause overshooting in junior ratings = the reward being removed for a long time after an initial over-reward, which you say is bad. If G2 does underrate some juniors mildly then while the reward they get for one success is only moderate, another success will bring another reward. That is what you say that you want. You have just justified the system you are complaining about.

Now when are you going to answer (and I mean answer seriously, not some stupid pseudo-funny question-dodge) my question about rating a player who performs at 1000 over 10 games in one rating period then 1200 over 10 games in the next? If you think you have any alternative to G2 then it should be trivially easy for you to answer the question. My guess is you'll keep dodging the question because any answer you give will either be obviously ludicrous or else will demolish the rest of your case. At least Matt had a crack at it, poor as his attempt was.


Generally generating increased participation requires local committment, mostly it can only be done at a local level without sophisticated infra-structure such as we see with the AFL. That requires money of course. What is important at a national level is structure.

Yes, you just said that, go start another thread about it, it has nothing to do with ratings.

Cat
09-09-2004, 07:16 PM
[QUOTE=Kevin Bonham]This is irrelevant because you can only give a player so much reward before their rating massively exceeds their performance, unless you allow the ratings to gradually inflate (in which case people will realise the rewards are hollow). Many of your ideas would cause overshooting in junior ratings = the reward being removed for a long time after an initial over-reward, which you say is bad. If G2 does underrate some juniors mildly then while the reward they get for one success is only moderate, another success will bring another reward. That is what you say that you want. You have just justified the system you are complaining about.

Nonsense, you clearly have no idea what you're talking about, just spouting out your usual pedantic diatribe. Stick to snails, you have more in common with the invertebrates.


Now when are you going to answer (and I mean answer seriously, not some stupid pseudo-funny question-dodge) my question about rating a player who performs at 1000 over 10 games in one rating period then 1200 over 10 games in the next? If you think you have any alternative to G2 then it should be trivially easy for you to answer the question. My guess is you'll keep dodging the question because any answer you give will either be obviously ludicrous or else will demolish the rest of your case. At least Matt had a crack at it, poor as his attempt was.

You're good behaviour didn't last very long did it? no reward for you Buddy.

Bill Gletsos
09-09-2004, 07:23 PM
Now when are you going to answer (and I mean answer seriously, not some stupid pseudo-funny question-dodge) my question about rating a player who performs at 1000 over 10 games in one rating period then 1200 over 10 games in the next? If you think you have any alternative to G2 then it should be trivially easy for you to answer the question. My guess is you'll keep dodging the question because any answer you give will either be obviously ludicrous or else will demolish the rest of your case.
Given all his crap about ratings you would think he could answer it off the top of his head.
It is fairly straighforward.

DR obviously just cannot bring himself to answer the question.

Kevin Bonham
09-09-2004, 07:26 PM
David, that was all unsubstantiated puerile rubbish. Thank you for forfeiting the debate again.

PHAT
09-09-2004, 10:52 PM
At least Matt had a crack at it, poor as his attempt was.


My crack was a cracker, knocker. What I showed was simple and justified - it was the epitomy of elegance. Now PM me your poor attempt.

Bill Gletsos
09-09-2004, 10:56 PM
My crack was a cracker, knocker. What I showed was simple and justified - it was the epitomy of elegance. Now PM me your poor attempt.
Your response was total rubbish.
For one thing you gave no justification for your dynamic figure of 1400.
Understandable since it cannot be justified.

PHAT
09-09-2004, 11:58 PM
Your response was total rubbish.
For one thing you gave no justification for your dynamic figure of 1400.
Understandable since it cannot be justified.

Hey, cat lover, what is the next number in the following series?

31, 37, ...

Bill Gletsos
10-09-2004, 12:00 AM
Hey, cat lover, what is the next number in the following series?

31, 37, ...
Yes I'm quite fond of pussies.

As for your question it could easily be 31. :owned:

Irrespective that is immaterial to the question at hand.

PHAT
10-09-2004, 12:10 AM
As for your question it could easily be 31. :owned:

Irrespective that is immaterial to the question at hand.

Not so. The next number could be 31, but, if you had to bet $1000 on any number between 1 and 100, which number would you pick?

PHAT
10-09-2004, 12:13 AM
Yes I'm quite fond of pussies.


And you have the hutzpah to call me crude. :rolleyes:

Bill Gletsos
10-09-2004, 12:15 AM
Not so. The next number could be 31, but, if you had to bet $1000 on any number between 1 and 100, which number would you pick?
The question has insufficient information.
Based on there being no reason why 37 is following 31 then any number between 1 to 100 is equally as likely.
That why in IQ tests they always give a minimum of 3 numbers.

Bill Gletsos
10-09-2004, 12:17 AM
And you have the hutzpah to call me crude. :rolleyes:
No, I have two cats.

Alan Shore
10-09-2004, 12:23 AM
Not so. The next number could be 31, but, if you had to bet $1000 on any number between 1 and 100, which number would you pick?

I think I'm missing some kind of hidden implication here...

Anyway.. the first thing that came to mind? 43. However I would probably bet on 41.. the next prime number in the sequence.

Bill Gletsos
10-09-2004, 12:28 AM
I think I'm missing some kind of hidden implication here...

Anyway.. the first thing that came to mind? 43. However I would probably bet on 41.. the next prime number in the sequence.
Matt was trying to make me suggest the next number was 43 except he posed the question extremely poorly.

As such given it went up by 6 it could just as easily go down by 6.
Likewise your answer is just as valid as would virtually any other number given as the answer.

PHAT
10-09-2004, 12:31 AM
The question has insufficient information.
Based on there being no reason why 37 is following 31 then any number between 1 to 100 is equally as likely.
That why in IQ tests they always give a minimum of 3 numbers.

You are such a drongo :hand:

Imagine that an inteligent person was given two numbers, and was told, "These numbers are data from the natural world. Now pick the third number." That person would say FMD, not much to go on, eh! They would think about it and most likely pick a number in the range 31 to 43.

If you don't agree with me on this point, I am afraid there is no hope for you.

Bill Gletsos
10-09-2004, 12:34 AM
You are such a drongo :hand:

Imagine that an inteligent person was given two numbers, and was told, "These numbers are data from the natural world. Now pick the third number." That person would say FMD, not much to go on, eh! They would think about it and most likely pick a number in the range 31 to 43.
If they did then they really arent intelligent at all.


If you don't agree with me on this point, I am afraid there is no hope for you.
I dont agree, but the only one there is no hope for is you.
Those first two numbers could be two temperature readings. For all anyone knows 37 may have been the daily maximum.

PHAT
10-09-2004, 12:39 AM
I think I'm missing some kind of hidden implication here...

Anyway.. the first thing that came to mind? 43. However I would probably bet on 41.. the next prime number in the sequence.

Actually, I picked primes because there is no obvious pattern to them. I just wanted a pair of values without the asthetic baggage of factors.

PHAT
10-09-2004, 12:44 AM
Those first two numbers could be two temperature readings.


