Page 1 of 5 123 ... LastLast
Results 1 to 15 of 67
  1. #1
    CC FIDE Master
    Join Date
    Jan 2004
    Posts
    762
    Welcome back IM Greg Hjorth.

    Does his 75% score after 2 rounds in Melbourne adjust his ACF rating from 2423?? to 2225?? (down 198!).

    See Barry’s Glicko calculator

    Glicko

  2. #2
    CC Candidate Master
    Join Date
    Jul 2005
    Posts
    247
    Quote: Does his 75% score after 2 rounds in Melbourne adjust his ACF rating from 2423?? to 2225?? (down 198!).

    Yes it does.

    For my opinion on the Glicko system see my interview with Elliot Renzies at http://www.chesschat.org/showthread.php?t=5069&page=10

    The Glicko is too volatile. It implies huge fluctuations in playing strength at the top level that simply don't exist. Yes, form does fluctuate somewhat, but only somewhat, not by hundreds of points at IM level. If Greg had lost that game, and let's not forget Voon is quite a tricky 1888 player, Greg would have lost about 400 points! Hjorth, on the basis of 2 games would be rated about 2040. Is that realistic? Is that sensible? I would say not.

    The Glicko probably helps underrated juniors get rated more accurately faster, (though it still doesn't seem to work all that well) but distorts strengths at the top end. In my opinion the top end is more important to have accurate as it can affect important things like Olympiad selections that can change whole careers.

    Imagine if the ATP tennis rankings were pretty good down in the hundreds, but the top 20 list was often out of whack due to bizarre fluctuations based on a small sample of matches. People would get very upset because there is a lot at stake, tournament invitations and the like. Chess is the same. Some of our top players are professionals and should not be placed at the mercy of extreme volatility.

    I understand the argument that it is self correcting, but all that means is that it will be wrong in either direction a roughly equal amount. I prefer a sytem that ensures a better chance of the top players ratings being about right. Yes, "about right" is subjective, but I mean a better predictor of how they will perform in the next few months. I don't want to be told that player X just had one terrible or one brilliant result, I want to be told roughly how strong they are on average.

    I'm not having a go at Bill, who has the rating system running very efficiently and well, I'm just saying that I have a strong preference for a less volatile system like the one that preceded the Glicko. Just like I have a strong preference for longer time limits rather than FIDE's new faster ones. It's my opinion only. I'm not going to get involved in yet another debate about the mathematics of the Glicko... my expertise is in chess, not maths. The Glicko may be the most beautiful mathematical creature in the universe, but it is less accurate in reflecting the relative strengths of the top players. I guess many players think that doesn't matter, whereas I think it does.

  3. #3
    CC Candidate Master Sutek's Avatar
    Join Date
    Jan 2004
    Posts
    189
    Quote Originally Posted by Gattaca
    Quote: Does his 75% score after 2 rounds in Melbourne adjust his ACF rating from 2423?? to 2225?? (down 198!).

    Yes it does.

    For my opinion on the Glicko system see my interview with Elliot Renzies at http://www.chesschat.org/showthread.php?t=5069&page=10

    The Glicko is too volatile. It implies huge fluctuations in playing strength at the top level that simply don't exist. Yes, form does fluctuate somewhat, but only somewhat, not by hundreds of points at IM level. If Greg had lost that game, and let's not forget Voon is quite a tricky 1888 player, Greg would have lost about 400 points! Hjorth, on the basis of 2 games would be rated about 2040. Is that realistic? Is that sensible? I would say not.

    The Glicko probably helps underrated juniors get rated more accurately faster, (though it still doesn't seem to work all that well) but distorts strengths at the top end. In my opinion the top end is more important to have accurate as it can affect important things like Olympiad selections that can change whole careers.

    Imagine if the ATP tennis rankings were pretty good down in the hundreds, but the top 20 list was often out of whack due to bizarre fluctuations based on a small sample of matches. People would get very upset because there is a lot at stake, tournament invitations and the like. Chess is the same. Some of our top players are professionals and should not be placed at the mercy of extreme volatility.

    I understand the argument that it is self correcting, but all that means is that it will be wrong in either direction a roughly equal amount. I prefer a sytem that ensures a better chance of the top players ratings being about right. Yes, "about right" is subjective, but I mean a better predictor of how they will perform in the next few months. I don't want to be told that player X just had one terrible or one brilliant result, I want to be told roughly how strong they are on average.

