PDA

View Full Version : Underrated Juniors



Pages : [1] 2 3

Paul S
18-04-2004, 05:31 PM
I experienced first-hand the problem (which I had tended to previously dismiss as whingeing by "sore loser" chess players) of underrated juniors at the Doeberl. Despite what has been previousy said (implied), this problem of underrated Juniors is NOT confined to just the ACT!

I played in the Minor at the Doeberl, and was seeded #20, so theoretically I should have ended up with 4.5/7 (instead of my actual score of 3.5/7).

My round 1 game was against Jessica Kinder (I think she is about 12 to 13 years old) from QLD, who was (supposedly!) only rated 933. As such, when I saw the rating difference between us (I'm 1432), I thought "OK, I just need to go through the motions here and I will win"! However, after "going through the motions (eg playing my moves too quickly)" until realising by about move 25 that Jessica was (by the way she played) in reality about 1300-1350, my game was lost, and Jessica had a well deserved win (well done, young Jessica!). BTW, Jessica ended up with 4/7.

I won my round 2 game against an ACT Junior (Aidan Lloyd - a boy of about 10) who was supposedly only 992 (but in reality was more like 1250-1300). It took me a while to "finish him off" (the "decisicive moment" came after about 60 moves). BTW, Aidan ended up with 3/7.

In my round 6 game, I played Eugene Schon from VIC who was supposedly rated 853. Well, for a player who is supposedly about 600 points weaker than me, I did well to get a draw!!! Here is my game with young Eugene (a boy of about 10) - tell me BB readers, is this the sort of game that one would expect a player with a (supposedly) 853 rating to play? White (Eugene Schon) vs Black (Myself) 1. e4 e5 2. Nf3 d6 3. d4 Bg4 4. de BxN 5. Qx B de 6. Bc4 Qe7 7. Bd2 Nd7 8. Qb3 Nc5 9. Qf3 0-0-0 10. Nc3 c6 11. 0-0-0 f6 12. Be3 h5 13. BxN QxB 14. Qf5+ Kc7 15. RxR KxR 16 Rd1+ Bd6 17. Qe6 Kc7 18. Qf7+ Ne7 19. Q x g7 Rd8 20. Bf7 Qxf2 21. Bxh5 Rg8 22. Qf7 Rxg2 23 Be2 b5 24. a3 Rxh2 25. Qe6 Qe3+ 26 Kb1 Qc5 27 Bg4 f5 28. ef Nd5 29. NxN c6xN 30. Rc1 a5 31. f6 b4 32. c4 ba 33. Qd7+ Kb6 34. ba d4 35. Bf3 Qc7 36. Qb5+ Ka7 37. c5 Qb8 38. Qb6+ QxQ 39. c5x Q+ Kb8 40. a4 Rf2 41. Rc6 RxB 42. RxB Kb7 43. Re6 e4 44. Rxe4 Rxf6 45 Rxd4 Kxb6 46. Ka2 Rf3 47. Rd5 Rf2+ Draw agreed (1/2 - 1/2). BTW, Eugene ended up with 4/7 at the Doeberl. IMHO the game he played was what I would have expected from someone rated around 1500-1600 (not 853!!!).

In round 7 I played David Jaksa of ACT (about 13-14 years old I think) who was supposedly rated 1051. After about 45 moves we agreed on a draw. From the way he played it seemed as though his true rating was around 1400-1450. Indeed when I said to David afterwards "you're a lot better than 1051", he told me that he had recently beaten a few players rated over 1500! BTW, David ended up with 3.5/7 at the Doeberl.

Fortunately I don't care (too much!) about my rating these days! However, my experiences with playing underrated Juniors at the 2004 Doeberl (which BTW will not affect whether or not I play in the 2005 Doeberl) indicates to me that there is something wrong with the rating system w.r.t. Juniors Australia wide (ie not just the ACT!), despite the ratings being updated nowadays every 3 months.

Bill Gletsos
18-04-2004, 06:00 PM
I experienced first-hand the problem (which I had tended to previously dismiss as whingeing by "sore loser" chess players) of underrated juniors at the Doeberl. Despite what has been previousy said (implied), this problem of underrated Juniors is NOT confined to just the ACT!

I played in the Minor at the Doeberl, and was seeded #20, so theoretically I should have ended up with 4.5/7 (instead of my actual score of 3.5/7).

My round 1 game was against Jessica Kinder (I think she is about 12 to 13 years old) from QLD, who was (supposedly!) only rated 933. As such, when I saw the rating difference between us (I'm 1432), I thought "OK, I just need to go through the motions here and I will win"! However, after "going through the motions (eg playing my moves too quickly)" until realising by about move 25 that Jessica was (by the way she played) in reality about 1300-1350, my game was lost, and Jessica had a well deserved win (well done, young Jessica!). BTW, Jessica ended up with 4/7.

I won my round 2 game against an ACT Junior (Aidan Lloyd - a boy of about 10) who was supposedly only 992 (but in reality was more like 1250-1300). It took me a while to "finish him off" (the "decisicive moment" came after about 60 moves). BTW, Aidan ended up with 3/7.

In my round 6 game, I played Eugene Schon from VIC who was supposedly rated 853. Well, for a player who is supposedly about 600 points weaker than me, I did well to get a draw!!! Here is my game with young Eugene (a boy of about 10) - tell me BB readers, is this the sort of game that one would expect a player with a (supposedly) 853 rating to play? White (Eugene Schon) vs Black (Myself) 1. e4 e5 2. Nf3 d6 3. d4 Bg4 4. de BxN 5. Qx B de 6. Bc4 Qe7 7. Bd2 Nd7 8. Qb3 Nc5 9. Qf3 0-0-0 10. Nc3 c6 11. 0-0-0 f6 12. Be3 h5 13. BxN QxB 14. Qf5+ Kc7 15. RxR KxR 16 Rd1+ Bd6 17. Qe6 Kc7 18. Qf7+ Ne7 19. Q x g7 Rd8 20. Bf7 Qxf2 21. Bxh5 Rg8 22. Qf7 Rxg2 23 Be2 b5 24. a3 Rxh2 25. Qe6 Qe3+ 26 Kb1 Qc5 27 Bg4 f5 28. ef Nd5 29. NxN c6xN 30. Rc1 a5 31. f6 b4 32. c4 ba 33. Qd7+ Kb6 34. ba d4 35. Bf3 Qc7 36. Qb5+ Ka7 37. c5 Qb8 38. Qb6+ QxQ 39. c5x Q+ Kb8 40. a4 Rf2 41. Rc6 RxB 42. RxB Kb7 43. Re6 e4 44. Rxe4 Rxf6 45 Rxd4 Kxb6 46. Ka2 Rf3 47. Rd5 Rf2+ Draw agreed (1/2 - 1/2). BTW, Eugene ended up with 4/7 at the Doeberl. IMHO the game he played was what I would have expected from someone rated around 1500-1600 (not 853!!!).

In round 7 I played David Jaksa of ACT (about 13-14 years old I think) who was supposedly rated 1051. After about 45 moves we agreed on a draw. From the way he played it seemed as though his true rating was around 1400-1450. Indeed when I said to David afterwards "you're a lot better than 1051", he told me that he had recently beaten a few players rated over 1500! BTW, David ended up with 3.5/7 at the Doeberl.

Fortunately I don't care (too much!) about my rating these days! However, my experiences with playing underrated Juniors at the 2004 Doeberl (which BTW will not affect whether or not I play in the 2005 Doeberl) indicates to me that there is something wrong with the rating system w.r.t. Juniors Australia wide (ie not just the ACT!), despite the ratings being updated nowadays every 3 months.
If we look at the rating history of the players you mention there is no indication that they are significantly stronger than their ratings with the exception of Schon.


Results For Player Aidan Lloyd Id number = 1291833
Current Rating = 992 Current Rapid Rating = Unr

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Aug 2001 Unr 208 2.5 7
Dec 2001 77 -241 0.0 3
Apr 2002 132 415 1.5 2
Aug 2002 329 759 4.5 8
Dec 2002 573 705 10.0 24
Mar 2003 574 589 1.0 5
Jun 2003 576 578 10.5 28
Sep 2003 588 605 8.0 23
Nov 2003 897 0 0.0 0
Dec 2003 955 1006 14.5 30
Mar 2004 992 652 2.0 6
Based on previous results there is no indication Aidan is underrated.


Results For Player David Jaksa Id number = 1292894
Current Rating = 1051 Current Rapid Rating = 559

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Dec 2002 Unr 604 1.0 1
Jun 2003 757 757 5.5 11
Sep 2003 745 740 11.0 24
Nov 2003 1054 0 0.0 0
Dec 2003 981 921 9.0 18

Players Historical Rating Performance Rapid
Period Rating Perf Score Games
Mar 2003 Unr 391 2.5 8
Mar 2004 559 635 5.0 12

Again based on previous results there is no indication David is underrated.



Results For Player Jessica Kinder Id number = 4175671
Current Rating = 933 Current Rapid Rating = 876

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Apr 2001 Unr 584 1.0 5
Aug 2001 479 388 1.5 13
Dec 2001 567 650 4.5 22
Apr 2002 577 638 1.0 5
Aug 2002 711 865 8.0 19
Dec 2002 798 914 9.0 24
Jun 2003 723 70 1.0 4
Nov 2003 751 0 0.0 0
Dec 2003 775 798 11.0 22
Mar 2004 933 1094 7.5 16

Players Historical Rating Performance Rapid
Period Rating Perf Score Games
Dec 2000 493 493 2.5 11
Aug 2001 568 586 15.0 38
Dec 2001 749 799 24.5 52
Apr 2002 771 877 2.5 11
Aug 2002 828 876 25.5 42
Dec 2002 866 983 10.0 17
Jun 2003 879 902 13.0 21
Sep 2003 793 642 9.5 19
Dec 2003 836 917 12.0 19
Mar 2004 876 659 2.0 6
There is no real evidence that Jessica is underrated. Prior to March 2004 she had never performed near the 1000 mark let alone the 1300. Even in the March period she only performed at the 1094 level over 16 games.
Her rating of 933 is reasonable based on her results.


Results For Player Eugene Schon Id number = 3115450
Current Rating = 853 Current Rapid Rating = 741

Players Historical Rating Performance Normal
Period Rating Perf Score Games
Mar 2004 853 1169 11.0 20

Players Historical Rating Performance Rapid
Period Rating Perf Score Games
Dec 2001 Unr 174 3.5 7
Apr 2002 237 279 4.0 7
Aug 2002 218 202 6.0 14
Dec 2002 350 458 15.0 29
Mar 2003 355 378 3.5 7
Jun 2003 473 732 8.5 14
Sep 2003 552 725 14.5 21
Dec 2003 567 621 11.0 21
Mar 2004 741 1143 11.0 21
Anybody who is really observant will realise that Eugene is the player starter was referring to in the March 2004 ratings thread.
Eugene may be underrated or he may not be. Prior to his March 2004 rating there was no indication he was near 1100. Its possible that result was an anomaly or that the Glicko just cannot correct him sufficently in just 1 rating period.

jay_vee
18-04-2004, 06:09 PM
Have these juniors played in tournaments against adults, or are they mainly playing amongst themselves (i.e. against other presumably under-rated juniors)?

jenni
18-04-2004, 06:18 PM
Have these juniors played in tournaments against adults, or are they mainly playing amongst themselves (i.e. against other presumably under-rated juniors)?