Exactly right! Just when I though that you were a total basket case, you go and utter a phrase that fuels hope.

Now, pick the next number.

Kevin Bonham
10-09-2004, 12:44 AM
My crack was a cracker, knocker. What I showed was simple and justified - it was the epitomy of elegance. Now PM me your poor attempt.

I'm not going to reveal it to you - not even offline - because I want to see if David can come up with anything without the help of any good answers.

Bill Gletsos
10-09-2004, 12:49 AM
Exactly right! Just when I though that you were a total basket case, you go and utter a phrase that fuels hope.

Now, pick the next number.
There is no correct next number to pick.

PHAT
10-09-2004, 01:01 AM
There is no correct next number to pick.

But there is a next number, what do you guess?

(That glimmer of hope is fading.)

Bill Gletsos
10-09-2004, 01:12 AM
But there is a next number, what do you guess?

(That glimmer of hope is fading.)
31 is as good as any.

PHAT
10-09-2004, 12:28 PM
31 is as good as any.

[A candle flickers and snuffs out]

Bill Gletsos
10-09-2004, 12:31 PM
[A candle flickers and snuffs out]
For you to suggest there is only one suitable number and that it must be greater than 37 lacks any logic.

PHAT
10-09-2004, 12:40 PM
For you to suggest there is only one suitable number and that it must be greater than 37 lacks any logic.

I did NOT say that it MUST be any particular number.


You have proved to the BB that overall, you have abnormal though processes. While most of your thinking is OK, some of it is profoundly subnormal.

Bill Gletsos
10-09-2004, 12:44 PM
I did NOT say that it MUST be any particular number.
You implied it because of your comment with respect to my selection of 31.
Based on your crappy question any number is valid.
The problem is you wanted a number greater than 37, preferably 43 bu 41 would have served your purpose, which was nothing more than to try and justify your 1400 response to kevins question with reference to your undefined "dynamic".



You have proved to the BB that overall, you have abnormal though processes. While most of your thinking is OK, some of it is profoundly subnormal.
This assessment coming from the likes of you means stuff all.

PHAT
10-09-2004, 02:58 PM
Bill, the senario is: You have to pick a number in the range 31 to 43. If you pick the wrong number, you have to drive to MtBuller with me in December. Pick the correct number and your fairy godmother wiil restore you brain to that of a normal person.

However, to help you choose, you ar given some clues. It reads, "The number you choose will be the 3rd in a series, of which, you only know two consecutive numbers, 31 and 37. These two numbers and the 3rd, are data points from the natural world."

Bill, what number do you guess?

Bill Gletsos
10-09-2004, 03:06 PM
Bill, the senario is: You have to pick a number in the range 31 to 43. If you pick the wrong number, you have to drive to MtBuller with me in December. Pick the correct number and your fairy godmother wiil restore you brain to that of a normal person.
There was nothoing in your original question that required the answer to be between 31 and 43. In fact based on your follwoing statement there is still nothing that makes that a given.
As for Mt Buller, I dont have to do any such thing.


However, to help you choose, you ar given some clues. It reads, "The number you choose will be the 3rd in a series, of which, you only know two consecutive numbers, 31 and 37. These two numbers and the 3rd, are data points from the natural world."
As pointed out previously the third number could be anything.


Bill, what number do you guess?
I'm not playing your stupid little game any longer.
As I pointed out above you already have a preconceived number. your a waste of time as usual.

Why dont you actually just answer kevins question reagrding your 1400 if the person was dynamic.
Define dynamic.
Justify the 1400 value with some actual maths.
Anything else just shows you are pulling the number out of thin air.

Garvinator
10-09-2004, 03:11 PM
time for this thread to also get the delete or move treatment too i think :doh:

Bill Gletsos
10-09-2004, 03:16 PM
time for this thread to also get the delete or move treatment too i think :doh:
Nope because it is just barely ratings related.
Matt is trying to draw an anaolgy.
In reality a pretty poor attempt at a forced predetermined result by analogy to justify his unexplained 1400 number.

rob
10-09-2004, 03:25 PM
I am interested in anyones opinion on the three questions below:

Do you consider that juniors generally are more volatile in strength than seniors?

If so, then does glicko2 adequately reflect this quickly enough in its calculations (for both the juniors sakes and future opponents)?

Would treating reliably rated juniors as RD! rather than RD!! be of any benefit (the juniors rating would change by more than their opponents)?

I believe that the accuracy of all players ratings has been improving :) , I would like to hear other ppl's perceptions (compared to other states we don't have many juniors playing rated games in WA).

Bill Gletsos
10-09-2004, 03:33 PM
I am interested in anyones opinion on the three questions below:

Do you consider that juniors generally are more volatile in strength than seniors?
I would suspect that "generally" they are. However their volatility can be determined from their actual results.


If so, then does glicko2 adequately reflect this quickly enough in its calculations (for both the juniors sakes and future opponents)?
I would say yes, provided the juniors are playing around 15 games a rating period. Anyhing less and you cannot really determine if thay are volatile or just having a good or bad run.


Would treating reliably rated juniors as RD! rather than RD!! be of any benefit (the juniors rating would change by more than their opponents)?
I would argue no. Their RD should be based on the calculations which is dependant on their opponents rating and their results and whether tehy are volatile or not..


I believe that the accuracy of all players ratings has been improving :) , I would like to hear other ppl's perceptions (compared to other states we don't have many juniors playing rated games in WA).
Some of those perceptions would just be wrong based on the recent changes implemented in the September ratings.

PHAT
10-09-2004, 03:34 PM
time for this thread to also get the delete or move treatment too i think :doh:

Don't be silly. This is exactly on topic. The whole ratings system is designed around fundamental assumptions. Right now, BG is demonstrating that he does not know what mathamatical modelling is, how it works, or why the world is so complicated.

PHAT
10-09-2004, 03:45 PM
I would suspect that "generally" they are. However their volatility can be determined from their actual results.

:eek: You don't know. FMD, you are the bloke with the figures. You are the bloke who says juniors are no different to seniors. And yet, you now say that you don't know if juniors are more volatile than seniors. FMD FMD FMD.



Anyhing less [than 15] and you cannot really determine if thay are volatile or just having a good or bad run.
...
[Juniors'] RD should be based on the calculations which is dependant on ... their results and whether tehy are volatile or not.


From this we can say that a junior with less than 15 games per period cannot have his RD determined. Therefore, I say, no junior (or senior) should have a !! RD if they have not played 15+ games in the previous period.

Go on, call me a moron. But what I have just said is perfectly consistant with what you have said.

Bill Gletsos
10-09-2004, 03:51 PM
:eek: You don't know. FMD, you are the bloke with the figures. You are the bloke who says juniors are no different to seniors. And yet, you now say that you don't know if juniors are more volatile than seniors. FMD FMD FMD.
Dont be an idiot.
Oh silly me, you cant help but be that.
I took it that rob was referring to the "perception".
Hence my answer that I suspect that is generally the case.
Of course perceptions or considerations are subjective.