    I'm not having a go at Bill, who has the rating system running very efficiently and well, I'm just saying that I have a strong preference for a less volatile system like the one that preceded the Glicko. Just like I have a strong preference for longer time limits rather than FIDE's new faster ones. It's my opinion only. I'm not going to get involved in yet another debate about the mathematics of the Glicko... my expertise is in chess, not maths. The Glicko may be the most beautiful mathematical creature in the universe, but it is less accurate in reflecting the relative strengths of the top players. I guess many players think that doesn't matter, whereas I think it does.
    Hi Gattaca,

    The other absurd thing about Glicko is if you have not played for many years and have a “??” against your rating then you can also go up hundreds of points.

    fex. According to the master file my current rating is "2240??"

    I haven’t played any otb chess for years yet if I happen to beat you one game on my return my new rating is 2410!
    But it gets better, if I beat Ian Rogers one game on my return my new rating is 2582!

    What sort of b/s is that?

    How is possible to have not played chess for many years and suddenly be Grandmaster strength!

    But regardless if you one thinks Glicko is good or bad it definitely discourages anyone from ever making a come back.

    How can that be good for Australian chess?

  4. #4
    CC Grandmaster ER's Avatar
    Join Date
    Jul 2006
    Location
    Melbourne - Australia
    Posts
    11,681

    Sutek vs Rogers not on!

    Quote Originally Posted by Sutek
    But it gets better, if I beat Ian Rogers one game on my return my new rating is 2582!
    Hi Sutek, just in case you ddn't know GM Ian Rogers has retired from competitive Chess! However, Glicko or not, if you manage to beat IM Guy West after such a long period of Chess inactivity, you will deserve your new rating, do go for it!
    ACF 3118316
    FIDE 3201457

  5. #5
    Account Suspended
    Join Date
    Mar 2007
    Posts
    928
    How about you play Ian Rogers, his phone rings and you win by the mobile phone rule, gaining you the rating points.


    Quote Originally Posted by Sutek
    Hi Gattaca,

    The other absurd thing about Glicko is if you have not played for many years and have a “??” against your rating then you can also go up hundreds of points.

    fex. According to the master file my current rating is "2240??"

    I haven’t played any otb chess for years yet if I happen to beat you one game on my return my new rating is 2410!
    But it gets better, if I beat Ian Rogers one game on my return my new rating is 2582!

    What sort of b/s is that?

    How is possible to have not played chess for many years and suddenly be Grandmaster strength!

    But regardless if you one thinks Glicko is good or bad it definitely discourages anyone from ever making a come back.

    How can that be good for Australian chess?

  6. #6
    Monster of the deep Kevin Bonham's Avatar
    Join Date
    Jan 2004
    Posts
    37,468
    This same debate comes up a lot so I thought I'd give it its own thread. Actually Gattaca's post covers a couple of different issues - firstly the question of Glicko's treatment of returning players and secondly whether Glicko is too volatile in general.

    Firstly the answer to Peter's original question:

    Does his 75% score after 2 rounds in Melbourne adjust his ACF rating from 2423?? to 2225?? (down 198!).
    Strictly the answer is no, for two reasons:

    1. A player's rating is not calculated game by game in Glicko. It doesn't even make sense to speak of a player's "live rating" in Glicko the way you can in ELO (in which you can know what impact a game will have on a player's rating as soon as that game is played). Rather, a player's rating in Glicko is calculated period by period and in determining how much a player's rating is affected by a result, their opponent's performance in the period is taken into account (among other factors).

    2. The ACF no longer uses Glicko-1 (which is the basis for Barry's calculator) but uses an advanced variant of Glicko-2. Glicko-1 is only an approximation.

    Most returning players will play a full tournament on their return not just one or two games, and one of the safeguards in the ACF system is that a player can't overshoot. So if a returning player performs way below their rating through the whole event then at worst their new rating will be their performance rating for that event.

    The answer to the broader question, of whether it's possible for a returning player to lose lots of points from just one or two games is yes - in certain rare circumstances. (However often when specific examples of this have been claimed, the claims have been incorrect or exaggerated.) It can only happen if the returning player only has those one or two games processed in the period in which they return, which is unusual.

    Some players think that returning players should not be able to lose 200 or so points from a single bad event. They seem to believe that a rating that is decades out of date should be just as protected from evidence of its own inaccuracy as one that is current - the way FIDE has done it. Obviously from a predictive viewpoint this is nonsense - chessplayers know that out of date ratings are usually inaccurate and rusty players usually don't perform to their rating (although some do) so why do players expect that an out-of-date rating should carry the same weight as a fresh one?

