PDA

View Full Version : Does Accelerating pairings affect Glicko?

firegoat7
25-01-2004, 10:05 AM
Dear Bill,

This question is directed to Bill Gletsos.

Firstly, I do not pretend to know anything about ratings. This is an honest question.

I decided to accelerate the pairings in the first round of the Australian day open, but I am interested in the question of what would happen if I had not.

Let us take two imaginative tournaments as an example.

Player Watersheep plays the same field in both tournaments except the first round, all players are 2000 except A1-1500 and A2-1800. Let us say for arguements sake that both tourneys are 5 rounds. As an example let us say in all tourneys Watersheep scores 3/5 against an average field of
1900 for tourney A1 and 1960 for tourney A2.I have presumed the results get avergaed out against the field (Is this correct?) This all seems reasonable to me and appears to show how a chessclub can manipulate ratings.

Please tell me if I have made a mistake here because this is the logic I have based the pairings on. Furthermore one entrant asked why I did this and I replied that in my opinion the average chess player at MCC was underated and this seemed like a simple way in which we could increase the average pool.

Bill Gletsos
25-01-2004, 12:09 PM
Dear Bill,

This question is directed to Bill Gletsos.

Firstly, I do not pretend to know anything about ratings. This is an honest question.

I decided to accelerate the pairings in the first round of the Australian day open, but I am interested in the question of what would happen if I had not.

Let us take two imaginative tournaments as an example.

Player Watersheep plays the same field in both tournaments except the first round, all players are 2000 except A1-1500 and A2-1800. Let us say for arguements sake that both tourneys are 5 rounds. As an example let us say in all tourneys Watersheep scores 3/5 against an average field of
1900 for tourney A1 and 1960 for tourney A2.I have presumed the results get avergaed out against the field (Is this correct?) This all seems reasonable to me and appears to show how a chessclub can manipulate ratings.

Please tell me if I have made a mistake here because this is the logic I have based the pairings on. Furthermore one entrant asked why I did this and I replied that in my opinion the average chess player at MCC was underated and this seemed like a simple way in which we could increase the average pool.

I'm sure you raised an issue about ratings on the old board and I thought I made it clear then, that games are rated on an individual basis and their is no averaging of opponents rating in the calculations.

Hence in that sense it makes no difference if you accelerate the pairings or not.

However if a player played a 1800 in the first round and beat him as opposed to playing a 1500 in the first round and beating him then obviously he gains more rating points.

firegoat7
25-01-2004, 12:45 PM
Hi Bill,

Thanks for you quick and accurate reply.
I guess tournament organisers can actually cater for different segments of their playing group.

So like if a club has promising underated juniors, who want to see their ratings improve quickly they probably should not accelerate pairings. Whilst if a club has older players they would like to protect then they would.

Does anybody know if there has been any long term studies completed on such a phenomenea?

I suppose this means that an arbiter(or whoever does the pairings) as an observer could actually contribute to significant differences in ratings.

Bill Gletsos
25-01-2004, 12:50 PM
fg7,

You may want to read starters thread on the old BB called "How competetive do you want it to be?".

A similarly titled thread exists on this board but the background information is on the old board.

ursogr8
25-01-2004, 02:38 PM
Dear Bill,

This question is directed to Bill Gletsos.

Firstly, I do not pretend to know anything about ratings. This is an honest question.

I decided to accelerate the pairings in the first round of the Australian day open, but I am interested in the question of what would happen if I had not.

Let us take two imaginative tournaments as an example.

Player Watersheep plays the same field in both tournaments except the first round, all players are 2000 except A1-1500 and A2-1800. Let us say for arguements sake that both tourneys are 5 rounds. As an example let us say in all tourneys Watersheep scores 3/5 against an average field of
1900 for tourney A1 and 1960 for tourney A2.I have presumed the results get avergaed out against the field (Is this correct?) This all seems reasonable to me and appears to show how a chessclub can manipulate ratings.