Thats why I then stated that you can of course determine it from their results.


From this we can say that a junior with less than 15 games per period cannot have his RD determined.
From this we can say you dont know what you are talking about as usual.


Therefore, I say, no junior (or senior) should have a !! RD if they have not played 15+ games in the previous period.
Who cares what you say, you are a fool who clearly does not understand the functioning of the glicko systems.


Go on, call me a moron.
If the shoe fits. :owned:


But what I have just said is perfectly consistant with what you have said.
No, its a total distortion.

Its complete crap like this that makes me seriously consider not answering anyones rating questions on the BB in future.

Bill Gletsos
10-09-2004, 03:56 PM
Don't be silly. This is exactly on topic. The whole ratings system is designed around fundamental assumptions. Right now, BG is demonstrating that he does not know what mathamatical modelling is, how it works, or why the world is so complicated.
No what it shows is that i'm too smart for you and your stupid attempts to force a given response to your weak attempt at anaolgy.

PHAT
10-09-2004, 04:03 PM
Dont be an idiot.
Oh silly me, you cant help but be that.
I took it that rob was referring to the "perception".
Hence my answer that I suspect that is generally the case.
"Perception" my eye.

rob asked a specific and simple question, "Do you consider that juniors generally are more volatile in strength than seniors?"

You answered. "I would suspect that "generally" they are."

The word "suspect" indicates that you don't actually know. DR is going to smash you to bits with this. Your tangled web of ratings deceit is slowly tieing you up.

Bill Gletsos
10-09-2004, 04:11 PM
"Perception" my eye.
Stop looking out your third eye.


rob asked a specific and simple question, "Do you consider that juniors generally are more volatile in strength than seniors?"

You answered. "I would suspect that "generally" they are."

The word "suspect" indicates that you don't actually know.
Ah so you are a mind reader now.
No wonder you and the witch doctor are tag team buddies.

You of course cannot know what I was thinking when I typed that.
If I had been a little more thorough I would have typed "I would suspect the perception is that "generally" they are."



DR is going to smash you to bits with this.

Dr would be lucky to smash his way out of a paper bag.


"Your tangled web of ratings deceit is slowly tieing you up.
What is abundantly clear is that both you and DR have absolutely no clue how the RD works in Glicko and Glicko2 and how the volatility works in Glicko2.
You are both a complete waste of time.

PHAT
10-09-2004, 04:11 PM
No what it shows is that i'm too smart for you and your stupid attempts to force a given response to your weak attempt at anaolgy.

No, "smart" is not the word I would use. You are asking for a to be allowed to forfiet after the clocks have started. You are, however, smart enough to know that you are gone for all money on this one.

Now, leadhead, what number do you pick and why? :owned:

Bill Gletsos
10-09-2004, 04:14 PM
No, "smart" is not the word I would use. You are asking for a to be allowed to forfiet after the clocks have started.
You arrived at the board to late and lost on forfeit.


You are, however, smart enough to know that you are gone for all money on this one.
the only one gone for all money here is you. You cannot justify your 1400 answer to kevin.


Now, leadhead, what number do you pick and why? :owned:
I have given you an answer twice now.

You however have not answered kevins question showing why you picked 1400 and how you calulate dit, nor have you defined your use of "dynamic".

You are a waste of time.

PHAT
10-09-2004, 04:33 PM
I have given you an answer twice now.


yes, you did say 31, twice. Looks like I'll be picking you up at ~8am for the MtBuller journey.

You however have not answered kevins question showing why you picked 1400 and how you calulated it, nor have you defined your use of "dynamic".


A natural phenominae is never fully static. All values are climbing or falling. Some are somewhere in a cycle, while otheres are going only one way. There values are dynamic.

1400 can be justified as being part of a trent in the values. 1000, 1200, 1400.

On the other hand. Natural phenominae display statisical noise to greater or lesser degrees. The 1000, 1200 data points may be for some phenominae that is almost static. Therefore, the 1000 to 1200 change is due prediminantly to statistical noise (random fluctuations.) In this case, the best pick for the next number is the midpoint. (X1+X2)/2 = 1100

However, we have no way of knowing what proportion of the +200 units difference is due to noise and how much due to dynamic reasons. In the absence of such information, I have abitrarily deemed the relative contributions at 50:50.

(1400+1100)/2 = 1250.

Bill Gletsos
10-09-2004, 04:50 PM
yes, you did say 31, twice. Looks like I'll be picking you up at ~8am for the MtBuller journey.
You will be travelling alone.


A natural phenominae is never fully static. All values are climbing or falling. Some are somewhere in a cycle, while otheres are going only one way. There values are dynamic.

1400 can be justified as being part of a trent in the values. 1000, 1200, 1400.
In chess ratings there is no way to predict a trend.
Just because a player went up 200 points in one period there is absolutely no reason to believe they will go up at all in the next. They could even fall.
You can only go by what they have shown via results, not some mythical future results.

The 1400 is totally unjustifiable.

PHAT
10-09-2004, 06:32 PM
In chess ratings there is no way to predict a trend.


This is so wrong that it must be classed as a lie. I cannot believe that you believe it.
Just because a player went up 200 points in one period there is absolutely no reason to believe they will go up at all in the next. They could even fall.
You can only go by what they have shown via results, not some mythical future results.

The 1400 is totally unjustifiable.

Here is a series of ratings. 800+/-50, 850+/-50, 900+/-50, 950+/-50, 1000+/-50, 1050+/-50. What is the next rating going to be?

You are losing it.

Bill Gletsos
10-09-2004, 06:47 PM
This is so wrong that it must be classed as a lie.
Your lack of knowledge of all things chessic is a wonder to behold.


I cannot believe that you believe it.
I've yet to see you get even close to proving it.



Here is a series of ratings. 800+/-50, 850+/-50, 900+/-50, 950+/-50, 1000+/-50, 1050+/-50. What is the next rating going to be?
You can play your silly games with yourself as much as you like.
I'm not taking any part in it.


You are losing it.
Certainly not to you.

PHAT
11-09-2004, 11:56 AM
You can only go by what they have shown via results, not some mythical future results.


What does, "only go by" mean?

Bill Gletsos
11-09-2004, 05:09 PM
What does, "only go by" mean?
Its obvious.
If you cannot work it out thats your problem.

PHAT
11-09-2004, 05:42 PM
Its obvious.
If you cannot work it out thats your problem.If you cannot answer a reasonable question why should I be reasonable with you? You are demonstrably unable to connect mathematics with the real world. Nor can you use civility when the opportunity to do so presents itself. Don't bother replying to this post - I have heard it all before.

Next time there is a knot in your shoe laces or you nick yourself shaving, remember that I said you deserve it.