Aidan Lloyd (who is turning 14 this year) and David Jaksa both play at Belconnen Chess Club. Thus they do play a number of adults, although a lot of other juniors as well (because of the number at BCC). Aidan plays all the JUnior weekenders as well.

Aidan does seem to perform well against adults - he got quite good results last year in Queensland. He is a bit erratic and often loses to juniors, so his rating is probably a bit lower than it could be.

I think the problem is the juniors are always going to seem under-rated, as they are improving and their rating lags their performance. At the moment I feel Glicko 2 is bringing improving juniors to their correct rating faster. However it is never going to be accurate, if a junior is rapidly improving. Both Jessica Kinder and Eugene Schon fit into the rapidly improving slot. I also think Juniors are more fiery and really want to win, so are alwasy harder to play - I would much rather play a 1200 adult than an 800 junior.

Bill Gletsos
18-04-2004, 06:30 PM
Have these juniors played in tournaments against adults, or are they mainly playing amongst themselves (i.e. against other presumably under-rated juniors)?
That issue was taken into consideration when the ACT juniors ratings and anyone interstate players they played were corrected back in November 2003.
Anyone in the list with a rating showing as Nov 2003 was a correction.

Paul S
18-04-2004, 06:31 PM
Aidan Lloyd (who is turning 14 this year) .......


Sorry, Aidan (assuming you read this BB). It seems that you are a few years older than I thought you were! :lol:


I also think Juniors are more fiery and really want to win, so are alwasy harder to play - I would much rather play a 1200 adult than an 800 junior.

Yes, I too would prefer to play a 1200 adult (reliable rating) than an 800 Junior (unreliable rating!). I like to know what I am up against! :lol:

Bill Gletsos
18-04-2004, 06:37 PM
I also think Juniors are more fiery and really want to win, so are alwasy harder to play - I would much rather play a 1200 adult than an 800 junior.
Actually i think its easier to play a 1200 rated adult than a 1200 rated junior.

Also the younger the juniors are the more erratic they are.

Garvinator
18-04-2004, 06:39 PM
how about not looking at the rating of your opponent before you play your game?

Paul S
18-04-2004, 06:47 PM
If we look at the rating history of the players you mention there is no indication that they are significantly stronger than their ratings with the exception of Schon.

Based on previous results there is no indication Aidan is underrated.

Again based on previous results there is no indication David is underrated.

There is no real evidence that Jessica is underrated. Prior to March 2004 she had never performed near the 1000 mark let alone the 1300. Even in the March period she only performed at the 1094 level over 16 games.
Her rating of 933 is reasonable based on her results.


:hmm: Perhaps. :hmm:

Nonetheless, that was not my experience in playing these players! Has anyone else played any of these Juniors (Jessica Kinder, Aidan Lloyd and David Jaksa) recently?


Anybody who is really observant will realise that Eugene is the player starter was referring to in the March 2004 ratings thread.
Eugene may be underrated or he may not be. Prior to his March 2004 rating there was no indication he was near 1100. Its possible that result was an anomaly or that the Glicko just cannot correct him sufficently in just 1 rating period.

I guess I wasn't that observant! Mind you, I don't read every post on this BB! :lol:

I have put up the game I played with Eugene (see my original post in this thread). Again I ask BB readers, is this the game one would expect a (supposedly) 853 rated player to play?

Garvinator
18-04-2004, 06:54 PM
to help Paul S out ;) :


White (Eugene Schon) vs Black (Paul S)

1. e4 e5 2. Nf3 d6 3. d4 Bg4 4. dxe5 Bxf3 5. Qxf3 dxe5 6. Bc4 Qe7 7. Bd2 Nd7 8. Qb3 Nc5 9. Qf3 0-0-0 10. Nc3 c6 11. 0-0-0 f6 12. Be3 h5 13. Bxc5 Qxc5 14. Qf5+ Kc7 15. Rxd8 Kxd8 16. Rd1 Bd6 17. Qe6 Kc7 18. Qf7 Ne7 19. Qxg7 Rd8 20. Bf7 Qxf2 21. Bxh5 Rg8 22. Qf7 Rxg2 23. Be2 b5 24. a3 Rxh2 25. Qe6 Qe3+ 26. Kb1 Qc5 27. Bg4 f5 28. exf5 Nd5 29. Nxd5 cxd5 30. Rc1 a5 31. f6 b4 32. c4 bxa3 33. Qd7+ Kb6 34. bxa3 d4 35. Bf3 Qc7 36. Qb5+ Ka7 37. c5 Qb8 38. Qb6+ Qxb6 39. cxb6 Kb8 40. a4 Rf2 41. Rc6 Rxf3 42. Rxd6 Kb7 43. Re6 e4 44. Rxe4 Rxf6 45. Rxd4 Kxb6 46. Ka2 Rf3 47. Rd5 Rf2+

Bill Gletsos
18-04-2004, 06:57 PM
:hmm: Perhaps. :hmm:

Nonetheless, that was not my experience in playing these players! Has anyone else played any of these Juniors (Jessica Kinder, Aidan Lloyd and David Jaksa) recently?
The point is you can only rate players based on their results and not the opinion of their vanquished opponents. ;)


I guess I wasn't that observant! Mind you, I don't read every post on this BB! :lol:
Shame on you. :hand:


I have put up the game I played with Eugene (see my original post in this thread). Again I ask BB readers, is this the game one would expect a (supposedly) 853 rated player to play?
At the very best he would only be rated 1169. You would still out rate him by 263 points. As jenni pointed out, some juniors just improve rapidly. There is not you can do to cater for them.

Bill Gletsos
18-04-2004, 06:59 PM
to help Paul S out ;)
Unfortunately most of it is not valid pgn hence it stops working.
e.g things like RxR and BxN etc.

Ian Rout
18-04-2004, 07:00 PM
One thing to remember is that a rating measures average performance, combining good and bad days. Juniors will tend to play at their peak for a special event like the Doeberl and not so hard at ordinary club or junior events.

Garvinator
18-04-2004, 07:04 PM
Unfortunately most of it is not valid pgn hence it stops working.
e.g things like RxR and BxN etc.

and i have been correcting the notation, but it wont allow me pass 16. Rd1 ??

Bill Gletsos
18-04-2004, 07:07 PM
to help Paul S out ;) :


White (Eugene Schon) vs Black (Paul S)

1. e4 e5 2. Nf3 d6 3. d4 Bg4 4. dxe5 Bxf3 5. Qxf3 dxe5 6. Bc4 Qe7 7. Bd2 Nd7 8. Qb3 Nc5 9. Qf3 0-0-0 10. Nc3 c6 11. 0-0-0 f6 12. Be3 h5 13. Bxc5 Qxc5 14. Qf5+ Kc7 15. Rxd8 Kxd8 16. Rd1 Bd6 17. Qe6 Kc7 18. Qf7 Ne7 19. Qxg7 Rd8 20. Bf7 Qxf2 21. Bxh5 Rg8 22. Qf7 Rxg2 23. Be2 b5 24. a3 Rxh2 25. Qe6 Qe3+ 26. Kb1 Qc5 27. Bg4 f5 28. exf5 Nd5 29. Nxd5 cxd5 30. Rc1 a5 31. f6 b4 32. c4 bxa3 33. Qd7+ Kb6 34. bxa3 d4 35. Bf3 Qc7 36. Qb5+ Ka7 37. c5 Qb8 38. Qb6+ Qxb6 39. cxb6+ Kb8 40. a4 Rf2 41. Rc6 Rxf3 42. Rxd6 Kb7 43. Re6 e4 44. Rxe4 Rxf6 45. Rxd4 Kxb6 46. Ka2 Rf3 47. Rd5 Rf2+

This should be correct.

Bill Gletsos
18-04-2004, 07:08 PM
and i have been correcting the notation, but it wont allow me pass 16. Rd1 ??
You have full stops missing from behind some of the numbers like 16, 23 and 45 to name a few.

Alan Shore
18-04-2004, 07:10 PM
how about not looking at the rating of your opponent before you play your game?

I've always been curious about the psychological effect of knowing an opponent's rating.. for some, if they see a low rated opponent they might take the game easier and not concentrate as they should have, or perhaps give no respect and just crush. If they meet a higher opponent they may abandon hope of winning or conversely lift their game and be determined to claim a high-rated scalp.

Having said that, I'd still prefer to know the strength of my opponent.. the more you know about your enemy the better chance you have of defeating him ;)

Alan Shore
18-04-2004, 07:12 PM
Thanks for the pgn correction Bill, was about to do it myself :)

Garvinator
18-04-2004, 07:24 PM
the more you know about your enemy the better chance you have of defeating him ;)
or her :hmm:

Bill Gletsos
18-04-2004, 07:33 PM
Well Paul S's problem is obvious. As others have noted in the analysis threads it is not wise to play Philidors. :whistle:

Alan Shore
18-04-2004, 08:25 PM
Well Paul S's problem is obvious. As others have noted in the analysis threads it is not wise to play Philidors. :whistle:

Well, not that line at least.. However, 1. e4 e5 2. Nf3 d6!

ursogr8
18-04-2004, 08:48 PM
In my round 6 game, I played Eugene Schon from VIC who was supposedly rated 853. Well, for a player who is supposedly about 600 points weaker than me, I did well to get a draw!!! Here is my game with young Eugene (a boy of about 10) - tell me BB readers, is this the sort of game that one would expect a player with a (supposedly) 853 rating to play? White (Eugene Schon) vs Black (Myself) 1. e4 e5 2. Nf3 d6 3. d4 Bg4 4. de BxN 5. Qx B de 6. Bc4 Qe7 7. Bd2 Nd7 8. Qb3 Nc5 9. Qf3 0-0-0 10. Nc3 c6 11. 0-0-0 f6 12. Be3 h5 13. BxN QxB 14. Qf5+ Kc7 15. RxR KxR 16 Rd1+ Bd6 17. Qe6 Kc7 18. Qf7+ Ne7 19. Q x g7 Rd8 20. Bf7 Qxf2 21. Bxh5 Rg8 22. Qf7 Rxg2 23 Be2 b5 24. a3 Rxh2 25. Qe6 Qe3+ 26 Kb1 Qc5 27 Bg4 f5 28. ef Nd5 29. NxN c6xN 30. Rc1 a5 31. f6 b4 32. c4 ba 33. Qd7+ Kb6 34. ba d4 35. Bf3 Qc7 36. Qb5+ Ka7 37. c5 Qb8 38. Qb6+ QxQ 39. c5x Q+ Kb8 40. a4 Rf2 41. Rc6 RxB 42. RxB Kb7 43. Re6 e4 44. Rxe4 Rxf6 45 Rxd4 Kxb6 46. Ka2 Rf3 47. Rd5 Rf2+ Draw agreed (1/2 - 1/2). BTW, Eugene ended up with 4/7 at the Doeberl. IMHO the game he played was what I would have expected from someone rated around 1500-1600 (not 853!!!).

Fortunately I don't care (too much!) about my rating these days! However, my experiences with playing underrated Juniors at the 2004 Doeberl (which BTW will not affect whether or not I play in the 2005 Doeberl) indicates to me that there is something wrong with the rating system w.r.t. Juniors Australia wide (ie not just the ACT!), despite the ratings being updated nowadays every 3 months.