    I note that Gattaca mentioned the ATP tennis ratings. The ATP list discards all evidence that is more than one year old. Win four grand slams in one year and go to #1 in the world, take three years off and you do not come back as #1 in the world again as Kasparov might if he suddenly came out of retirement. No, in tennis you would come back with no rating. Of course, chess strength changes much more slowly than tennis strength but why is it that chessplayers think ratings decades out of date should carry any more than a small amount of weight? If long-inactive players tend to perform way below their ratings for a while on their return then a lower rating will more accurately predict their performance than one that mainly reflects old glories.

    Of course, the problem is that under any rating system that treats inactive players accurately, they are likely to lose lots of points. Systems that don't respond rapidly to their new playing strength might mollycoddle more players who are sensitive about their ratings into returning, but at the cost of those players carrying unreasonably high ratings well into their return.

    I suggested a way round this before, and Peter didn't like it one iota. I debunked his objections to it here and heard no more from him on the issue. My proposal is that once a rating is more than five years out of date, call it an expired rating. It stays on the master list and in the system as an internal rating for calculation of ratings when the player returns in the normal Glicko way, but the player becomes officially unrated. That way, instead of the thought of loss of points acting as a deterrent, the player has an incentive to return - they would have a rating again instead of no longer having one. Also, a player would have an incentive not to stop playing in the first place - if they stopped for too long, their old rating would expire and they would become officially unrated.

    Also, if the number of games a player played on return is less than, say, five, perhaps don't publish their new rating (on the active list at least) until they've played more games on return. I'd say there's actually less confidence about the accuracy of the rating of a player who has played one game now and had a rating 20 years ago than there is about the rating of a new player who has played nine games. But these cases are very very rare. Mostly the problem is that a player plays a whole tournament 300 points below their old rating, having not touched a pawn in 20 years, and thinks they should only lose about 30 points.

    In terms of predicting performance, I have not seen any data showing whether so-called "glickoed" ratings (those that drop 100+ points following the first ratings period after inactivity) tend to overstate the damage. There is a theoretical argument that they might, because a player is less rusty a few months after return than immediately after. (A case in point is Julian Steward who lost c. 150 points from a couple of rusty tournaments on return after not playing for 14 years. In the next period he won two tournaments and recovered all his losses - but I suspect on that evidence that he is actually now still underrated.) However even if the Glicko approach might overdo it a bit (and it might not), I'm certain it would be better at predicting the returning player's new strength than the ELO approach.

    As for the separate issue of whether Glicko is too volatile in general, this parallels the debate (if you can call it that) between Macieja and Sonas and many others on one hand, and Nunn and some others on the other, about whether the FIDE K-factor should be increased as foreshadowed, or whether it should remain at 10. Those arguing for increasing it are winning because it is apparently true that a higher k-factor makes for a rating that more accurately predicts a player's scores in the next ratings period.

    There are those who suggest a rating should measure some other kind of "strength" that is rather more long-term than that, and that for these purposes systems that are less dynamic might be more appropriate at least at the top of a global or local pool. That's especially the case if it's more important to you to prevent a flash-in-the-pan from being world #1 than it is to prevent a player past their peak from losing that status promptly. So there is sometimes a pragmatic argument for using lower k-factors at the top of a rating list, even if it makes the list a bit less accurate. Indeed, the ACF (though not for this reason) already has a degree of protection from rapid rating change for the higher rated players, compared with the rest of the pool.

    I set up a thread for discussing the question of objectives of rating systems over here, following the correct point by Rincewind that progress in these debates is not possible when people are not explicit about what they believe a rating system is for.

    There were a number of helpful suggestions, but those who prefer ELO's treatment of inactive players to Glicko's were conspicuous in their failure to visit the thread and state the basis of their views about what a rating system is actually supposed to do.

    To summarise my views about this, a rating should measure how good you are at the time it was issued. It should not measure how good you were a number of decades ago, and it should not measure how good you might be in a year after you shake off the rust.
    Last edited by Kevin Bonham; 07-05-2009 at 12:36 AM.

  7. #7
    Illuminati Bill Gletsos's Avatar
    Join Date
    Jan 2004
    Location
    Sydney
    Posts
    16,544
    Greg Hjorth has not played a ACF rated game since June 1993.
    Sutek (Stephen Kerr) hasnt played a ACF rated game since December 1991.

    Their respective ratings are over 15 and 17 years old.

    In fact with that in mind I think I will add 3 new sysmbols to the rating lists.

    A ??? indicates your rating is essentially meaninmgless as an indicator of your current strength.

    A ?*4 indicates your rating is totally meaningless as an indicator of your current strength.

    A ?*∞ indicates your rating is absolutely meaningless as an indicator of your current strength.
    The Force can have a strong influence on the weak-minded.