Bill Gletsos
11-09-2004, 07:59 PM
If you cannot answer a reasonable question why should I be reasonable with you?
As far as I am concerned there was nothing reasonable about your question.
In fact your pathetic attempt to force a given number in your 31, 37 question demonstrates you havent got a clue.


You are demonstrably unable to connect mathematics with the real world.
You have demonstrated nothing.
You childishly attempted to force you already predetermined response.



Nor can you use civility when the opportunity to do so presents itself.
You offered no opportunity.
You have deliberately tried to steer my response to what you perceive as acceoptable answers.
I simply refuse to play that game.



Don't bother replying to this post - I have heard it all before.
You didnt really expect me to pay any attention to this now did you.
After all when you sprough your usual rubbish, I will always highlight it.


Next time there is a knot in your shoe laces or you nick yourself shaving, remember that I said you deserve it.
Next time you actaual have even a half clever idea, be thankful to the cosmos that even highly improbable events can actually happen. :owned:

antichrist
12-09-2004, 12:00 PM
Then a strange peace came over the land, lights got turned off earlier, wives received sex again, ISPs went broke.

Bill Gletsos
12-09-2004, 01:46 PM
Then a strange peace came over the land, lights got turned off earlier, wives received sex again, ISPs went broke.
Stick to the non chess threads you fool as your contribution to the chess related ones is zilch.

rob
13-09-2004, 09:36 PM
I would suspect that "generally" they are. However their volatility can be determined from their actual results.


I would say yes, provided the juniors are playing around 15 games a rating period. Anyhing less and you cannot really determine if thay are volatile or just having a good or bad run.


I would argue no. Their RD should be based on the calculations which is dependant on their opponents rating and their results and whether tehy are volatile or not..


Some of those perceptions would just be wrong based on the recent changes implemented in the September ratings.

Thanks to Bill for his response, but I'm sorry that MS played attack Bill - very creditable (not) :( , he didn't even bother giving his three answers.

However, I'm not sure that MS idea of RD! rather than RD!! for juniors playing less than 15 games in a rating period is that terrible, but the way it was presented/expressed leaves plenty to be desired.

Who dares to consider that one day an MS idea to improve Aus chess may actually have some merit, who would couragiously support him? :hmm: I'm sure he 'means well' or is he 'mean so well' or does he feel 'swell to be mean' :wall:

In reply to Bill's 2rd reply, I do think that expecting juniors to play around 15 games in a rating period (13 weeks) a bit much, especially for those in some states & territories. I'd hope that around 10-12 (two 6 round swisses or one 12 player all-play-all) would be enough. This slightly different opinion to Bill's should not be considered as a weapon for use by MS or DR.

PHAT
13-09-2004, 09:58 PM
This slightly different opinion to Bill's should not be considered as a weapon for use by MS or DR.

Oh yeah, sure thing, :rolleyes: You have crossed the rubicon. You are persona nongrata and your black soul will burn in hell for all eternity. Welcome to the dark side :evil: :evil: :evil:

rob, I can see you are a bit of a pragmatist, so you will have to get used to doing :wall: a lot when it comes to talking with BG. Don't worry too much about the pain. Eventually you will realise that it isn't actually a wall, it is BGs head, and it is hurting him just as much ;)

EDIT PS The idea of !! being changed to ! for juniors has already been done by a tricky new fangled euphermism called the "Volatility Factor". Brought in after howls of protests at the sloooooow Glicko reaction to juniors who improve quickly.

Of course, BG spent 50,000 words tell anyone who would listen (and those who didn't want to) that "Glicko will handles it." :rolleyes:

Down here in the firey pits, we slowly chip away at his stupidness and eventually things happen.

Bill Gletsos
13-09-2004, 10:05 PM
However, I'm not sure that MS idea of RD! rather than RD!! for juniors playing less than 15 games in a rating period is that terrible, but the way it was presented/expressed leaves plenty to be desired.
As far as Matt's suggestion on this he really has no clue, he is just picking theories out of the air.
That said although on the surface this might appear to have a benefit of increasing a juniors rating quicker the opposite is also true in that if they have a bad result their rating will plumment quicker to the same degree.
Based on testing Graham and I did ages back on an idea very similar to this I'm virtually certain the predictive accuracy goes south.


Who dares to consider that one day an MS idea to improve Aus chess may actually have some merit, who would couragiously support him? :hmm: I'm sure he 'means well' or is he 'mean so well' or does he feel 'swell to be mean' :wall:
That has about the same chance of happening as Matt's rating reaching 2000.


In reply to Bill's 2rd reply, I do think that expecting juniors to play around 15 games in a rating period (13 weeks) a bit much, especially for those in some states & territories. I'd hope that around 10-12 (two 6 round swisses or one 12 player all-play-all) would be enough.
For the static for slowly improving junior it probably isnt an issue. If the junior is rapidly improving then 12 games is insufficient.


This slightly different opinion to Bill's should not be considered as a weapon for use by MS or DR.
That chance of that happening is pretty much zero. :lol:

Bill Gletsos
13-09-2004, 10:08 PM
Oh yeah, sure thing, :rolleyes: You have crossed the rubicon. You are persona nongrata and your black soul will burn in hell for all eternity. Welcome to the dark side :evil: :evil: :evil:

rob, I can see you are a bit of a pragmatist, so you will have to get used to doing :wall: a lot when it comes to talking with BG. Don't worry too much about the pain. Eventually you will realise that it isn't actually a wall, it is BGs head, and it is hurting him just as much ;)
The trouible with you, you fool is that you have no apparent clue who rob is.
If you did you would realise he and I are much more likely to agree than disagree and even when we disagree the mutual respect remains.


EDIT PS The idea of !! being changed to ! for juniors has already been done by a tricky new fangled euphermism called the "Volatility Factor". Brought in after howls of protests at the sloooooow Glicko reaction to juniors who improve quickly.

Of course, BG spent 50,000 words tell anyone who would listen (and those who didn't want to) that "Glicko will handles it."

Down here in the firey pits, we slowly chip away at his stupidness and eventually things happen.
Just shows what you know you fool.

Firstly Glicko was far more responsive to improving juniors than Elo could ever have hoped to be.
Secondly we started working on Glicko2 which included the volatility factor as soon as Professor Glickman released his paper on it to the public domain.

PHAT
13-09-2004, 10:16 PM
The trouible with you, you fool is that you have no apparent clue who rob is.
If you did you would realise he and I are much more likely to agree than disagree and even when we disagree the mutual respect remains.

Frankly, I could give a tinker's cuss who rob is in reality. He seems OK to me, and if he is starting to see your block-o-cement head through the green and brown abuse camoflage, then all the better. :P

rob
13-09-2004, 10:17 PM
As far as Matt's suggestion on this he really has no clue, he is just picking theories out of the air.
That said although on the surface this might appear to have a benefit of increasing a juniors rating quicker the opposite is also true in that if they have a bad result their rating will plumment quicker to the same degree.
Based on testing Graham and I did ages back on an idea very similar to this I'm virtually certain the predictive accuracy goes south.:
Thanks for explaining - I thought it worthy of consideration.