PaulS
Eugene Schon’s rating has already been subject of multiple posts on the ACF Ratings thread. Essentially, I asked Bill to investigate and he showed that the rapid ratings were stable for a few years in Box Hill Sunday events. This rapid rating was then used as an ‘initiator’ for the normal rating when Eugene graduated to longer games. However, this occurred just at the time that Eugene received intensive coaching to change his skill level; so, in fact the initiator was artificially low for the normal rating. Bill half-agrees that it would have been better in this instance to treat him as UNRATED for normal ratings. Bill has said that he will experiment.
Thus, there is no excuse for you PaulS; this player had featured on the BB already as a player to be carefully watched.
Also, you will find Eugene mentioned as a upset RISK in the Box Hill weekly bulletins that we publish on the web-site, and often here also. I think you should keep in touch with the rest of Australia instead of tending to Sydney-centric issues.

Next year PaulS, send me a PM before Doeberl and I will mail you a list of who to watch out for. (Or alternatively just ask jeffrei for a list of his 20 students).

starter

ursogr8
18-04-2004, 08:57 PM
Have these juniors played in tournaments against adults, or are they mainly playing amongst themselves (i.e. against other presumably under-rated juniors)?

jay_vee

In Victoria, the only rapid rated events that juniors can play in are senior events.
For normal rated events, with the exception the once/per/year VIC JUNIOR OPEN, the only rated events that Juniors can play in are SENIOR events.

In summary, the answer to your question is NO, for rated tournaments.

starter

ursogr8
18-04-2004, 09:05 PM
One thing to remember is that a rating measures average performance, combining good and bad days. Juniors will tend to play at their peak for a special event like the Doeberl and not so hard at ordinary club or junior events.
hi Ian

Your theory reminds me of that old saying where a passer-by says "Does that dog bite?" And the owner says "No, he does not bite". The passer-by says "Does the dog know that".

I wish some of the juniors in the current Box Hill Championship knew that they should not play so hard at 'just the BH Championship'.

starter

Alan Shore
18-04-2004, 09:25 PM
jay_vee

In Victoria, the only rapid rated events that juniors can play in are senior events.
For normal rated events, with the exception the once/per/year VIC JUNIOR OPEN, the only rated events that Juniors can play in are SENIOR events.

In summary, the answer to your question is NO, for rated tournaments.

starter

That surprises me, I thought the junior chess scene was quite big in Victoria what with Cordover, Jammo, Daz etc. operations happening.. how come there are so few junior tournaments? I think we're spoiled here in QLD, there seems to be on average one every week or two!

Cat
18-04-2004, 09:32 PM
Looks to me like you simply played crap Paul


Have these juniors played in tournaments against adults, or are they mainly playing amongst themselves (i.e. against other presumably under-rated juniors)?


Yes, exactly. This issue was originally raised on the old BB by Jonathon from Cambridge, UK. Glicko uses Bayesian methodology, which assumes the likelihood of one player playing any other is approximately equal. As you suggest, Juniors are essentially pooled in certain areas of Australia, so there is a greatly increased chance they will be playing another Junior. In addition, as juniors often play in Junior tournements, so the chances of playing another junior is further increased. In addition, juniors are more likely to be represented in the bottom half of an open draw.

These pooling effects create extreme rating distorsions amongst the the Junior population, and obviously is not factored into the Glicko calculations. I did explain all this on the BB last year, you guys must have short memories. I'm sure Bill remembers though, don't you Bill?

Cat
18-04-2004, 09:36 PM
What's happened to my Avator?

Bill Gletsos
18-04-2004, 10:01 PM
Yes, exactly. This issue was originally raised on the old BB by Jonathon from Cambridge, UK. Glicko uses Bayesian methodology, which assumes the likelihood of one player playing any other is approximately equal. As you suggest, Juniors are essentially pooled in certain areas of Australia, so there is a greatly increased chance they will be playing another Junior. In addition, as juniors often play in Junior tournements, so the chances of playing another junior is further increased. In addition, juniors are more likely to be represented in the bottom half of an open draw.

These pooling effects create extreme rating distorsions amongst the the Junior population,
Supposition on your part.
As usual you offered no actual proof.


and obviously is not factored into the Glicko calculations. I did explain all this on the BB last year,
Nope, no explanation just an unproven theory.


you guys must have short memories.
Yes, well one tries to forget crap. ;)


I'm sure Bill remembers though, don't you Bill?
I dont recall anyone agreeing with you then.

Cat
18-04-2004, 10:27 PM
Supposition on your part.
As usual you offered no actual proof.

Bill, if you actually opened your eyes for once, you'd clearly see the evidence for yourself. Which bit do you think is supposition?

79% of participator on the Gold Coast are junior -FACT
around 70% participants in Australia are adult - FACT

I don't know the statistics in Box Hill or ACT, but unless you are able to produce figure otherwise, stop whinging.

Juniors on the Gold Coast generally compete in Junior competition far more often than they do in open competition - FACT

Adults are more likely to be playing another adult at the ADULT club - FACT.

I have the results of the last 5 yrs of the Gold Coast Classic. Juniors are overwhelmingly represented in the lower half of the draw - FACT



Nope, no explanation just an unproven theory.

Oh, so Glicko now has a correction for closed systems does it? Now that's something I'd missed, perhaps you could enlighten us?


Yes, well one tries to forget crap. ;)

Loosing the plot again Bill? It doesn't take you long to slip into profanity does it? The typical response of the intellectually bankrupt.


I dont recall anyone agreeing with you then.

Isn't it ironic that 12/12 ago, when all this ratings stuff blew up, you denied then there was a problem. Since then there have been significant corrections, my own rating has gone up 200 points without playing a game, and yet you remain fossilised in this plane of denial? I actually think you did a good job on the Gold Coast, but unless you wake up and listen, the problems are set to return in 18 months or so.

Kevin Bonham
18-04-2004, 10:29 PM
I reckon White played this game at about 1400 strength and Black played at about 1200 and was very lucky to get away with a draw. Paul, your opening was horrible. 3...Bg4 is very dodgy and if you are going to play it at all the one half-decent move that gives Black some sort of game is 6...Qf6! Your 6...Qe7 is very average, but not as bad as the commonly played 6....Nf6?, which loses. (See: "One impressive game" thread for more gory details.)

So after being given a free +/- out of the opening and not doing too much wrong for a while your opponent got a stronger and stronger position with obvious moves and then when he was building up some pressure you ignored it with the totally extravagent 12...h5. He could have put you away on the spot for that with 18.Na4 and then again with 27.b4 and 36.c5+ (both moves a 1500-1600 player could well have seen.) He also missed 38.Qxb8+ and pushing f7, which quickly wins. So he missed four wins, three of which were not all that difficult to see - although a real 800 player probably wouldn't have even got positions that good in the first place.

You had a chance too - 20...Qxc3! winning a piece.

The stats and the game suggest that young Eugene may have just made the quantum leap. Some juniors do this, suddenly gain several hundred points of playing strength for no obvious reason over one ratings period. The system needs to wait for more evidence, because they could just as easily have played at their normal strength and been lucky over a small sample of games.

Cat
18-04-2004, 10:36 PM
I reckon White played this game at about 1400 strength and Black played at about 1200 and was very lucky to get away with a draw. Paul, your opening was horrible. 3...Bg4 is very dodgy and if you are going to play it at all the one half-decent move that gives Black some sort of game is 6...Qf6! Your 6...Qe7 is very average, but not as bad as the commonly played 6....Nf6?, which loses. (See: "One impressive game" thread for more gory details.)

So after being given a free +/- out of the opening and not doing too much wrong for a while your opponent got a stronger and stronger position with obvious moves and then when he was building up some pressure you ignored it with the totally extravagent 12...h5. He could have put you away on the spot for that with 18.Na4 and then again with 27.b4 and 36.c5+ (both moves a 1500-1600 player could well have seen.) He also missed 38.Qxb8+ and pushing f7, which quickly wins. So he missed four wins, three of which were not all that difficult to see - although a real 800 player probably wouldn't have even got positions that good in the first place.

You had a chance too - 20...Qxc3! winning a piece.

The stats and the game suggest that young Eugene may have just made the quantum leap. Some juniors do this, suddenly gain several hundred points of playing strength for no obvious reason over one ratings period. The system needs to wait for more evidence, because they could just as easily have played at their normal strength and been lucky over a small sample of games.

KB, there are just so many of these anomalies aren't there? Funny that. It does beg the question however, what handle does the ACF have on the demographics? Perhaps if we were to analyse these anomalies, we find find a pattern?

Alan Shore
18-04-2004, 10:49 PM
What's happened to my Avator?

Maybe the moderators didn't like Johnny.

I can't disguise that I'm pleased to see it gone though :whistle:

I mean, you don't see Machiavelli with a Mark Latham avatar.. that would be funny though :D

Bill Gletsos
18-04-2004, 11:07 PM
Bill, if you actually opened your eyes for once, you'd clearly see the evidence for yourself. Which bit do you think is supposition?
All of your statements.


79% of participator on the Gold Coast are junior -FACT
around 70% participants in Australia are adult - FACT

I don't know the statistics in Box Hill or ACT, but unless you are able to produce figure otherwise, stop whinging.

Juniors on the Gold Coast generally compete in Junior competition far more often than they do in open competition - FACT

Adults are more likely to be playing another adult at the ADULT club - FACT.

I have the results of the last 5 yrs of the Gold Coast Classic. Juniors are overwhelmingly represented in the lower half of the draw - FACT
So what.
All of that has nothing to do with ratings theory.


Oh, so Glicko now has a correction for closed systems does it? Now that's something I'd missed, perhaps you could enlighten us?
I never said there was a correction. You really arent very good at comprehension are you.
In fact why would one have a correction for your unsupported theory.


Loosing the plot again Bill? It doesn't take you long to slip into profanity does it? The typical response of the intellectually bankrupt.
If you are going to post unsubstantiated rubbish, that is hardly my fault.
All you are doing is peddling your own unsupported pet theory. :hand:



Isn't it ironic that 12/12 ago, when all this ratings stuff blew up, you denied then there was a problem. Since then there have been significant corrections, my own rating has gone up 200 points without playing a game, and yet you remain fossilised in this plane of denial?
Now you are just falsifing the data. The 70 point uplift has nothing to do with any correction. It is was done to try and a;ogn the ACF ratings with FIDE. Everybody got it. Its not a correction for any error or problem with Glicko.
Your other 120 point uplift in November was to correct perceived problems with Gold Coast, however as I pointed out before if we had been totally accurate we would have reduced the ratings of a significant number of players. We chose not to do this and allow over correction to be redistributed over future rating periods.


I actually think you did a good job on the Gold Coast, but unless you wake up and listen, the problems are set to return in 18 months or so.
Remind me in 18mths. :whistle:

Cat
18-04-2004, 11:16 PM
I see what your doing here Bill - filling the thread with drivel to divert attention away from from the real issue. Well as you've only responded to my original post with denial, I'll repost it to see if you can deliver a proper response.

'Yes, exactly. This issue was originally raised on the old BB by Jonathon from Cambridge, UK. Glicko uses Bayesian methodology, which assumes the likelihood of one player playing any other is approximately equal. As you suggest, Juniors are essentially pooled in certain areas of Australia, so there is a greatly increased chance they will be playing another Junior. In addition, as juniors often play in Junior tournements, so the chances of playing another junior is further increased. In addition, juniors are more likely to be represented in the bottom half of an open draw.