  8. #8
    Monster of the deep Kevin Bonham's Avatar
    Join Date
    Jan 2004
    Posts
    37,468


    And one could add:

    ?*∞*∞ means your rating is Peter Parr's.

  9. #9
    CC Grandmaster ER's Avatar
    Join Date
    Jul 2006
    Location
    Melbourne - Australia
    Posts
    11,681

    is he on the way back?

    Quote Originally Posted by Bill Gletsos
    Sutek (Stephen Kerr) hasnt played a ACF rated game since December 1991.
    Can we hope that all of the above discussion can be interpreted as an indication of Stephen Kerr's possible comeback to OTB chess?
    The return to competitive Chess of the former Aus and British Champ as well as a CC IM Stephen Kerr, will definitely be one of the highlights of the year!
    ACF 3118316
    FIDE 3201457

  10. #10
    CC Grandmaster ER's Avatar
    Join Date
    Jul 2006
    Location
    Melbourne - Australia
    Posts
    11,681
    CORRECTION CORRECTION CORRECTION
    Quote Originally Posted by justaknight
    Can we hope that all of the above discussion can be interpreted as an indication of Stephen Kerr's possible comeback to OTB chess?
    The return to competitive Chess of the former Aus and British Junior Champ as well as a CC IM Stephen Kerr, will definitely be one of the highlights of the year!
    ACF 3118316
    FIDE 3201457

  11. #11
    CC Rookie
    Join Date
    Feb 2008
    Posts
    17

    Please no nustiness!!

    Quote Originally Posted by Kevin Bonham


    And one could add:

    ?*∞*∞ means your rating is Peter Parr's.
    A lot of former chess players want to come back and play and support chess.
    Do not discourage them by childish remarks please.

  12. #12
    Monster of the deep Kevin Bonham's Avatar
    Join Date
    Jan 2004
    Posts
    37,468
    Quote Originally Posted by Tzoglanis
    A lot of former chess players want to come back and play and support chess.
    How are my comments (I'll ignore the unsubstantiated maturity flame) discouraging former chess players from coming back and playing?

    Actually what my comments are doing is discouraging former players from sitting on long-inactive ratings, believing that those ratings still mean something and expecting anybody else to care.

    I am encouraging inactive players who believe that their old ratings are a meaningful sign of their chess strength to get back into the ring and prove it over the board and I am supporting proposals that I believe do that without compromising the accuracy of the ratings.

  13. #13
    CC Grandmaster
    Join Date
    Nov 2005
    Location
    Melbourne
    Posts
    5,531
    Let me add two points:

    1. Player who has been inactive for a long time might play much weaker in the first few tournaments. However, his/her old rating is a good indicator of the "danger" level.

    2. If a 2200 player loses 500 points in the first two tournaments and gets "very reliable" rating of 1700, how long does he have to perform at 2200 to approach this rating? Much more then just a couple of tournaments.
    For private coaching (IM, four times VIC champion) call or SMS 0417519733
    Computer tells you what to play. Good coach explains why.

  14. #14
    CC Grandmaster ER's Avatar
    Join Date
    Jul 2006
    Location
    Melbourne - Australia
    Posts
    11,681

    uncalled for!

    Quote Originally Posted by Tzoglanis
    A lot of former chess players want to come back and play and support chess.
    Do not discourage them by childish remarks please.
    Tzoglanis, with all due respect your remark was out of place and uncalled for! Kev's bit was a tongue-firmly-in-cheek comment which would make Peter himself smile! Furthermore, I do not believe that former chess players' return to chessl could be in any way be hindered by "childish remarks"!
    ACF 3118316
    FIDE 3201457

  15. #15
    CC Grandmaster Basil's Avatar
    Join Date
    Mar 2006
    Location
    Subtropical Brisbane, Australia
    Posts
    11,231
    Quote Originally Posted by Tzoglanis
    A lot of former chess players want to come back and play and support chess.
    I don't think so. Where are they?

    Quote Originally Posted by Tzoglanis
    Do not discourage them by childish remarks please.
    No. Childish would be changing your nic in the quote to 'Tockly'. Kev has summed up his counter-position more than adequately to deal with your evacuation. I just wanted to post the word tockley (childishly).

    Carry on everyody - you're all doing very well.
    There is no cure for leftism. Its infestation of the host mostly diminishes with age except in the most rabid of specimens.

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Similar Threads

  1. ACF June 2005 Ratings
    By Bill Gletsos in forum Ratings Arena
    Replies: 136
    Last Post: 19-06-2005, 08:44 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •