PDA

View Full Version : Is it time to benchmark?



ursogr8
17-09-2005, 10:32 PM
hi Bill

The Whitehorse Chess Club has been submitting tourneys for rapid rating for a couple of years. Not may other junior tourneys are submitted in Victoria, so there is a risk that it is a bit of a 'closed' pool. There is a little bit of 'realism' in the ratings as about 10 out of 60 players in each monthly tournament are adults with fairly stable (normal as well as rapid) ratings.

Those 10 adults and (a small) number of juniors also play in Box Hill Chess Club events...rapid and normal.
Here is an analysis of the current 175 BHCC members

> First...have to discard 67 as they have only one rating. That leaves 108 with both a rapid and normal rating.

>> 1 has a difference (normal - rapid) greater than 500
>> 2 have ..... in the range 500-400
>> 7 have........................400-300
>> 20..............................300-200
>> 30..............................200-100
>> 28..............................100-0
>>> 13............................0 (-100)
>>> 4.............................(-100)_(-200)
>>> 3.............................(-200)_(300)


I can send the more detailed analysis if you PM me a mailbox.

regards
starter

Alan Shore
17-09-2005, 11:21 PM
What I was originally annoyed at years ago (couldn't care less now really) was how QLD Junior ratings were automatically imported into the ACF Rapid rating pool, thus overrating GC players etc. etc. as I alluded to some 6 years ago now. So if you do standardise, make sure it's at least going to be consistent - dunno who made that poor decision.

ursogr8
18-09-2005, 09:33 AM
What I was originally annoyed at years ago (couldn't care less now really) was how QLD Junior ratings were automatically imported into the ACF Rapid rating pool, thus overrating GC players etc. etc. as I alluded to some 6 years ago now. So if you do standardise, make sure it's at least going to be consistent - dunno who made that poor decision.

hi BD

My request to Bill is not about importing/initiating new players into a pool.

Instead, I am looking a collection of 108 players who have long-standing normal ratings and rapid ratings, with a clear difference between the averages in both measures.

(But I obviously agree with standardisation of process).

starter

ursogr8
18-09-2005, 06:40 PM
Read from the top down to see chronological order >>

starter:

>Bill...are you going to engage on the benchmarking question on the new ratings thread?...


Bill Gletsos:

you seem to be under the misapprehension that normal and rapid ratings of players should be similar



starter:

> Bill........ah...so you are going to engage; and the first debating point is whether the calibration scales should be aligned. I look forward to your argument about the need for independence.



Bill Gletsos:

Nope. There is no need for alignment. In fact when FIDE was running their Rapid list it was only a 3 digit number



Bill Gletsos:

The reason for this was so that it was not confused as being related to the players normal rating


starter:

> Bill; now that I have got your valued attention, do you think we could adjourn to the actual thread....instead of such a public place.


Bill Gletsos:

Nope. I see it as a non issue and not worthy of wasting my time on.

Frank Walker
18-09-2005, 07:18 PM
oooooooooooooooh
Fight of the BIGGEST and the Big BIG Men

ursogr8
18-09-2005, 07:41 PM
So, Bill, as the official ACF Ratings Officer (representative on this bb), can you just advise me of a fact (without controversy) >> Is there an ACF policy on the mission/vision of the rapid ratings scale? And where is it visible? In other words, what is the objective of the process that collects money from us?

regards
starter

Bill Gletsos
18-09-2005, 08:07 PM
So, Bill, as the official ACF Ratings Officer (representative on this bb), can you just advise me of a fact (without controversy) >> Is there an ACF policy on the mission/vision of the rapid ratings scale? And where is it visible? In other words, what is the objective of the process that collects money from us?The rapid ratings list simply provides rapid ratings for players. The rapid list meets that purpose.

ursogr8
18-09-2005, 08:14 PM
The rapid ratings list simply provides rapid ratings for players. The rapid list meets that purpose.

This is a good start by you Bill. No stonewalling. (a/c to note...it apparently in the way you ask the questions).

Now we move from fact to discourse.
You have listed one reason that the rapid need to be independent from the normal ratings; viz, so that the average Joe does not confuse one or the other. Are there any other reasons to put on the table at the beginning? Or just this one reason.....so we don't confuse the fact that we have two ratings?


still, with regards
starter

Bill Gletsos
18-09-2005, 08:17 PM
This is a good start by you Bill. No stonewalling. (a/c to note...it apparently in the way you ask the questions).

Now we move from fact to discourse.
You have listed one reason that the rapid need to be independent from the normal ratings; viz, so that the average Joe does not confuse one or the other. Are there any other reasons to put on the table at the beginning? Or just this one reason.....so we don't confuse the fact that we have two ratings?Not sure what you mean.
Do you mean why is there a seperate rapid list at all?
The answer is that if there wasnt then rapid games would not be rated and many players mainly juniors would have no rating whatsoever.

ursogr8
18-09-2005, 08:24 PM
Not sure what you mean.
Do you mean why is there a seperate rapid list at all?
The answer is that if there wasnt then rapid games would not be rated and many players mainly juniors would have no rating whatsoever.

OK
I will re-phrase like this.

You, as ACF Ratings Officer (and cohorts) must have decided that a scale and an update formula were required (for rapids). As you point out, you could have picked a 3 digit scale, a 4 digit scale, a 5 digit scale, or even alphabetic.
You chose basically a 4 digit scale.
Next choice is whether to choose whether to align the rapid with the normal. That is, a player with equal skill at rapid and normal would end up with very similar measures for the two if he was active and if he moved across closed pools.
I think you are saying you choose to NOT align because you did not want players to confuse their two numbers.
Correct so far?

w.k.r,
starter

PHAT
18-09-2005, 08:29 PM
There is no need for alignment. In fact when FIDE was running their Rapid list it was only a 3 digit number. The reason for this was so that it was not confused as being related to the players normal rating


Most of the time you heap scorn upon the FIDE rating system (as well as other FIDE decissions.) Now, when starter called you to justify the non-alignment of rapid and normal ratings in the ACF system, you cite soley the FIDE methods and reasons as your own justification.

It is an error to, on one hand ridicule the FIDE methods, while on the other hand using similarity to FIDE methods as evidence that your (non)policy is sound.

ursogr8
18-09-2005, 08:32 PM
Most of the time you heap scorn upon the FIDE rating system (as well as other FIDE decissions.) Now, when starter called you to justify the non-alignment of rapid and normal ratings in the ACF system, you cite soley the FIDE methods and reasons as your own justification.

It is an error to, on one hand ridicule the FIDE methods, while on the other hand using similarity to FIDE methods as evidence that your (non)policy is sound.

Yeh. I saw this illogicality earlier too. But, at the moment I am just picking the long-hanging fruit. ;)

But, in the interest of discipline in the ring....TAG.


starter

Bill Gletsos
18-09-2005, 08:32 PM
OK
I will re-phrase like this.

You, as ACF Ratings Officer (and cohorts) must have decided that a scale and an update formula were required (for rapids). As you point out, you could have picked a 3 digit scale, a 4 digit scale, a 5 digit scale, or even alphabetic.
You chose basically a 4 digit scale.
Next choice is whether to choose whether to align the rapid with the normal. That is, a player with equal skill at rapid and normal would end up with very similar measures for the two if he was active and if he moved across closed pools.
I think you are saying you choose to NOT align because you did not want players to confuse their two numbers.
Correct so far?You miss the point.
The formula used is the same as for normal games.
The rapid measures players results in rapid games.

In fact if every players rapid strength matched their normal strength their would be no need for a rapid list.

However players normal and rapid strengths can vary quite a lot. Hence a player rated 1600 normal may be 1800 rapid whilst another rated 1600 normal may be 1400 rapid. Likewise a player rated 1600 normal plays like a 1600 rapid.

You cannot align these three players rapid ratings to the normal scale.

There is no reason to expect that a player rated 1600 at rapid is equal in skill as a player rated 1600 in normal.

It is a non issue.
Thus endeth the discussion.

Bill Gletsos
18-09-2005, 08:33 PM
Yeh. I saw this illogicality earlier too. But, at the moment I am just picking the long-hanging fruit. ;)

But, in the interest of discipline in the ring....TAG.You and your tag partner can play with yourselves.

ursogr8
18-09-2005, 11:02 PM
You miss the point.

I am willing to listen to the point...so thanks in advance for what is your response.


The formula used is the same as for normal games.
Of course; quite logical. I have no issue with this.


The rapid measures players results in rapid games.
Of course; quite logical. I have no issue with this.


In fact if every players rapid strength matched their normal strength their would be no need for a rapid list.
Of course; quite logical. I have no issue with this. And it is a strawman you have introduced, not my suggestion at all.


However players normal and rapid strengths can vary quite a lot. Hence a player rated 1600 normal may be 1800 rapid whilst another rated 1600 normal may be 1400 rapid. Likewise a player rated 1600 normal plays like a 1600 rapid.

But of course, we would all agree with this empirical data.



You cannot align these three players rapid ratings to the normal scale.
Another strawman of yours.


There is no reason to expect that a player rated 1600 at rapid is equal in skill as a player rated 1600 in normal.

This is a self evident truth if you don't align the scales.
But it is a different matter if the ACF Ratings Officer uplifted the rapids so that the mean of the population (with both ratings) matched the mean of the normal.


It is a non issue.

So what does this mean. You don't discuss because it is only one of us? You don't discuss because I am barking up the wrong tree? You don't discuss because it is awkward for you to do the re-alignment. Other?


Thus endeth the discussion.

That will be a first for you Bill.


starter

ursogr8
19-09-2005, 08:18 AM
Bill

Bump a part of post #13
I asked "I think you are saying you choose to NOT align because you did not want players to confuse their two numbers.
Correct so far?".

starter

Davidflude
19-09-2005, 11:53 AM
As one of the adults who play in the Box Hill allegros I will make the following points

My ratings are Highest ICCF next normal Australian rating lowest Australian rapid play rating. This in my opinion reflects my ability well. Sure I lose points in Allegro to rapidly improving underrated juniors, so what. If they play well enough
to beat me they deserve the rating points.

ursogr8
20-09-2005, 10:59 AM
Bill Gletsos:

The reason for this was so that it was not confused as being related to the players normal rating

Therefore Bill, why do we not start the rapid ratings at quite a different base point like 7000, instead of zero. That would distinguish them from the normal ratings.
Or as you remarked elsewhere, have rapid ratings as a three digit number (for all players); that would be a bit more distinguishing.


As you have it now
> normal ratings 0 to 2600
> rapid ratings 0 to 2600.



Each Club member I have spoken to, since the scaling issue first came to notice, has intuitively expected that
the average of the normals would be close to the average of the rapids (for players who have both a rapid and a normal rating).

And as we can see from the data in the first post, this is not near the case for the (near-)closed population in Victoria.

regards
starter

ursogr8
21-09-2005, 04:28 PM
You miss the point.
The formula used is the same as for normal games.
The rapid measures players results in rapid games.

In fact if every players rapid strength matched their normal strength their would be no need for a rapid list.

However players normal and rapid strengths can vary quite a lot. Hence a player rated 1600 normal may be 1800 rapid whilst another rated 1600 normal may be 1400 rapid. Likewise a player rated 1600 normal plays like a 1600 rapid.

You cannot align these three players rapid ratings to the normal scale.

There is no reason to expect that a player rated 1600 at rapid is equal in skill as a player rated 1600 in normal.

It is a non issue.
Thus endeth the discussion.

Bill
Reference this sentence in your post.

In this old post of yours on planned rating changes (http://chesschat.org/showpost.php?p=18967&postcount=1) why do you contemplate using the rapid rating to SEED a new players normal rating if they are not somewhat related?

starter

PHAT
21-09-2005, 04:36 PM
Bill
Reference this sentence in your post.

In this old post of yours on planned rating changes (http://chesschat.org/showpost.php?p=18967&postcount=1) why do you contemplate using the rapid rating to SEED a new players normal rating if they are not somewhat related?

starter

I saw this new word in another thread and like any good meme I have absorbed it .....

ZING!

ursogr8
21-09-2005, 10:24 PM
I saw this new word in another thread and like any good meme I have absorbed it .....

ZING!

Matt

Does this new word
a) describe the deal where a poster is annointed as a MOD on the condition that he forsakes all personal opinions and debate on the bb,
or
b) the sound of gelding with a sharp knife
or
c) the sight of a poster reduced to quoting facts from files, and no human analysis allowed.

starter

PHAT
21-09-2005, 11:07 PM
Multiple choice usually has 4 options. May I suggest choice (d) as being "All of the above."

Ian Rout
22-09-2005, 11:39 AM
Although it's not a major issue, I think it would be convenient if the scales were roughly aligned so that somebody who turns up for a tournament with only a Rapid rating could be given that rating for pairing and maybe rating prize purposes.

In the ACT on the last list there were 64 players who had ! or !! for both normal and rapid ratings. On average the rapid was 89 points lower; or if the top and bottom 10% were removed to discount extreme cases (or people who are just much better at one form) little changed, the rapid was 90 points lower. So I would suggest (in ACT) as a simple transformation Normal = Rapid + 90 for seeding new players.

ursogr8
22-09-2005, 01:41 PM
Although it's not a major issue, I think it would be convenient if the scales were roughly aligned so that somebody who turns up for a tournament with only a Rapid rating could be given that rating for pairing and maybe rating prize purposes.

In the ACT on the last list there were 64 players who had ! or !! for both normal and rapid ratings. On average the rapid was 89 points lower; or if the top and bottom 10% were removed to discount extreme cases (or people who are just much better at one form) little changed, the rapid was 90 points lower. So I would suggest (in ACT) as a simple transformation Normal = Rapid + 90 for seeding new players.

Thanks Ian, your post adds another interesting metric.....the average difference.
I calculated this new metric for Box Hill (108 players as described above) and our average was a 113 differential for normal ratings being higher than rapid ratings.


starter

ursogr8
25-09-2005, 12:12 PM
The rapid ratings list simply provides rapid ratings for players. The rapid list meets that purpose.

Bill

On the FIDE Forum site (http://forum.fide.com/viewforum.php?f=1&topicdays=0&start=250) has there been any discussion on the calibration of rapid ratings, in the sense whether there is a target average for the ratings. For example, have FIDE declared that over time that the average should be 1500 (or some such target), or is it just allowed to float. And is it forever decreasing or forever increasing?

regards
starter