These pooling effects create extreme rating distorsions amongst the the Junior population, and obviously is not factored into the Glicko calculations. I did explain all this on the BB last year, you guys must have short memories. I'm sure Bill remembers though, don't you Bill?'

Bill Gletsos
18-04-2004, 11:28 PM
I see what your doing here Bill - filling the thread with drivel to divert attention away from from the real issue. Well as you've only responded to my original post with denial, I'll repost it to see if you can deliver a proper response.
Post it as often as you like, my response wont change.


'Yes, exactly. This issue was originally raised on the old BB by Jonathon from Cambridge, UK. Glicko uses Bayesian methodology, which assumes the likelihood of one player playing any other is approximately equal. As you suggest, Juniors are essentially pooled in certain areas of Australia, so there is a greatly increased chance they will be playing another Junior. In addition, as juniors often play in Junior tournements, so the chances of playing another junior is further increased. In addition, juniors are more likely to be represented in the bottom half of an open draw.
So what.
Where is your proof that this has any effect on a rating system, whether it be Elo, Glicko or something else.


These pooling effects create extreme rating distorsions amongst the the Junior population,
Where is your proof this causes distortions.
Just because you say so does not make it so.
All you are doing is pushing your own insubstatiated pet theory.



and obviously is not factored into the Glicko calculations.
Why would one factor in anything for some so called effect that has not been proven to occur, let alone have an effect on the rating system.



I did explain all this on the BB last year, you guys must have short memories. I'm sure Bill remembers though, don't you Bill?'
It was unproven rubbish back then.
It still is.

Cat
18-04-2004, 11:39 PM
Where is your proof that this has any effect on a rating system, whether it be Elo, Glicko or something else.

So now you're not disputing that the juniors are more likely to play each other, but simply questioning how this should effect the rating system.

ok, which of these issues would you object to;

1. Glicko uses Bayesian methodology.
2. Bayesian methodology assumes that all combinations are equally likely - that is that any player is just as likely to play any other.
3. Bayesian methodology alone is inadequate for accurately predicting ratings in a semi-closed system, ie when a player is more likely to play certain players repeatedly.

Garvinator
18-04-2004, 11:58 PM
i have a question for david, what do you propose to alleviate/solve these 'problems'?

Cat
19-04-2004, 12:07 AM
i have a question for david, what do you propose to alleviate/solve these 'problems'?

Garvin, I've made many suggestions, Bill's full cogniscent of many options he has available, but it's become a matter of honour to him to solve these issues within the Glicko system.

Bill Gletsos
19-04-2004, 12:09 AM
So now you're not disputing that the juniors are more likely to play each other,
How do you draw this conclusion from my statements above.


but simply questioning how this should effect the rating system.

ok, which of these issues would you object to;

1. Glicko uses Bayesian methodology.
2. Bayesian methodology assumes that all combinations are equally likely - that is that any player is just as likely to play any other.
3. Bayesian methodology alone is inadequate for accurately predicting ratings in a semi-closed system, ie when a player is more likely to play certain players repeatedly.
So what.
That does not prove that there is a problem or even likely to be a problem.

The Elo system with varying K values would be very similar to the Glicko system. Can you show me a reference where the ELO system uses Bayesian methodology.


Bottom line is, this was discussed to death on the old BB.
You wasted Kevin, Barry and my time then.
You are wasting it now.

Bill Gletsos
19-04-2004, 12:12 AM
Garvin, I've made many suggestions, Bill's full cogniscent of many options he has available, but it's become a matter of honour to him to solve these issues within the Glicko system.
You do sprout a load of rubbish.
Certainly you havent proposed any useful options, thats for sure.

Oh thats tight, the "maturation factor".
Hell kids get older, they must be improving, just give them rating points.
Dont let their results get in the way.

What a load of cods wallop that was.

Cat
19-04-2004, 12:20 AM
How do you draw this conclusion from my statements above.


So what.
That does not prove that there is a problem or even likely to be a problem.

The Elo system with varying K values would be very similar to the Glicko system. Can you show me a reference where the ELO system uses Bayesian methodology.


Bottom line is, this was discussed to death on the old BB.
You wasted Kevin, Barry and my time then.
You are wasting it now.

Answer the question Bill. What's ELO got to do with the question?

It's impossible to rationally accept that a semi-closed system exists in Australia and at the same time argue that Bayesian methodology is sufficient to describe dynamic movement adequately. True or false.

Bill Gletsos
19-04-2004, 12:28 AM
Answer the question Bill. What's ELO got to do with the question?
Its simple.
You are arguing that the make up of the opponents is important as to whether they are adult or junior.
This has no bearing on the ELO system.
The Elo system does not use a bayesian model.
Yet the Elo system is a special case of the Glicko system.
Why should it therefore have an effect on the Glicko.



It's impossible to rationally accept that a semi-closed system exists in Australia and at the same time argue that Bayesian methodology is sufficient to describe dynamic movement adequately. True or false.
Well you believe its true and I guess thats all that really matters , now isnt it.

The USCF has a far larger pool of players and juniors and junior only events than Australia. There has been no argument put forward by the USCF Ratings Commision that juniors should be treated differently to adults.

Guess they are unfortuante to have good ol Doc Richards pushing his theory onto them.
Those damn yanks just dont know what they are missing. :whistle:

Cat
19-04-2004, 12:30 AM
You do sprout a load of rubbish.
Certainly you havent proposed any useful options, thats for sure.

Oh thats tight, the "maturation factor".
Hell kids get older, they must be improving, just give them rating points.
Dont let their results get in the way.

What a load of cods wallop that was.


No, that's simply a distortion of what I said. I've said many things;

1. A maturation factor related to performance.
2. A differential effective-k factor for adults and juniors.
3. The volatility should be directional.
4. That regional performance should be referenced to a standard, such as Sydney.
5. That the ACF studies rating changes in developing juniors to identify the most appropriate dynamic model.

All of these were suggestions, not complete models. But I guess we'll just continue with the 100 point corrections every couple of years.

Cat
19-04-2004, 12:36 AM
Its simple.
You are arguing that the make up of the opponents is important as to whether they are adult or junior.
This has no bearing on the ELO system.
The Elo system does not use a bayesian model.
Yet the Elo system is a special case of the Glicko system.
Why should it therefore have an effect on the Glicko.


Bill, your nonsense just gets worse. So you still can't answer the question, all it takes is a T or F.

Bill Gletsos
19-04-2004, 12:43 AM
No, that's simply a distortion of what I said. I've said many things;

1. A maturation factor related to performance.
No basis in theory.


2. A differential effective-k factor for adults and juniors.
Why.
Oh thats right, the Doc thinks its a good idea.
Unfortunately no basis in theory. The K should not depend on chess and their results not the age of the player.


3. The volatility should be directional.
Why.
Sane reason as 2. I suspect.


4. That regional performance should be referenced to a standard, such as Sydney.
Possibly, but there are many many variables. Simply being in a particular region isnt suffiecent reason for any sort of tweak.


5. That the ACF studies rating changes in developing juniors to identify the most appropriate dynamic model.
Pay me, and I might consider it. ;)
After all my time is valuable. :hand:


All of these were suggestions, not complete models. But I guess we'll just continue with the 100 point corrections every couple of years.
I wouldn't bet on it if I were you.
Oh thats right, you are not me. :hand:
Praise the lord. :owned:

Cat
19-04-2004, 12:44 AM
The USCF has a far larger pool of players and juniors and junior only events than Australia. There has been no argument put forward by the USCF Ratings Commision that juniors should be treated differently to adults.

Yes, they have
1. A much larger pool of players.
2. Less regionalisation.
3. A long chess playing tradition with a large bedrock of adult players and with comparatively less juniors as a proportion of participants, certainly when compared to the relatively isolated regions of the Gold Coast and ACT, etc..

But I do believe they are having some junior rating problems.

Bill Gletsos
19-04-2004, 12:45 AM
Bill, your nonsense just gets worse.
The only one speaking nonsense is you.

In fact all you are doing is raking over old coals from last year and in the process destroying what was a reasonable thread.

Bill Gletsos
19-04-2004, 12:47 AM
Yes, they have
2. Less regionalisation.
What proof do you have of this statement?
Cite an actual reference to back this up.


3. A long chess playing tradition with a large bedrock of adult players and with comparatively less juniors as a proportion of participants, certainly when compared to the relatively isolated regions of the Gold Coast and ACT, etc..
Again what proof do you have of this statement.
Cite an actual reference to back this up.

Alan Shore
19-04-2004, 12:48 AM
Yes, they have
1. A much larger pool of players.
2. Less regionalisation.
3. A long chess playing tradition with a large bedrock of adult players and with comparatively less juniors as a proportion of participants, certainly when compared to the relatively isolated regions of the Gold Coast and ACT, etc..

But I do believe they are having some junior rating problems.

Where are you getting all your US stats from David?

Bill Gletsos
19-04-2004, 12:49 AM
Where are you getting all your US stats from David?
I suspect out of some orifice. ;)

Cat
19-04-2004, 12:55 AM
No basis in theory.

Whose theory Bill, your theory eh? There is plenty of theory available on describing dynamic changes in human populations that is available for examination. There are theories on child development, theories about mathematical representations of changing human demographics. To believe that chess perfomance is somehow unique and different to every other human pursuit is simple ignorance. This is how is always been done, so why should we change it? Because the demographics in Austrlia are unique and we need specific solutions for our specific needs.

Garvinator
19-04-2004, 12:56 AM
i have a question that could be for both bill and david at opposite ends of the spectrum. I would be willing to bet money on the fact that their are rapidly improving adults too. How come the debate is about junior players all the time? Why dont we debate rating points and speciation etc etc for the improving adults too? Why is this 'problem' only been debated in a junior context?

Bill Gletsos
19-04-2004, 12:57 AM
Whose theory Bill, your theory eh? There is plenty of theory available on describing dynamic changes in human populations that is available for examination. There are theories on child development, theories about mathematical representations of changing human demographics. To believe that chess perfomance is somehow unique and different to every other human pursuit is simple ignorance.
If I recall corrrectly, Kevin shot you down on this one on the basis that there was nothing to show that chess developemnt followed the same developmental cirves.

I'm sure Kevin will correct me or clarify if I'm misrepresenting him..

Cat
19-04-2004, 01:05 AM
Where are you getting all your US stats from David?

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 - that the large adult population would act as a stabilising effect in the system and that developing juniors would move more rapidly to this much larger population mass.

You know as well as I do the US has a population of around 250 million with urban conurbations on the East and West Coast in excess of the Australian population. Public transport in the US is more easily accessible and the level of funding for the USCF is much greater relatively than the ACF- hence its able to hold more prestigious events and encourage greater participation.

I'm sure there would be areas of the US that could be problematic, of course.

Alan Shore
19-04-2004, 01:06 AM
i have a question that could be for both bill and david at opposite ends of the spectrum. I would be willing to bet money on the fact that their are rapidly improving adults too. How come the debate is about junior players all the time? Why dont we debate rating points and speciation etc etc for the improving adults too? Why is this 'problem' only been debated in a junior context?

Volume - 90% of those players with very large increases are juniors. It does happen at adult level, just a lot more infrequently.

David's arguments I find a little too idealistic and complex - this is a standardised rating system, you cannot account for all these individual differences. The primary counter to individual differences in psychological testing is simply having a sufficiently large sample size. The number of players on the rating list is certainly a sufficient sample size - the calculations you would want to implement would take a lot of time, yielding what I believe would be little reward.

Bill Gletsos
19-04-2004, 01:06 AM
i have a question that could be for both bill and david at opposite ends of the spectrum. I would be willing to bet money on the fact that their are rapidly improving adults too. How come the debate is about junior players all the time? Why dont we debate rating points and speciation etc etc for the improving adults too? Why is this 'problem' only been debated in a junior context?
You wont see me just restricting it to juniors.
In fact its the use of an RD and the volatility factor in Glicko2 that will handle rapidly improving players.
However one has to take into consideration that you have to be sure the player is improving and not just performing at the extreme of his true strength. After all the 1500 who performs like a 1700 in one rating period over 30 games is much more likely to have actually improved than the 1500 who performs like a 1700 over 7 games.

Cat
19-04-2004, 01:16 AM
If I recall corrrectly, Kevin shot you down on this one on the basis that there was nothing to show that chess developemnt followed the same developmental cirves.

I'm sure Kevin will correct me or clarify if I'm misrepresenting him..

Yes, well Dr Kev maybe pretty smart about snails, but I think he may be a little less qualified to speak about human child development. There are a number of seminal events in the intellectual development of the child, in particular related to the onset of puberty. Under the influence of the sex hormones, the childs ability to think mathematically, abstractly and their language abilities suddenly explode. These intellectual skills profoundly effect the child's reasoning abilities and processing skills. In fact, all the skills prerequisite to chess performance. Amazing isn't it? Look, at the risk of boring the reader, I can post references to studies not only on child development, but also to some papers on thinking skills and chess. I'll have to dig them out though, so it won't be tonight.

Cat
19-04-2004, 01:24 AM
Volume - 90% of those players with very large increases are juniors. It does happen at adult level, just a lot more infrequently.

David's arguments I find a little too idealistic and complex - this is a standardised rating system, you cannot account for all these individual differences. The primary counter to individual differences in psychological testing is simply having a sufficiently large sample size. The number of players on the rating list is certainly a sufficient sample size - the calculations you would want to implement would take a lot of time, yielding what I believe would be little reward.

We have around 120 ACF rated players on the Gold Coast, and as I say last time I checked 79% were juniors. Most of these players play at home. Its virtually impossible for them to progress in the rating system unless their outstanding, are prepared to travel, or recieve the biannual rating boost. Consequently, many juniors turn their attention to other concerns where the rewards are better. Given places like the Gold Coast are so important for the future of chess in Australia its certainly a handicap.

Garvinator
19-04-2004, 01:28 AM
We have around 120 ACF rated players on the Gold Coast, and as I say last time I checked 79% were juniors. Most of these players play at home. Its virtually impossible for them to progress in the rating system unless their outstanding, are prepared to travel, or recieve the biannual rating boost. Consequently, many juniors turn their attention to other concerns where the rewards are better. Given places like the Gold Coast are so important for the future of chess in Australia its certainly a handicap.

unless im mistaken, the gardiner chess centre ran the christmas swiss last year and also the reserve section for the fide tournament. Both had, in my opinion, very poor turnouts, especially from the juniors. How can juniors progress in the rating system when they arent turning up for adult tournaments?

Cat
19-04-2004, 01:30 AM
However one has to take into consideration that you have to be sure the player is improving and not just performing at the extreme of his true strength. After all the 1500 who performs like a 1700 in one rating period over 30 games is much more likely to have actually improved than the 1500 who performs like a 1700 over 7 games.

Yes and with the same reasoning, an individual improves and learns a task fairly quickly, but looses that skill very slowly. The assumption that RD is normally distributed is just that, an assumption. It seems much more likely that RD is skewed, i.e. that a player may improve quickly but decline more slowly - particularly a child. Of course that was another suggestion I gave Bill, that RD should be skewed to more accurately reflect the liklihood of change in performance.

Cat
19-04-2004, 01:38 AM
unless im mistaken, the gardiner chess centre ran the christmas swiss last year and also the reserve section for the fide tournament. Both had, in my opinion, very poor turnouts, especially from the juniors. How can juniors progress in the rating system when they arent turning up for adult tournaments?

The FIDE tournament was restricted to 10 participants, and only individuals with realisitic expectations were invited to play. The reserve competition, which I'm pleased to say I won against stiff competition, also had 10 participants, but it was an adult event.

The Gardiner Chess centre has made monday and wednesday adult competition night and the remaining nights are junior. Partly its because the adults demand certain social requirements, but juniors rated over 1200 are invited to play. Kelvin Finke and Alex Jule have both been regular attenders and I must say have performed very well.

I think the Christmas Swiss may have been affected by other engagements, I was in Europe for example

Bill Gletsos
19-04-2004, 01:39 AM
Yes and with the same reasoning, an individual improves and learns a task fairly quickly, but looses that skill very slowly. The assumption that RD is normally distributed is just that, an assumption. It seems much more likely that RD is skewed, i.e. that a player may improve quickly but decline more slowly - particularly a child.
You seemed to have this lack of comprehension regarding the RD and the volatility before.

The RD is just the standard deviation of the players rating.
Why should it be skewed.

If a player is improving rapidly this will be handled by the volatility factor of the glicko2 system.



Of course that was another suggestion I gave Bill, that RD should be skewed to more accurately reflect the liklihood of change in performance.
Yes with no basis in fact. Just some hackneyed theory.

Bill Gletsos
19-04-2004, 01:46 AM
We have around 120 ACF rated players on the Gold Coast, and as I say last time I checked 79% were juniors. Most of these players play at home. Its virtually impossible for them to progress in the rating system unless their outstanding, are prepared to travel, or recieve the biannual rating boost.
This seems contradictory.
So most stay home. Therefore most will have accuarte ratings relative to other pool members.
Those that play away (ourside the pool) will gain points from those away games if they are truly underrated.

As highlighted previously last year, the Gold Coast seems to have many junior only rapids but very few junior only tournaments at normal time controls that can be rated in the Normal rating system. Hence the juniors only progress slowly on the normal list because they have a lack of results in the normal system.

This is not a rating system problem but one due to tournament logistics.

They should play more normal rated games.

Rincewind
19-04-2004, 01:48 AM
A few observations...

No rating pool approximates the Richards' ideal of all players having equal likelyhood of playing all other in the rating pool. All rating pools have some form of biasing in the players they are paired against. Most players are more likely to play against players of similar rating strength due to finishing in similar streams in a swiss tournaments, playing in restricted tournaments (like the doeberl minor, major, etc).

In fact, it's not clear to me that Bayesian analysis requires or makes the assumption that there be equal probability of pariings throughout the entire rating pool. Bayesian method is to begin with a prior probability (which is the player's rating at the start of a rating period) and adjust it based on the information gathered in the present period to generate a posterior probablity (which is the new rating). Sufficient pairings between players in the rating pool is necessary, but much less than uniform probability would still provide sufficent interaction for a reliable posterior estimate.

Regarding regionalisation...

I imagine that the US has similar issues as are present in Australia. While they have generally better public transport infrastructure in large urban areas, the interurban system seems to be comparable with Australa's. Moderate proportions of both populations travelling large distances (by European standards) in personal motor vehicles. Yes the do have larger natoinal competitions but this is mainly due to their larger population and not testimony to their superior public transport systems.

In fact, Australia has the advantage of a greater population concentration on the SE seaboard which means regionalisation is probably less of an issue in Asutralia anywhere between Brisbane and Melbourne (including Canberra) than in most of the US.

Regions well outside of that area may have an issue of regionalisation comparable to the US (EG Perth, Darwin and to a lesser extent Adelaide and Hobart).

Kevin Bonham
19-04-2004, 01:52 AM
KB, there are just so many of these anomalies aren't there?

No. :hand:

In those cases where a junior player "jumps" there is often no real warning. Many of them muddle along with not much sign of improvement for years and then suddenly put in a performance that spectacularly exceeds past results with no real warning. This may be a case in point. How is a rating system supposed to predict that and how is it supposed to distinguish between that and simply a purple patch that doesn't last?

I suppose you're going to tell me it's more likely to happen at certain ages. Maybe so but I've seen it happen to eight year olds or eighteen year olds, I doubt it's that predictable.

Kevin Bonham
19-04-2004, 02:02 AM
It's impossible to rationally accept that a semi-closed system exists in Australia and at the same time argue that Bayesian methodology is sufficient to describe dynamic movement adequately. True or false.

False.

You are attempting to argue that juniors playing other juniors disproportionately will produce underrated juniors out of thin air.

There is actually no reason at all why this should be so. It is where the other juniors were seriously underrated to begin with that problems may kick in.

Kevin Bonham
19-04-2004, 02:22 AM
Yes, well Dr Kev maybe pretty smart about snails, but I think he may be a little less qualified to speak about human child development.

That is true but I have done enough science and other studies to know a clearly invalid or incomplete argument when I see one no matter what the field. The ability to spot BS irrespective of field is supposed to be one of the benefits of a sufficiently long incarceration in a uni environment. I am sure if I made claims about snail science that were as flawed in their reasoning as some of your claims, that any other good scientist from any field would shoot me down too.


There are a number of seminal events in the intellectual development of the child, in particular related to the onset of puberty. Under the influence of the sex hormones, the childs ability to think mathematically, abstractly and their language abilities suddenly explode. These intellectual skills profoundly effect the child's reasoning abilities and processing skills. In fact, all the skills prerequisite to chess performance.

If it was as simple as that, any child developing exceptional reasoning and processing skills at that age would be an outstanding chess player.

Yet many children with such exceptional skills do not even choose to play chess at all, or do not become interested enough in the game to take it seriously. Others who are more interested but have less basic mental talent outperform them.

A chessplayer's performance and improvement is about far more than their mental potential to be a good player. It can also be about things like their interest level in the game, their access to competition and coaching that can help them improve, whether they are supported in their interest and so on.

Therefore you simply cannot assume that a medical argument about potential, made in complete neglect of social factors that might influence performance, will tell you enough about chess performance to determine a correction for a ratings system.

You might, however, use such a medical argument to make predictions. But then you need to use actual performance data to show that those predictions hold up, and if you want a change to the ratings system you should have a specific proposed change that can be tested.

Cat
19-04-2004, 07:53 AM
They should play more normal rated games

Agreed


In fact, it's not clear to me that Bayesian analysis requires or makes the assumption that there be equal probability of pariings throughout the entire rating pool. Bayesian method is to begin with a prior probability (which is the player's rating at the start of a rating period) and adjust it based on the information gathered in the present period to generate a posterior probablity (which is the new rating). Sufficient pairings between players in the rating pool is necessary, but much less than uniform probability would still provide sufficent interaction for a reliable posterior estimate.


Yes, much less than uniform probablity, I agree. But much less is what exists in areas of Oz.


That is true but I have done enough science and other studies to know a clearly invalid or incomplete argument when I see one no matter what the field. The ability to spot BS irrespective of field is supposed to be one of the benefits of a sufficiently long incarceration in a uni environment. I am sure if I made claims about snail science that were as flawed in their reasoning as some of your claims, that any other good scientist from any field would shoot me down too.

Have you ever seen a child KB. Do they have them at your University?


You might, however, use such a medical argument to make predictions. But then you need to use actual performance data to show that those predictions hold up, and if you want a change to the ratings system you should have a specific proposed change that can be tested.

Yes, thats right. And a wise man might have regard for information available from a variety of disciplines to do just that.

Commentator
19-04-2004, 08:30 AM
I have put up the game I played with Eugene (see my original post in this thread). Again I ask BB readers, is this the game one would expect a (supposedly) 853 rated player to play?

Mr Sike,
I see that KB has commented on the relative strengths of the two players exhibited in this game.
Further, Ian Rout has suggested that junior players can play higher than their rating because of the ‘importance factor’.
And your tourist descriptions of where and what you ate in Canberra indicate your attention was elsewhere.

Combine these three effects and the game score was about correct.

C

ursogr8
19-04-2004, 08:41 AM
That surprises me, I thought the junior chess scene was quite big in Victoria what with Cordover, Jammo, Daz etc. operations happening.. how come there are so few junior tournaments? I think we're spoiled here in QLD, there seems to be on average one every week or two!

BD
I did not comment on the number of junior events; and I certainly did not say there are few junior events.
What I did say is that there are very few rated junior events. I can only think of one > the VIC JUNIOR OPEN…a week-ender.


I would imagine that ChessGURU runs many junior events but according to the ACF ‘Tournaments rated’ paragraph none are rated by my observation.
The Interschools is not rated as far as I know.
At Box Hill, the juniors play in adult events after they have had about 6 month of coaching and trial games.
At Drouin, their big event was all-in.

I have not seen MCC nor Dandenong, nor Frankston, nor Mentone, nor Ballarat, nor Geelong, nor Chess Ideas submit junior-only games for rating.

starter

Rincewind
19-04-2004, 10:22 AM
Yes, much less than uniform probablity, I agree. But much less is what exists in areas of Oz.

Was there supposed to be a question mark at the end of that sentence?

As I said in my first message, all rating pools contain sub-pools with more limited interaction between the sub-pools. Is there any eviidence that the problem is particularly worse in Australia than anywhere else? Furthermore, do you have a method which will produce a posterior probability with greater predictive accuracy?

The Elo system is certainly not immune from the problems caused by closed sub-pools. There was the well documented cases of several vastly overrated Burmese players on the official FIDE list only a couple of years ago.

PHAT
19-04-2004, 11:32 AM
I dont recall anyone agreeing with you [David Richards] then.

I did, a lot. And I am about to start a new campaigne.

Bill Gletsos
19-04-2004, 11:43 AM
I did, a lot.
Yes, well unfortunately you were no more convincing that David. ;)

And I am about to start a new campaigne.
I can hardly wait. :rolleyes:
Let us hope it is more successful than the idea of predictiveness based on past trends. :whistle:

Lucena
19-04-2004, 11:43 AM
I hate to say it Matthew but I think underrated juniors are a fact of life and there's not an awful lot you can do about it :doh:

Bill Gletsos
19-04-2004, 11:45 AM
I hate to say it Matthew but I think underrated juniors are a fact of life and there's not an awful lot you can do about it :doh:
Yes, we also have to be careful not to over generalise.
Not all juniors are underrated.
Even some of the higher rateed juniors are still erratic, especially the younger ones.

Garvinator
19-04-2004, 11:46 AM
I hate to say it Matthew but I think underrated juniors are a fact of life and there's not an awful lot you can do about it :doh:

part of reducing the effect of underrated juniors is to have the junior only tournaments played under the normal time control and also to have juniors playing more in 'open' events against adults. Some may complain that would increase the amount of adults having points 'ripped' off them, but in return the junior would increase their reliability of their rating faster and so any genuine improvement or under rating would be easier to detect.

PHAT
19-04-2004, 12:05 PM
This is not a rating system problem but one due to tournament logistics.

They should play more normal rated games.

:lol: Onya. When your Glicko fails to track players' ability, lay the blaim on them for not playing in the events that you think they need to.

Bill, DR is belting you all over the ground. Take a deep breath and repete after me, "I will admit my mistakes."

Garvinator
19-04-2004, 12:07 PM
:lol: Onya. When your Glicko fails to track players' ability, lay the blaim on them for not playing in the events that you think they need to.

Bill, DR is belting you all over the ground. Take a deep breath and repete after me, "I will admit my mistakes."

how can any system 'keep up' when provided with insufficent and incomplete data?

Bill Gletsos
19-04-2004, 12:14 PM
:lol: Onya. When your Glicko fails to track players' ability, lay the blaim on them for not playing in the events that you think they need to.
Matt, mate dont be silly.
If QLD juniors are playing mostly rapid games(which they are) and not normal games, then it is hardly the fault of the rating system if there is insufficient data.


Bill, DR is belting you all over the ground.
I disagree.
As it appears do Kevin and Barry.


Take a deep breath and repete after me, "I will admit my mistakes."
After careful thought I considered I may have made a mistake, but upon deeper reflection I realised that my mistake was to think I was mistaken. ;)

PHAT
19-04-2004, 12:43 PM
Many of them muddle along with not much sign of improvement for years and then suddenly put in a performance that spectacularly exceeds past results with no real warning ... How is a rating system supposed to predict that and how is it supposed to distinguish between that and simply a purple patch that doesn't last?


Your retoric is correct - we cannot accuritely predict leaps, and nor can we immeadiately identify a good performance as a leap or a fluke. However, some games results are more relavent than others for estimating a player's ability. For example, a 60m+10s club knockout games is taken less seriously by many players, than a 90m+30s at a far away venue. The Glicko treats both games as being equaly valuable data points.

A simple way to account for these types of phenominae is to have games weighted according to some algorithm that factors: home/away, length, championship, mean rating of the two players, et cetera. At the moment, the Glicko factors only: past performance, activity, and difference in rating.

As David Richards says, there is so much information that can be used in ratings calculations, but it is not used because such theorectical factors based on meta analysis - factors that Bill denies exist.

Kerry Stead
19-04-2004, 12:55 PM
I think the 'problem' of underrated juniors is simply one of when players are most likely to improve - when they have time to study the game and have minimal distractions while playing their games ... children fit very neatly into this category.
As Bill says, just because someone is a junior does not mean that they will automatically improve, but the opposite also applies. Typically improvement happens when a player gets some coaching, plays more events (simply exposure to the game and learning from mistakes can be very beneficial) and is generally motivated to do well.
I think some of these factors may also apply to university students, as they too tend to have some time on their hands to study chess, and typically do not have quite the level of distarctions that others.
Once you get past university level, other factors tend to play a more important role in life, which takes away from any potential chess improvement. Things like work, family/relationships, financial needs, etc become prioritised, so that when one gets to the chess board, these things remain in the back of the mind, so that the focus can not be 100% chess.
One also has to keep in mind the 'dropout factor' of older kids - yes they might improve significantly, but if they stop playing chess when they hit 16, what good does it do to the rating pool as a whole if they are given extra points for being a junior if there is no chance of these points being redistributed amongst the rating pool?
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.
As for the kids mentioned, I think Jessica Kinder's game has improved enormously since the Australian Clubs Teams Championship! :D ;)

ursogr8
19-04-2004, 01:07 PM
Bill, DR is belting you all over the ground. Take a deep breath and repete after me, "I will admit my mistakes."





After careful thought I considered I may have made a mistake, but upon deeper reflection I realised that my mistake was to think I was mistaken. ;)


As eagerly awaited as the first sightings of swallows in Spring, are the first few balls of a Test Match, or the first attempt by Matt to arc-up Bill in 2004.
The first ball was indeed head-high, but sadly it is down leg-side and Bill watches it whistle by, and takes the opportunity to sledge the wicket-keeper as well.

Alan Shore
19-04-2004, 01:15 PM
This is not a rating system problem but one due to tournament logistics.

They should play more normal rated games.

Well this is one comment I found puzzling to put as part of your argument.. sure they *should* but the fact is they won't - juniors prefer rapid games to longer time controls. This is why the adult long time controls turnouts are significantly lower than rapid tourns and to cater for the popularity more rapid tourns are held. This is why so many juniors play blitz and lightning games on the net rather than standard.

arosar
19-04-2004, 01:17 PM
Yo starter mate....I was flicking through some of me own old issues of CIA and on one cover, 1977 I think, there's a certain Ron Barassi opening a game on behalf of a certain jammo. That must surely warm the hearts of the likes of youse AFL lovin' Mexicans.

AR

PS: Any1 of youse know what happened to the Alex Szirt Cup?

Bill Gletsos
19-04-2004, 01:19 PM
Your retoric is correct - we cannot accuritely predict leaps, and nor can we immeadiately identify a good performance as a leap or a fluke. However, some games results are more relavent than others for estimating a player's ability. For example, a 60m+10s club knockout games is taken less seriously by many players, than a 90m+30s at a far away venue. The Glicko treats both games as being equaly valuable data points.
I disagree. I think its an exaggeration to suggest the 60m+10s games are taken less seriously.
Most players take very seriously any game that effects their normal ratings.
Why should those players who take all their games seriously be penalised because of the attitudes of a few.


A simple way to account for these types of phenominae is to have games weighted according to some algorithm that factors: home/away, length, championship, mean rating of the two players, et cetera. At the moment, the Glicko factors only: past performance, activity, and difference in rating.
You wouuld have to prove conclusively that the home/away , length, type of tournament has any actual effect on how the majority of players play.

After all if players dont like the way 60m + 10s might effect their rating, the answer is simple. Dont play in such tournaments. Dont penalise the majority to satisfy the minority.


As David Richards says, there is so much information that can be used in ratings calculations, but it is not used because such theorectical factors based on meta analysis - factors that Bill denies exist.
Factors that David cannot adequately quantify.

Bill Gletsos
19-04-2004, 01:23 PM
Well this is one comment I found puzzling to put as part of your argument.. sure they *should* but the fact is they won't - juniors prefer rapid games to longer time controls. This is why the adult long time controls turnouts are significantly lower than rapid tourns and to cater for the popularity more rapid tourns are held. This is why so many juniors play blitz and lightning games on the net rather than standard.
Well thats fine, but then dont bitch about a juniors normal rating if they are not playing many normal games.

If you want a junior to have a reliable normal rating they need to play a reasonable number of normal games.

Bill Gletsos
19-04-2004, 01:25 PM
Yo starter mate....I was flicking through some of me own old issues of CIA and on one cover, 1977 I think, there's a certain Ron Barassi opening a game on behalf of a certain jammo. That must surely warm the hearts of the likes of youse AFL lovin' Mexicans.

AR

PS: Any1 of youse know what happened to the Alex Szirt Cup?
It used to be a knockout tournament held yearly at St. George Leagues club.
My understanding is the sponsorship finally died and the tournament folded.
Someone like Parr would know the details better. Ask him next time you see him and report back.

Alan Shore
19-04-2004, 01:28 PM
Well thats fine, but then dont bitch about a juniors normal rating if they are not playing many normal games.

If you want a junior to have a reliable normal rating they need to play a reasonable number of normal games.

Settle down Bill, I agree with you!

PHAT
19-04-2004, 01:29 PM
Why should those players who take all their games seriously be penalised because of the attitudes of a few.


"penalised". How so?

I am only suggesting that particular games might be weighted for rating purposes. (BTW this is not the idea I will be pushing - even though it has merrit.)



You wouuld have to prove conclusively that the home/away , length, type of tournament has any actual effect on how the majority of players play.


"conclusively". What standard of evidence, exactly, would you concider to make such a claim "conclusive". For example, would you ask me to prove a claim that AFL players play harder in an MCG grandfinal than in the 3rd round of the season?

PHAT
19-04-2004, 01:39 PM
I hate to say it Matthew but I think underrated juniors are a fact of life and there's not an awful lot you can do about it :doh:

In deed, there are under-rated juniors, 'tis a fact of life. However there is a lot that can be done about it. Bill is too much of a scaredy-cat to do any of them.

Bill Gletsos
19-04-2004, 01:48 PM
Settle down Bill, I agree with you!
I should have been clearer.
My dont bitch comment wasnt in regards to you.
It was a comment aimed at others.

Bill Gletsos
19-04-2004, 01:55 PM
"penalised". How so?

I am only suggesting that particular games might be weighted for rating purposes. (BTW this is not the idea I will be pushing - even though it has merrit.)
If player A treats all his games seriously irrespective of the time limit and player B only treats 90m + 30s seriously then if 60m+10s games are weighted less this is probably ok from player B's viewpoint but not from palyer A's.

This is sort of like the crap argument that you should get x points if you beat a player rated 200 above you but lose less than x points if you lose to a player rated 200 below you.

There is no justification for it.



"conclusively". What standard of evidence, exactly, would you concider to make such a claim "conclusive". For example, would you ask me to prove a claim that AFL players play harder in an MCG grandfinal than in the 3rd round of the season?
Its very hard to draw an analogy. Chess is an individual sport.
After all does a player at Wimbeldon play harder in the quarter, semi or final than in the earlier rounds. Logically one would say no, otherwise said player is likely not to even make the quaters etc.

ursogr8
19-04-2004, 02:58 PM
Yo starter mate....I was flicking through some of me own old issues of CIA and on one cover, 1977 I think, there's a certain Ron Barassi opening a game on behalf of a certain jammo. That must surely warm the hearts of the likes of youse AFL lovin' Mexicans.

AR


And I think someone said to me that Barassi also played Karpov. But I can't be sure.

PHAT
19-04-2004, 04:04 PM
If player A treats all his games seriously irrespective of the time limit and player B only treats 90m + 30s seriously then if 60m+10s games are weighted less this is probably ok from player B's viewpoint but not from palyer A's.

This is sort of like the crap argument that you should get x points if you beat a player rated 200 above you but lose less than x points if you lose to a player rated 200 below you.

There is no justification for it.


Don't compare my idea with that inflationary "crap". Not only is what I am saying reasonable, but it has no analogy with that other "crap". So, lumping in my idea with that "crap" is gratuitous mud sport.




Its very hard to draw an analogy. Chess is an individual sport.
After all does a player at Wimbeldon play harder in the quarter, semi or final than in the earlier rounds. Logically one would say no, otherwise said player is likely not to even make the quaters etc.

No, not logically at all.
1. The finals player is likely to have had played only players seeded below him.
2. Since on average, the skill difference is greater in the early rounds, the games are likely to be shorter.
3. His opponents are less skilled and therefore the ball comes slower and bounces less close to the line - thus physical exersion is less.

I suggest you put a ban on yourself from using the words, "logical" and "clearly", because you use them as weapons. Unfortunately, what you are defending is the status quo. And that ain't a good policy in a rapidly changing world.

arosar
19-04-2004, 04:12 PM
Reading that post from Matt reminded me to say that perhaps I was a little unfair in characterising me mate Bill as "obstructionist". Recent events, not presently in the public light, clearly prove me wrong. I hereby withdraw that remark.

AR

Bill Gletsos
19-04-2004, 04:13 PM
Don't compare my idea with that inflationary "crap". Not only is what I am saying reasonable, but it has no analogy with that other "crap". So, lumping in my idea with that "crap" is gratuitous mud sport.
You can say what you like, there is no justification for penalising players who take all games seriously just because a few dont.



No, not logically at all.
1. The finals player is likely to have had played only players seeded below him.
2. Since on average, the skill difference is greater in the early rounds, the games are likely to be shorter.
3. His opponents are less skilled and therefore the ball comes slower and bounces less close to the line - thus physical exersion is less.
Not necessarily true at all.
Just look at the number of seeded players who lose prior to the quaterfinals.


I suggest you put a ban on yourself from using the words, "logical" and "clearly", because you use them as weapons. Unfortunately, what you are defending is the status quo. And that ain't a good policy in a rapidly changing world.
Sounds like the person complaining because their arguments are neither clear, logical or justifiable.

PHAT
19-04-2004, 04:14 PM
If player A treats all his games seriously irrespective of the time limit and player B only treats 90m + 30s seriously then if 60m+10s games are weighted less this is probably ok from player B's viewpoint but not from palyer A's.


A game that lasts 1 hour has half the predictive capacity of one that lasts 2 hours. For example, which data set would give you a the most accurite measure of a player's strength: 10 blitz games, or 10 x 1 hour games or 10 x 4 hour games.

Therefore, games should be weighted according to their length.

Bill Gletsos
19-04-2004, 04:18 PM
A game that lasts 1 hour has half the predictive capacity of one that lasts 2 hours.
Where is your proof of this statement.



For example, which data set would give you a the most accurite measure of a player's strength: 10 blitz games, or 10 x 1 hour games or 10 x 4 hour games.
There is a big difference between blitz games and games of an hour of greater. However there is no evidence that games of 1hr or games of 90mins + 30 to finish (sudden death) or 30secs per move are significantly different with regards determining playing strength.


Therefore, games should be weighted according to their length.
Only in your opinion.

PHAT
19-04-2004, 04:22 PM
Sounds like the person complaining because their arguments are neither clear, logical or justifiable.

Maybe so, but I am being clear, logical and justifying my statements.

Now tell me how making, say, 4 hr games "worth" twice the rating change of a 2 hr game, would penalise any player. Come on, be logical, be clear.

Bill Gletsos
19-04-2004, 04:26 PM
Maybe so, but I am being clear, logical and justifying my statements.

Now tell me how making, say, 4 hr games "worth" twice the rating change of a 2 hr game, would penalise any player. Come on, be logical, be clear.
Because it does not provide twice as much information as to the players strength. In fact there is probably little difference between the strength of a player for a G/60 game and a G/120.

PHAT
19-04-2004, 04:39 PM
However there is no evidence that games of 1hr or games of 90mins + 30 to finish (sudden death) or 30secs per move are significantly different with regards determining playing strength.


There being "no evidence" does not disprove my assertion. On the other hand, I can make a case as follows.

If we were to judge a players strength by looking at their n th move in round 4 of the XYZ Cup, we could rightly be said to be wankers - how can we judge on the basis of a move that took 1 or 2 minutes of thinking time and where there are probably more than one good move that could be made. Well, what about looking at 10 randomly picked moves from the game - that would be better. Now lets say we will look at all the moves - better still. The longer the game, and/or the more time spent on each move, the less the random error. in the final result. Less random error equates to a more reliable the result. Thus, longer games are more reliable than shorter games. Hence, weighting according to length is logical.

PHAT
19-04-2004, 04:40 PM
Because it does not provide twice as much information as to the players strength. In fact there is probably little difference between the strength of a player for a G/60 game and a G/120.

Where is your proof? Or is this just your un varified assumption.

frogmogdog
19-04-2004, 04:48 PM
fab, another incessant ratings discussion.

sorry haven't read most of the preceding but a few things come to mind.

1. i'm posting because unfortunately ratings matter. eg in xmas 2002, at the qld junior club champs, moulthun ly was supposed to play board 5 for our junior club on account of his 700 rapid rating (whereas he was already easily the strongest player, beat a 2300 not long after)

2. one kid had a single digit rating. luckily he thought it was hilarious to be rated lower than stanley the chimp, a chessmaster 9000 random move generator, but i think it's both a potential insult to give players ratings with fewer than 3 digits and also somewhat meaningless because it's easy to improve beyond that level.

3. i think an RD floor for ratings under a few hundred (to keep them volatile) and a rating floor of somewhere between 100 and 300 should be considered. this reduces the problem of kids with ratings of 50 beating 800's, who then beat 1200 adults and so on up.

them's me thorts.

ps bill, i do agree glicko2 > glicko > elo

Kerry Stead
19-04-2004, 05:46 PM
Correct me if I'm wrong, but didn't there used to be an additional weighting given to games played in the Australian Championship (and potentially events like state championships, etc)?
Personally I can see some merit in such a system (although looking at my performance in the most recent Australian Champs I might want to reconsider), as it rewards a good performance in a big event. Surely Bill you can see that a game in a category 1 GP tournament has a higher weighting (from a GP standpoint) than a catefory 3 tournament? Wouldn't something similar also hold true from a ratings point of view as well? (ie: a game at say Doeberl is more important than one at say the Drouin Open)

Rincewind
19-04-2004, 05:52 PM
Well this is one comment I found puzzling to put as part of your argument.. sure they *should* but the fact is they won't - juniors prefer rapid games to longer time controls. This is why the adult long time controls turnouts are significantly lower than rapid tourns and to cater for the popularity more rapid tourns are held.

I agree with all that.


This is why so many juniors play blitz and lightning games on the net rather than standard.

I disagree with that. The reason is the online environment and culture is simply more condusive to rapid and blitz play. I don't go out of my way to play rapid games and am even a member of the STC Bunch but play many more blitz games than standard games online.

Bill Gletsos
19-04-2004, 05:52 PM
Where is your proof? Or is this just your un varified assumption.
You are the one with the theory that a 4 hr game is twice as good as a 2hr game.
The onus is on you to prove it, not me to disprove it.

Bill Gletsos
19-04-2004, 05:54 PM
There being "no evidence" does not disprove my assertion. On the other hand, I can make a case as follows.

If we were to judge a players strength by looking at their n th move in round 4 of the XYZ Cup, we could rightly be said to be wankers - how can we judge on the basis of a move that took 1 or 2 minutes of thinking time and where there are probably more than one good move that could be made. Well, what about looking at 10 randomly picked moves from the game - that would be better. Now lets say we will look at all the moves - better still. The longer the game, and/or the more time spent on each move, the less the random error. in the final result. Less random error equates to a more reliable the result. Thus, longer games are more reliable than shorter games. Hence, weighting according to length is logical.
That does not make the case at all.
Its not evidence.
Just supposition.

Rincewind
19-04-2004, 05:55 PM
Correct me if I'm wrong, but didn't there used to be an additional weighting given to games played in the Australian Championship (and potentially events like state championships, etc)?
Personally I can see some merit in such a system (although looking at my performance in the most recent Australian Champs I might want to reconsider), as it rewards a good performance in a big event. Surely Bill you can see that a game in a category 1 GP tournament has a higher weighting (from a GP standpoint) than a catefory 3 tournament? Wouldn't something similar also hold true from a ratings point of view as well? (ie: a game at say Doeberl is more important than one at say the Drouin Open)

I can see where you're coming from if we want to predict performances at big events. But we want to predict performances at all events so why shouldn't all events receive an equal weighting.