That has about the same chance of happening as Matt's rating reaching 2000.

Good job you said Matt and not me :lol:


For the static for slowly improving junior it probably isnt an issue. If the junior is rapidly improving then 12 games is insufficient.
:wall: :evil: (just for Matt) :lol:


That chance of that happening is pretty much zero. :lol:
:doh:

Bill Gletsos
13-09-2004, 10:20 PM
Frankly, I could give a tinker's cuss who rob is in reality. He seems OK to me, and if he is starting to see your block-o-cement head through the green and brown abuse camoflage, then all the better. :P
He isnt seeing any of the rubbish you are sprouting you do nothing useless piece of crap.

You are so stupid you dont seem to realise he attacked you far more by his words than anything he did to me.

PHAT
13-09-2004, 10:30 PM
...useless piece of crap.

Are you sure you want to start flaming in this useless piece of crap way?

You think you are Mensa material, but you are finished with us. :lol:



I think you are the most unimaginative flamer on the www. Just shut the slit in your head and keep fibbing with your finger tips about the perfection of Glicko.

Bill Gletsos
13-09-2004, 10:42 PM
Are you sure you want to start flaming in this useless piece of crap way?
You were the one who restarted the flames in this thread with your reply to rob. I just reciprocated.


You think you are Mensa material, but you are finished with us. :lol:
All you have ever demonstrated is your complete and utter stupidty.


I think you are the most unimaginative flamer on the www.
I could not care less what a fool like you thinks.


Just shut the slit in your head and keep fibbing with your finger tips about the perfection of Glicko.
The only person consistently distorting the facts and spreading misinformation is you.

PHAT
13-09-2004, 10:57 PM
All you have ever demonstrated is your complete and utter stupidty.

Twas a criptic pun ya bloody idiot.

PHAT
13-09-2004, 11:01 PM
EDIT PS The idea of !! being changed to ! for juniors has already been done by a tricky new fangled euphermism called the "Volatility Factor". Brought in after howls of protests at the sloooooow Glicko reaction to juniors who improve quickly.

Of course, BG spent 50,000 words tell anyone who would listen (and those who didn't want to) that "Glicko will handles it." :rolleyes:

Down here in the firey pits, we slowly chip away at his stupidness and eventually things happen. ,,,

Cat
13-09-2004, 11:08 PM
In reply to Bill's 2rd reply, I do think that expecting juniors to play around 15 games in a rating period (13 weeks) a bit much, especially for those in some states & territories. I'd hope that around 10-12 (two 6 round swisses or one 12 player all-play-all) would be enough. This slightly different opinion to Bill's should not be considered as a weapon for use by MS or DR.

I don't know about you Rob, but most of our juniors play against each other. Of course they could play 10 000 games against each other and it wouldn't make a blind bit of difference to the average rating of the pool. All that would happen is their RD would reduce, so when they do finally get to play outside their pool, the dynamic effect would be reduced. So simply increasing the number of games juniors play makes no difference anyway, unless you can get them some competition from outside, unless that competition is against adults or juniors outside your club.

Bill Gletsos
13-09-2004, 11:11 PM
Twas a criptic pun ya bloody idiot.
That cryptic you fool.
And in case you were under any illusion my reply was an insult you damn fool.

Bill Gletsos
13-09-2004, 11:18 PM
EDIT PS The idea of !! being changed to ! for juniors has already been done by a tricky new fangled euphermism called the "Volatility Factor". Brought in after howls of protests at the sloooooow Glicko reaction to juniors who improve quickly.

Of course, BG spent 50,000 words tell anyone who would listen (and those who didn't want to) that "Glicko will handles it."

Down here in the firey pits, we slowly chip away at his stupidness and eventually things happen.
Just shows what you know you fool.

Firstly Glicko was far more responsive to improving huniors than Elo could ever have hoped to be.
Secondly we started working on Glicko2 which included the volatility factor as soon as Professor Glickman released his paper on it to the public domain.

Bill Gletsos
13-09-2004, 11:23 PM
I don't know about you Rob, but most of our juniors play against each other. Of course they could play 10 000 games against each other and it wouldn't make a blind bit of difference to the average rating of the pool. All that would happen is their RD would reduce, so when they do finally get to play outside their pool, the dynamic effect would be reduced. So simply increasing the number of games juniors play makes no difference anyway, unless you can get them some competition from outside, unless that competition is against adults or juniors outside your club.
As usual you havent got a damn clue what you are rabbitting on about.

All you are doing is making excuses why you arent trying to get your juniors to play more normal rated games.

Fortunately parents like ChessMum arent listening to your crap.

BTW when are you going to stop avoiding the issue and attempt to answer Kevins rating question.
The more you avoid it the more you look like the complete joke you are.

Kevin Bonham
13-09-2004, 11:25 PM
However, we have no way of knowing what proportion of the +200 units difference is due to noise and how much due to dynamic reasons. In the absence of such information, I have abitrarily deemed the relative contributions at 50:50.

(1400+1100)/2 = 1250.

In one of Martin Gardiner's books there is a cartoon of an individual locked up in what is called the "principle of indifference prison". You remind me a lot of that cartoon with these comments - arbitraily applying a 50% probability to each of two possible events instead of making some attempt to think about the likely probabilities in detail is not a good idea.

Bill Gletsos
13-09-2004, 11:34 PM
In one of Martin Gardiner's books there is a cartoon of an individual locked up in what is called the "principle of indifference prison". You remind me a lot of that cartoon with these comments - arbitraily applying a 50% probability to each of two possible events instead of making some attempt to think about the likely probabilities in detail is not a good idea.
Of course he still hasnt even remotely justified how he got his 1400 figure.
Obviously we can all see how he selected it but its selection is totally unjustified.

Garvinator
13-09-2004, 11:48 PM
BTW when are you going to stop avoiding the issue and attempt to answer Kevins rating question.
The more you avoid it the more you look like the complete joke you are.
david is getting very good at avoiding questions asked of him, i still want to know what was the result of the concerns i gave him that he said he would bring to the attention of the gold coast chess club.

Bill Gletsos
13-09-2004, 11:53 PM
david is getting very good at avoiding questions asked of him, i still want to know what was the result of the concerns i gave him that he said he would bring to the attention of the gold coast chess club.
Yes, even though he has been on the BB and seen fit to post his usual crap, I noticed he had avoided responding to your questions.

Garvinator
13-09-2004, 11:55 PM
Yes, even though he has been on the BB and seen fit to post his usual crap, I noticed he had avoided responding to your questions.
doesnt really matter, ill just speak to Greame myself this weekend and also ill ask him about David's posting as well. :whistle:

PHAT
14-09-2004, 12:02 AM
... arbitraily applying a 50% probability to each of two possible events instead of making some attempt to think about the likely probabilities in detail is not a good idea.

I know that! And what makes you so much of a snake is that you know that I know that, yet you still choose to pretend that you think I don't know that.

I was asking the Bildo what do we do when there is insufficient data on hand. BG's brain stops like a dunny roll in the s-bend when confronted with such a task. Poor bugger - perhaps he realy is abnormal.

Bill Gletsos
14-09-2004, 12:10 AM
I know that! And what makes you so much of a snake is that you know that I know that, yet you still choose to pretend that you think I don't know that.
You continually demonstrate that you know nothing.
Therefore it is natural Kevin would spell it out for you.


I was asking the Bildo what do we do when there is insufficient data on hand. BG's brain stops like a dunny roll in the s-bend when confronted with such a task. Poor bugger - perhaps he realy is abnormal.
Dont flatter yourself, you moron.

You were not asking anything, you were just trying to force your stupid predetermined answer as a response to justify your stupid 1400 figure.
Your p.ss poor attempt was visible for all to see.

Alan Shore
14-09-2004, 12:20 AM
doesnt really matter, ill just speak to Greame myself this weekend and also ill ask him about David's posting as well. :whistle:

Is that still contingent upon you getting a lift to the Classic with me though? ;)

Actually it's a really silly situation... and it's like this:

Garvin can only play if he gets a lift with me.

I only want to play if we can get a good team together. Hence I need two more players.

My mate [Alex] says he's interested but only wants to play if he gets the chance to rough up Sean Karita (my flatmate).

Sean says he'll only play if Nathan Anderson plays, since he wants him to be part of the team.

Nathan keeps saying he's retired from chess tourns (despite playing the GC Open in June) and is not likely to play.

Therefore, this breaks down the whole process and now, we have a situation extended further by Garvin not getting to chat to Graeme all because of Nathan not playing!

Causal chains are a funny thing :confused:


P.S. I would probably play regardless yet this depends on the outside factor of if I can get my assignments done in time!

Garvinator
14-09-2004, 12:27 AM
P.S. I would probably play regardless yet this depends on the outside factor of if I can get my assignments done in time!
i have some thoughts on getting a team together, but need to other players on thursday first ;) i have sent you email too.

Alan Shore
14-09-2004, 12:43 AM
i have some thoughts on getting a team together, but need to other players on thursday first ;) i have sent you email too.

Really? It hasn't arrived yet...

Garvinator
14-09-2004, 12:45 AM
Really? It hasn't arrived yet...
you sure, i sent it to the t(first letter) address.

Kevin Bonham
14-09-2004, 04:45 AM
I know that! And what makes you so much of a snake is that you know that I know that, yet you still choose to pretend that you think I don't know that.

Actually I'm too much of a snake to give you that much credit.

Well, at least I know you know it now. :lol:

rob
14-09-2004, 09:39 AM
I don't know about you Rob, but most of our juniors play against each other. .
Hi David, Unlike the Gold Coast, most of the WA juniors play ACF rated games in (senior) club and weekend tourneys. The only ACF rated junior only events these days are the state junior ch'ships and then only the U14, U16, U18 are ACF rated. All 5 age groups are 'Junior Chess Council' rated, as are any official games that juniors play in that are 20mins or more per player. In the late 1990's I ran a ACF rated junior invitation tourney each year for our stronger juniors that was held in early December as a preparation for the National juniors (this hasn't been held in recent years).



Of course they could play 10 000 games against each other and it wouldn't make a blind bit of difference to the average rating of the pool. All that would happen is their RD would reduce, so when they do finally get to play outside their pool, the dynamic effect would be reduced. So simply increasing the number of games juniors play makes no difference anyway, unless you can get them some competition from outside, unless that competition is against adults or juniors outside your club.
My feeling is that very inexperienced juniors that would not utilise their 1 hour or greater (playing 'micky mouse chess' should not play ACF rated games (it gives rise to the phrase 'garbage in, garbage out'). As you correctly state playing heaps of junior only games just reduces their RD. Although when some of them do become better and play other players (for example seniors) Glicko2 will move them and the seniors to their appropriate relative ratings, these relative ratings will be lower than they would if the juniors RD was higher. Plus with a high turnover of juniors many of these 'points gained' will also be lost. Hence your feelings of juniors not moving up quickly enough, seniors falling too much, and a general air of deflation (I'm using my antena here - I don't really know you feel). With the greatest respect though (I hope I don't sound too much like Sir Humphrey Appleby), this isn't really the fault of Glicko2, the newish juniors should be introduced into ACF rated games only once they have more experience - until then I believe you have a perfectly fine Qld junior rating system (it used to be maintained by Ian Murray, I think) and ACF rapid ratings to accomodate their need for a rating. But their main need should be to continue enjoying and improving their game and making chess friends :)

I also consider that what you may think is happening in the gold coast is wide-spread, but I very much doubt this is the case.

rob
14-09-2004, 10:23 AM
Frankly, I could give a tinker's cuss who rob is in reality. He seems OK to me, and if he is starting to see your block-o-cement head through the green and brown abuse camoflage, then all the better. :P

I think you are being somewhat unfair on Bill as he has demonstrated that he is not afraid to enhance the ACF rating system if there is clear benefit and the full implications have been considered and tested. Making changes opens oneself up to criticism - he has shown willing when it has been found that there can be logical improvement. Almost couragous ('Yes Minister') :)

Expressing ideas as you do Matt is no bad thing - but when you play the man and not the ball you tend to get everyone off-side. But your ideas do get ppl considering other options and looking at things from a variety of angles, kind of stimulating - but when you grab the bull by the horns you can be frustrating.

Thanks for the advise Matt 'I know the score' (Porridge).

I can talk in cliches until the cows come home :)

Cat
14-09-2004, 12:59 PM
Hi David, Unlike the Gold Coast, most of the WA juniors play ACF rated games in (senior) club and weekend tourneys. The only ACF rated junior only events these days are the state junior ch'ships and then only the U14, U16, U18 are ACF rated. All 5 age groups are 'Junior Chess Council' rated, as are any official games that juniors play in that are 20mins or more per player. In the late 1990's I ran a ACF rated junior invitation tourney each year for our stronger juniors that was held in early December as a preparation for the National juniors (this hasn't been held in recent years).


My feeling is that very inexperienced juniors that would not utilise their 1 hour or greater (playing 'micky mouse chess' should not play ACF rated games (it gives rise to the phrase 'garbage in, garbage out'). As you correctly state playing heaps of junior only games just reduces their RD. Although when some of them do become better and play other players (for example seniors) Glicko2 will move them and the seniors to their appropriate relative ratings, these relative ratings will be lower than they would if the juniors RD was higher. Plus with a high turnover of juniors many of these 'points gained' will also be lost. Hence your feelings of juniors not moving up quickly enough, seniors falling too much, and a general air of deflation (I'm using my antena here - I don't really know you feel). With the greatest respect though (I hope I don't sound too much like Sir Humphrey Appleby), this isn't really the fault of Glicko2, the newish juniors should be introduced into ACF rated games only once they have more experience - until then I believe you have a perfectly fine Qld junior rating system (it used to be maintained by Ian Murray, I think) and ACF rapid ratings to accomodate their need for a rating. But their main need should be to continue enjoying and improving their game and making chess friends :)

Yes, that's exactly right. Glicko 2, with it's variable RD does however, exacerbate the problem.

Yes we have a well run Queensland Junior Rating system, and recently it underwent a face lift, I think it's looking quite good.



I also consider that what you may think is happening in the gold coast is wide-spread, but I very much doubt this is the case.

That's right, it's really the areas like The Gold Coast, Canberra and Box Hill. It's also a problem for kids in the country. These areas are, however, where the future of Australian Chess lies, and it we can't get it right there, then we have some problems. Where the adult/junior ratio is near the norm, say around 2:1, there's no major problem. The ratio's are reversed in our area, that's where the problem lies.

Cat
14-09-2004, 01:05 PM
Bill as he has demonstrated that he is not afraid to enhance the ACF rating system if there is clear benefit and the full implications have been considered and tested. Making changes opens oneself up to criticism - he has shown willing when it has been found that there can be logical improvement. Almost couragous ('Yes Minister') :)

Expressing ideas as you do Matt is no bad thing - but when you play the man and not the ball you tend to get everyone off-side. But your ideas do get ppl considering other options and looking at things from a variety of angles, kind of stimulating - but when you grab the bull by the horns you can be frustrating.

Thanks for the advise Matt 'I know the score' (Porridge).

I can talk in cliches until the cows come home :)

I agree, but Matt is not the initiator, he's the responder. He's playing the rule laid down by those who should know better. Personally, I welcome Matt's imaginative straight-talking, if there wasn't so much double-speak going on I'm sure Matt's mind would prove itself to be very fertile pasture.

Bill Gletsos
14-09-2004, 01:20 PM
My feeling is that very inexperienced juniors that would not utilise their 1 hour or greater (playing 'micky mouse chess' should not play ACF rated games (it gives rise to the phrase 'garbage in, garbage out'). As you correctly state playing heaps of junior only games just reduces their RD.
The point is the Gold Coast Juniors are not playing 10,000 of normal rated games. In fact even within a 12 month period they are playing no where near that many normal rated games.
Glicko2 will correctly reduce their RD's indicating that based on their results their ratings are reliable. However if the juniors results (or for that matter also an adults) performs significantly differently from their previous rating then the volatility factor will kick in to increase their RD and hence increase their rise or fall in rating. As I noted previously for rapidly improving players be they junior or adult although the volatility factor will still kick in for low numbers of games, 12-15 games in a rating period isnt really enough to determine if the player has significantly improved or just going thru a good or bad patch.

Bill Gletsos
14-09-2004, 01:22 PM
Yes, that's exactly right. Glicko 2, with it's variable RD does however, exacerbate the problem.
As usual you have no clue what you are rabbitting on about.

Bill Gletsos
14-09-2004, 01:29 PM
I agree, but Matt is not the initiator, he's the responder.
More often than not he is the initiator.
His responses tend to the crude and vulgar.
You of course obviously condone Matt's crude and vulgar responses because you have never criticised him for them.



He's playing the rule laid down by those who should know better.
No one is laying down rules regarding Matts crude and foul mouthed respones.
In fact he is the only BB resident who regularly responds in that way.


Personally, I welcome Matt's imaginative straight-talking,
Good to see you finally acknowledge you support his crude and vulgar responses.
Just shows what a complete hypocrite you are when it comes to complaining about BB abuse.



if there wasn't so much double-speak going on I'm sure Matt's mind would prove itself to be very fertile pasture.
You and matt are the only ones involved in misreprenting the facts.
As for fertile pastures, they are generally the ones covered in cow manure.

PHAT
14-09-2004, 06:18 PM
I think you are being somewhat unfair on Bill as he has demonstrated that he is not afraid to enhance the ACF rating system if there is clear benefit and the full implications have been considered and tested. rob, this goes to the heart of our conflict. BG demand crystal clear and infinite contigency plans. I, on the other hand, demand that risks be taken and mistakes accepted. My record on accepting people's mistakes is the best here. If BG made more changes and more mistakes, I would leave him alone.


Expressing ideas as you do Matt is no bad thing - but when you play the man and not the ball you tend to get everyone off-side.

Actually, it is oly BG and KB with whom I get into incendury diplomacy. In any case, if Bill desisted in his 1:1 sentence to moron insult ratio, I would cease my 1:1 post to napam attack ratio.

arosar
14-09-2004, 06:28 PM
If BG made more changes and more mistakes, I would leave him alone.

I find this quite ridiculous. This is why it's sometimes hard to support you, though I feel that you could actually be a force in NSW chess. You got some good ideas man - very passionate and all that. I voted for you, didn't I? As a result of your overly risky attitude - we'll oppose you instead. You're too bloody dangerous.


. . .incendury [sic] diplomacy . . .

What do they call that - an oxymoron or something?

AR

Bill Gletsos
14-09-2004, 06:44 PM
rob, this goes to the heart of our conflict. BG demand crystal clear and infinite contigency plans. I, on the other hand, demand that risks be taken and mistakes accepted. My record on accepting people's mistakes is the best here. If BG made more changes and more mistakes, I would leave him alone.
Yes, but you are simply being a fool.
The majority of players would not accept changes made to the rating system without adequate testing taking place. Before we make changes we always do this.
Only a complete idiot would suggest otherwise.



Actually, it is oly BG and KB with whom I get into incendury diplomacy. In any case, if Bill desisted in his 1:1 sentence to moron insult ratio, I would cease my 1:1 post to napam attack ratio.
The reason why others probably dont confront you is ebcause they are not prepared to be on a receiving end of a barrage of your crude and vulgar responses.
Kevin and I of course are prepared to confront you irrespective of this because of your continual posting of misinformation and attempts at beatups.
You are also the only regulat BB poster who reverts to crudeness and vulgarity.

Bill Gletsos
14-09-2004, 06:52 PM
I find this quite ridiculous. This is why it's sometimes hard to support you, though I feel that you could actually be a force in NSW chess. You got some good ideas man - very passionate and all that. I voted for you, didn't I? As a result of your overly risky attitude - we'll oppose you instead. You're too bloody dangerous.
Well said AR.



What do they call that - an oxymoron or something?
Amazing how the word moron appears in reference to Matt's posts isnt it. ;)

PHAT
14-09-2004, 07:01 PM
The majority of players would not accept changes made to the rating system without adequate testing taking place. . The majority of players (~95% ?) don't know if you do testing or not. Nor do they care if you think it is or is not adequate. They don't know if mistakes are being made and/or rectified. All they know is that they have no say in how responsive or resitant their rating is to their performance.

I reckon there is a case to be made for players electing to have a hi/med/low RD each 12 months. At least it would shut them up.



You are also the only regular BB poster who reverts to crudeness and vulgarity.

So? Am I supposed to be mortified. Am I supposed to beg forgiveness. Am I supposed to conform to your sneering elitist standards of properness. FY!

Bill Gletsos
14-09-2004, 07:05 PM
The majority of players (~95% ?) don't know if you do testing or not.
They dont need to know, because any intelligent person would expect it to happen.


Nor do they care if you think it is or is not adequate.
No, they would expect it to be tested. Only a fool like you would suggest otherwise.


They don't know if mistakes are being made and/or rectified. All they know is that they have no say in how responsive or resitant their rating is to their performance.

I reckon there is a case to be made for players electing to have a hi/med/low RD each 12 months. At least it would shut them up.
Too bad there isnt something that would shut you up. :owned:


So? Am I supposed to be mortified. Am I supposed to beg forgiveness. Am I supposed to conform to your sneering elitist standards of properness. FY!
No but perhaps eventually the admins will tire of your crude and vulgar responses and decide you have had enouugh warnings and that enough is enough and remove you permanently.

PHAT
14-09-2004, 07:14 PM
As a result of your overly risky attitude - we'll oppose you instead. You're too bloody dangerous.


:lol: I agree, Ameil.

However, remember this; There are horses for courses, but if you bet on nags, you will always lose. Or another one I just made up; There are horses for courses, but if bet only on the favorites, you will eventually go broke.

PHAT
14-09-2004, 07:17 PM
No but perhaps eventually the admins will tire of your crude and vulgar responses and decide you have had enouugh warnings and that enough is enough and remove you permanently.

That'll be the day.

PHAT
14-09-2004, 07:19 PM
No but perhaps eventually the admins will tire of your crude and vulgar responses and decide you have had enouugh warnings and that enough is enough and remove you permanently.

That'll be the day, ya flaming Edgar.

Garvinator
14-09-2004, 07:21 PM
you dont get two posts to abuse the same quote ;)

Bill Gletsos
14-09-2004, 08:13 PM
That'll be the day.
Yes, well the chance of that day arriving far exceeds that of a day arriving when you would ever be put in charge of the NSWCA finances.

Bill Gletsos
14-09-2004, 08:14 PM
That'll be the day, ya flaming Edgar.
Yes, well the chance of that day arriving far exceeds that of a day arriving when you would ever be put in charge of the NSWCA finances, you useless do nothing individual. :hand:

Bill Gletsos
14-09-2004, 08:15 PM
you dont get two posts to abuse the same quote ;)
You dont?
Are you sure? ;)

Kevin Bonham
14-09-2004, 08:29 PM
Down here in the firey pits, we slowly chip away at his stupidness and eventually things happen.

But they're generally not the things you call for. Take the Gold Coast. There was an issue there but David completely misidentified what it was and hence the solutions he proposed were wrong.

The above quote reminds me of the way cryptozoologists try to rub it in to science every time a new species is discovered, even though the new species that scientists find are virtually never the ones the cryptos said were out there.

Garvinator
14-09-2004, 08:39 PM
You dont?
Are you sure? ;)
starter wont be happy at you two rigging his betting scams :P :owned:

rob
14-09-2004, 10:11 PM
rob, this goes to the heart of our conflict. BG demand crystal clear and infinite contigency plans. I, on the other hand, demand that risks be taken and mistakes accepted. My record on accepting people's mistakes is the best here. If BG made more changes and more mistakes, I would leave him alone.

You cannot seriously consider that heaps of whimsical ideas should be applied to the ratings all the time - ratings would end up with no credibility and criticism would be wide-spread. Why make mistakes when they can be avoided with some checking? If you were applying for a job would you use spell-check? Dare I ask what you wish to achieve (your objective) with the changes that you are seeking? It isn't always obvious what you are aiming for except perhaps chaos - your messages and any logical reasoning you may have to support your ideas get lost in your bickering.

It is nice to hear that you are so accepting of people's mistakes :lol: . So judging by your last sentence (If BG made more changes and more mistakes, I would leave him alone.), have you been trying to make Bill appear to be wrong just in order to force yourself to leave him alone?

PHAT
14-09-2004, 10:54 PM
You cannot seriously consider that heaps of whimsical ideas should be applied to the ratings all the time No, not whimsical ideas, interesting ideas.
... ratings would end up with no credibility and criticism would be wide-spread. Sounds like what we have now.
Why make mistakes when they can be avoided with some checking? How much is checking required is proportional to the damage that could be done. I don't think that the kinds of changes DR, myself and others (including yourself) would kill the system.


So judging by your last sentence (If BG made more changes and more mistakes, I would leave him alone.), have you been trying to make Bill appear to be wrong just in order to force yourself to leave him alone?

THAT is too acurite to be funny ........ :lol: :clap:

Garvinator
14-09-2004, 11:03 PM
Sounds like what we have now. If ratings werent accurate a very high majority of the time, then there would be more upsets, instead of the first round of most monster swisses being junk rounds.

Alan Shore
14-09-2004, 11:12 PM
If ratings werent accurate a very high majority of the time, then there would be more upsets, instead of the first round of most monster swisses being junk rounds.

NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO

No mentioning this first round stuff again, your comments are complete rubbish, we've been over it so many times and you and starter are wrong, the system we have now is fine. Thank you and good night.

Garvinator
14-09-2004, 11:13 PM
NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO

No mentioning this first round stuff again, your comments are complete rubbish, we've been over it so many times and you and starter are wrong, the system we have now is fine. Thank you and good night.
i was slapping matthew round, not mentioning competitiveness. I was actually saying that i think ratings are accurate in the high majority of occassions

Alan Shore
14-09-2004, 11:14 PM
i was slapping matthew round, not mentioning competitiveness. I was actually saying that i think ratings are accurate in the high majority of occassions

I know you were but you slipped it in anyway.

Garvinator
14-09-2004, 11:21 PM
I know you were but you slipped it in anyway.
stop it bruce :P unlike most ppl, i was trying to give an example to back up my point.

Alan Shore
14-09-2004, 11:26 PM
stop it bruce :P unlike most ppl, i was trying to give an example to back up my point.

OK I forgive you.

Garvinator
14-09-2004, 11:30 PM
OK I forgive you.
i dont need your forgiveness ;) :lol: :uhoh:

Bill Gletsos
14-09-2004, 11:50 PM
No, not whimsical ideas, interesting ideas.
No matter how you try and sugar coat it they are stupid ideas none the less.


Sounds like what we have now.
What we have now from some people is uninformed and unjustifed comments.
What we would have if we implemented your ideas would be uninformed but justified criticism.


How much is checking required is proportional to the damage that could be done. I don't think that the kinds of changes DR, myself and others (including yourself) would kill the system.
The critical part of the above sentnece is I[Matt] dont think.
You of course have demonstrated that capability many times over the since the BB inception.



THAT is too acurite to be funny ........ :lol: :clap:
If that is so, then if just shows what an absolute fool you are.