Please tell me if I have made a mistake here because this is the logic I have based the pairings on. Furthermore one entrant asked why I did this and I replied that in my opinion the average chess player at MCC was underated and this seemed like a simple way in which we could increase the average pool.

I'm sure you raised an issue about ratings on the old board and I thought I made it clear then, that games are rated on an individual basis and their is no averaging of opponents rating in the calculations.

Hence in that sense it makes no difference if you accelerate the pairings or not.

However if a player played a 1800 in the first round and beat him as opposed to playing a 1500 in the first round and beating him then obviously he gains more rating points.

Bill
Yes, I can recall you answering this exact question before on the old BB.
starter

ursogr8
26-02-2004, 10:17 AM
Hi Bill,

I guess tournament organisers can actually cater for different segments of their playing group.

Box Hill cater for the A and B segments in SWISSes by permanent acceleration of the B group by 2 points, but with the proviso no B player can win an A prize. This has the effect of alllowing B players to win a lot more games in a 7-round SWISS (thus making the results much fairer ) while still able to be matched with 2nd-quartile players.

So like if a club has promising underated juniors, who want to see their ratings improve quickly they probably should not accelerate pairings.

In the Box Hill SWISS described above the juniors ratings improve quickly because they get to play 7 competitive games in a 7 round SWISS instead of the usual 4 competitive games and 3 junk rounds where the junior has no hope of a win or draw.
Thus the acceleration system we have designed maximises the juniors chances of getting an appropriate rating and avoiding being under-rated. It has the flow-on benefit that seniors in the 3rd quartile are not playing under-rated juniors and hence the senior’s rating is not at risk.

Whilst if a club has older players they would like to protect then they would.

We do protect the older players by ensuring the junior ratings are representative of strength.

Does anybody know if there has been any long term studies completed on such a phenomenea?

We have been running this permanently accelerated SWISS for 3 years and Bill’s analysis (on a case-by-case basis we have submitted to him) show we have no seniors whose ratings have been hit to leg.

I suppose this means that an arbiter(or whoever does the pairings) as an observer could actually contribute to significant differences in ratings.

In our case we think the tournament designer, not the Arbiter, holds the key to solving the problem. Essentially the key to ensuring reliable ratings is to encourage players to be involved in as many rated games as possible. Thus if 7 games in a 7-round SWISS can be won or lost by a player then this gives more data-points for reliability than a SWISS where 3 rounds are JUNK rounds and the player has little chance of their rating being regularised. And if junior ratings are reliable then so also will be senior ratings of those who have to play them

starter

eclectic
26-02-2004, 11:46 AM
I myself would NOT object to starter's super accelerated system provided it was applied universally across all events in Australia.

Under the old system a novice of any age could enter open events in the hope that after 20 to 30 games they would have a decent rating.

If there were to be this S-A system where people are segregated into divisions and told that because of their rating they have they have no right to win an event then let's go all the way and say that any novice has to be rated at 500 or 1000 and be put in the lowest division and then have to climb their way up through the ranks.

Let's make all major chess events in Australia consist of 3 divisions i e like the Doeberl Cup except I would prefer the divisions to be Over 1900 , 1401 - 1900 , and Under 1400

Lower rated players could not play in higher divisions and possibly unrated players in the top division only - unless a rule was introduced that players had to get their rating established at club level before they could enter class 3 grand prix events

I wonder if the Ballarat Begonia organisers would like that.

I find the term "junk round" interesting. Didn't certain events at Box Hill use to be 7 rounds ? Now there's 9 !! The maximum number of entries allowed there is usually 120 which is under the 128 that a 7 round normal swiss caters for anyway. It would seem that 2 junk rounds were already introduced to have competitors at their "proper" level by round 3.

I feel the need to weep at those poor high rated players who suffer early round upsets at the hands of Box Hill ( or visiting ) juniors or micro juniors.

Never mind !! They're only junk rounds !! Just write them off !!

Sounds a bit like negative gearing - writing off investment property losses at tax time.

:hmm:

eclectic

Actually,
What makes a game a junk round ?
The opponent ?
Or your attitude to the game ?

ursogr8
26-02-2004, 12:51 PM
I myself would NOT object to starter's super accelerated system

Well that is a start, eclectic. You are on the way to becoming a convert.

provided it was applied universally across all events in Australia.

But this ‘strawman’ suggestion of ours is a bit centralist. Personally I think tournament designers should have a free hand at finding what suits them and what suits their chess population. I don’t think you will get far proposing a universal solution as you suggest.

I find the term "junk round" interesting. Didn't certain events at Box Hill use to be 7 rounds ? Now there's 9 !!

We have had 7 round SWISSes and 9 round SWISSes for the better part of 20 years. There is nothing new in this. The choice between 7 and 9 used be one of scheduling. For example, the Championship played in the Winter months has often been 9 rounds because 9 clear rounds were available. But the Autumn cup has had to be adjustable depending where school holidays, Begonia and Easter fall in the calendar.
Both 7 round and 9 rounds usually find a clear tournament winner. And our choice of 7 or 9 has not usually been governed by this one particular factor.
The choice between 7 and 9 does not change the likelihood of junk rounds; what does change that likelihood is the acceleration decision. Two rounds only of acceleration usually sees two rounds at least of junk rounds. And in a seven round SWISS that means at most 5 competitive games to get ratings correctly adjusted and prizes fairly awarded.
But an A and B 2-point permanent acceleration gives 7 competitive games and fairness in the B Division prizes.

The maximum number of entries allowed there is usually 120 which is under the 128 that a 7 round normal swiss caters for anyway.

A 7-round SWISS finds a winner eclectic. But what it fails to do is a) give 7 data points for rating, and b) fair prizes in B Division.

It would seem that 2 junk rounds were already introduced to have competitors at their "proper" level by round 3.

Not sure what you meant here.

I feel the need to weep at those poor high rated players who suffer early round upsets at the hands of Box Hill ( or visiting ) juniors or micro juniors.

Come on electic, name names and name games. Let us have a full examination of this theory of yours. How many? Who? When?

Never mind !! They're only junk rounds !! Just write them off !!

Not at Box Hill mate. You don’t get junk rounds. We don’t have to write off any results; we send all results to Bill.

Actually,
What makes a game a junk round ?
The opponent ?
Or your attitude to the game ?

A junk round is where the mean absolute deviation of pairing differences is higher than 400. See the thread, ‘How competitive do you want it to be’ for examples and analysis.

And they are junk because they (the whole round) rarely add one iota to ratings reliability and they mean a missed opportunity to have the prize distribution based on reliable data, rather than a fluke pairing in round 7.

starter

Garvinator
26-02-2004, 01:50 PM
also a junk round to me could also be defined as- a round where the result of every game in that round is almost a formality.

for instance- in some monster swiss first rounds, most of the pairings in that first round have a rating differential of 500 points. this almost always means that each game's results are pretty well known before the round is even played. The higher rated player didnt get to test his/her skills in the first round game and the lower rated player got squashed. So that is no fun for either player.

ursogr8
26-02-2004, 02:22 PM
also a junk round to me could also be defined as- a round where the result of every game in that round is almost a formality.

for instance- in some monster swiss first rounds, most of the pairings in that first round have a rating differential of 500 points. this almost always means that each game's results are pretty well known before the round is even played. The higher rated player didnt get to test his/her skills in the first round game and the lower rated player got squashed. So that is no fun for either player.

Yes ggrayggray, what you describe is exactly what a tournament designer seeks to avoid because he knows that makes the event relativity unattractive. I agree with all you say,except the metric choice. I use 400 as the distinguishing M.A.D. of pairings to declare a junk round.
starter

ps And BTW I notice the post differential is under 100 for the first time.