Remember that a rating increase should not be viewed as a reward for playing well, it is merely a number used to predict the results of future pairings. You performance indicates a greater probability of doing well in future games, therefore the rating goes up.

shaun
19-04-2004, 06:23 PM
Correct me if I'm wrong, but didn't there used to be an additional weighting given to games played in the Australian Championship (and potentially events like state championships, etc)?
Personally I can see some merit in such a system (although looking at my performance in the most recent Australian Champs I might want to reconsider), as it rewards a good performance in a big event. Surely Bill you can see that a game in a category 1 GP tournament has a higher weighting (from a GP standpoint) than a catefory 3 tournament? Wouldn't something similar also hold true from a ratings point of view as well? (ie: a game at say Doeberl is more important than one at say the Drouin Open)

Yes. Under the ACF Ratings System that ran from 1981 the k factors were weighted as follows: Australian Championship/Open 30, State Championships 22.5, Normal adult events and Australian Junior 15, junior events 7.5.
The reasoning behind it wasn't so much the importance of events as to do with the reliability of the field. The idea was that "elite" events like National and State Championships attracted not only the best players, but players with reliable ratings, and that performances in these events accurately reflected a players "real" rating. Therefore the more rapid change to players ratings from these events. Conversely junior events consisted of "erratic" players whose performnaces couldn't be relied on, and therefore their ratings shouldn't change so much.
When I took over as ACF Ratings Officer I think the ACF was already phasing this part of the system out. Even if they weren't, one of my first actions as ratings officer was to eliminate it completely. I couldn't see (and still can't see) why the nature of the event allows a player to put in a lesser (or greater) effort into their play.

Bill Gletsos
19-04-2004, 06:31 PM
Yes. Under the ACF Ratings System that ran from 1981 the k factors were weighted as follows: Australian Championship/Open 30, State Championships 22.5, Normal adult events and Australian Junior 15, junior events 7.5.
The reasoning behind it wasn't so much the importance of events as to do with the reliability of the field. The idea was that "elite" events like National and State Championships attracted not only the best players, but players with reliable ratings, and that performances in these events accurately reflected a players "real" rating. Therefore the more rapid change to players ratings from these events. Conversely junior events consisted of "erratic" players whose performnaces couldn't be relied on, and therefore their ratings shouldn't change so much.
In actual fact for those who actually believe there should be any sort of weightings the converse really should actually apply.
After all since juniors being more "erratic" they should have a higher K factor than the default K-factor and the "elite" players being more stable should have a lower K-factor.



When I took over as ACF Ratings Officer I think the ACF was already phasing this part of the system out. Even if they weren't, one of my first actions as ratings officer was to eliminate it completely. I couldn't see (and still can't see) why the nature of the event allows a player to put in a lesser (or greater) effort into their play.
The April 1987 Rating booklet points out that weightings were as you described above.
The August 1987 Rating booklet notes that all events are weighted equally.

Kevin Bonham
19-04-2004, 06:32 PM
Have you ever seen a child KB. Do they have them at your University?

Do Resistance members qualify?

(I'm not there anymore anyway. I've more or less escaped, though they're about to make me a dishonorary research associate.)


Yes, thats right. And a wise man might have regard for information available from a variety of disciplines to do just that.

You're the one claiming to have the information. Produce a proposed improvement to the system that can be tested and if it's tested and works I'll listen.

Bill Gletsos
19-04-2004, 06:41 PM
Correct me if I'm wrong, but didn't there used to be an additional weighting given to games played in the Australian Championship (and potentially events like state championships, etc)?
Personally I can see some merit in such a system (although looking at my performance in the most recent Australian Champs I might want to reconsider), as it rewards a good performance in a big event.
Surely Bill you can see that a game in a category 1 GP tournament has a higher weighting (from a GP standpoint) than a catefory 3 tournament? Wouldn't something similar also hold true from a ratings point of view as well? (ie: a game at say Doeberl is more important than one at say the Drouin Open)
The rating system and the GP are two different things.
The rating system is trying to indicate a players strength relative to others.

The GP system is allocating points based on where a player finishes in an event. The reason a GP3 is worth more poinst than a GP1 is because it is expected that the players participating are stronger, hence a placing is harder to acheive.
However just because player A rated 2000 beats Player B rated 1900 in a GP3 and also in a GP1 does not mean he should receive a greater change in his rating.

Kevin Bonham
19-04-2004, 06:54 PM
However, some games results are more relavent than others for estimating a player's ability. For example, a 60m+10s club knockout games is taken less seriously by many players, than a 90m+30s at a far away venue. The Glicko treats both games as being equaly valuable data points.

A simple way to account for these types of phenominae is to have games weighted according to some algorithm that factors: home/away, length, championship, mean rating of the two players, et cetera. At the moment, the Glicko factors only: past performance, activity, and difference in rating.

There are indeed all kinds of things that might make useful additions to the model, but they still have to be tested empirically to see that they actually work.

An example of a counter-intuitive scenario for using game length: You might think that downweighting short games is a good idea, especially as it also solves junior-vs-junior problems (if any) by probably downweighting games against other juniors. However juniors more often improve their performances in faster games before they improve their performances at slow games. So their first signs of rapid improvement (either vs other juniors or in those faster games played against adults) could be downweighted by such a system, resulting in more conservative junior ratings.

chesslover
19-04-2004, 06:55 PM
You are the one with the theory that a 4 hr game is twice as good as a 2hr game.
The onus is on you to prove it, not me to disprove it.

that is a good point Supreme Leader :clap:

It is upto those who have the new theory to show the proof

chesslover
19-04-2004, 07:01 PM
Don't compare my idea with that inflationary "crap". Not only is what I am saying reasonable, but it has no analogy with that other "crap". So, lumping in my idea with that "crap" is gratuitous mud sport.

I suggest you put a ban on yourself from using the words, "logical" and "clearly", because you use them as weapons. Unfortunately, what you are defending is the status quo. And that ain't a good policy in a rapidly changing world.

It is good to see the old Matt back. Glad to see that the Valium you took has worn off after about 4 months ;)

But play hard, play long but play nice okay?

Alan Shore
19-04-2004, 07:11 PM
2. one kid had a single digit rating. luckily he thought it was hilarious to be rated lower than stanley the chimp, a chessmaster 9000 random move generator, but i think it's both a potential insult to give players ratings with fewer than 3 digits and also somewhat meaningless because it's easy to improve beyond that level.

That's nothing - kids have had negative ratings in the past! Kind of like saying 'I am worse than someone who has no idea of the moves of chess'. I think Tamzin Oliver was one of these negatives.



3. i think an RD floor for ratings under a few hundred (to keep them volatile) and a rating floor of somewhere between 100 and 300 should be considered. this reduces the problem of kids with ratings of 50 beating 800's, who then beat 1200 adults and so on up.

I'd support that if it were feasible, what do you think Bill?

PHAT
19-04-2004, 07:12 PM
Just supposition.
No it isn't, it is deduction.

I see I am going to have to s-p-e-l-l it out for your poor old brain.

It is a matter sample size. A longer game is a larger sample size of chess skill/thinking. Therefore it carries more information.

If you cannot concur on this simple fact, you have some fairly specific type brain damage.

Kevin Bonham
19-04-2004, 07:13 PM
A game that lasts 1 hour has half the predictive capacity of one that lasts 2 hours.

Half the predictive capacity for predicting what?

For predicting one hour games? Two hour games? Games played at a random mix of time limits?

I'm confident your claim is exaggerated no matter what the time limit, but it's not possible to make real progress in debating predictive capacity when the thing being predicted isn't even defined.

Alan Shore
19-04-2004, 07:22 PM
I disagree with that. The reason is the online environment and culture is simply more condusive to rapid and blitz play. I don't go out of my way to play rapid games and am even a member of the STC Bunch but play many more blitz games than standard games online.

This may be true. It may also be the case that this 'net culture' is overlapping to OTB chess. (FYI, I've played over 10,000 games on FICS and 0 standard!)

chesslover
19-04-2004, 07:25 PM
Half the predictive capacity for predicting what?

For predicting one hour games? Two hour games? Games played at a random mix of time limits?

I'm confident your claim is exaggerated no matter what the time limit, but it's not possible to make real progress in debating predictive capacity when the thing being predicted isn't even defined.

he means and he can correct, half the predictive capacity of a person's true rating and chess playing ability

jenni
19-04-2004, 07:25 PM
That's nothing - kids have had negative ratings in the past! Kind of like saying 'I am worse than someone who has no idea of the moves of chess'. I think Tamzin Oliver was one of these negatives.


Yes Tamzin was - I think her rating went down to something like -170, before she started climbing back. Which means her rating has gone up about 1200 points in 5 years!

Bill has fixed that problem - negative ratings are no longer possible and also until they perform in a positive region, no past history is collected.

Bill Gletsos
19-04-2004, 07:30 PM
That's nothing - kids have had negative ratings in the past! Kind of like saying 'I am worse than someone who has no idea of the moves of chess'. I think Tamzin Oliver was one of these negatives.
She was not the first to do so.
In fact I remember the first time a player appeared on the old ACF Elo system with a negative rating. Since it was an ACT junior I spoke to Jenni Oliver at the time about it and asked her was it even remotely possible that this rating was accurate. Jenni's response was something like "yes, that rating is reasonably accurate. They had scored something like 0.5 from around 27 games against an avergae rating of 500-600.


I'd support that if it were feasible, what do you think Bill?
Under Glicko2 there should be no need to artificially play with the RD since the Volatility factor should handle it.

Alan Shore
19-04-2004, 07:33 PM
The rating system and the GP are two different things.
The rating system is trying to indicate a players strength relative to others.

The GP system is allocating points based on where a player finishes in an event. The reason a GP3 is worth more poinst than a GP1 is because it is expected that the players participating are stronger, hence a placing is harder to acheive.
However just because player A rated 2000 beats Player B rated 1900 in a GP3 and also in a GP1 does not mean he should receive a greater change in his rating.

I agree, a game of chess is a game of chess and ratings should not be dependent upon where this game was played or how strong the field is - thank goodness it was changed in 1987 to what it is now.

Kevin Bonham
19-04-2004, 07:34 PM
No it isn't, it is deduction.

I see I am going to have to s-p-e-l-l it out for your poor old brain.

It is a matter sample size. A longer game is a larger sample size of chess skill/thinking. Therefore it carries more information.

If you cannot concur on this simple fact, you have some fairly specific type brain damage.

Good old proof by intimidation, where would we be without it? :p

The idea that a longer game must predict chess skill better follows if you assume the purpose of ratings is to predict performance in some kind of hypothetical condition where the players can take as long as they like - CC sans books, databases etc.

But this is not the purpose of ratings. Ratings measure and predict performance under tournament conditions.

Hence my comment above.

Bill Gletsos
19-04-2004, 07:34 PM
Yes Tamzin was - I think her rating went down to something like -170, before she started climbing back. Which means her rating has gone up about 1200 points in 5 years!
It was her first rating on the list back in Aug 1999 when she was -97. Of course she was not yet 8 years old at that stage.


Bill has fixed that problem - negative ratings are no longer possible and also until they perform in a positive region, no past history is collected.
Correct. ;)

Alan Shore
19-04-2004, 07:39 PM
Under Glicko2 there should be no need to artificially play with the RD since the Volatility factor should handle it.

I'm sorry I read the original message incorrectly.. I was not thinking of an RD floor but a rating floor, of around 100-300 rather than 0, to at least perhaps cut down the supposed problems with junior improvement. The QLD Junior rating system has a floor of about 500 I think (it's been changed around a couple of times).

Bill Gletsos
19-04-2004, 08:07 PM
No it isn't, it is deduction.

I see I am going to have to s-p-e-l-l it out for your poor old brain.

It is a matter sample size. A longer game is a larger sample size of chess skill/thinking. Therefore it carries more information.

If you cannot concur on this simple fact, you have some fairly specific type brain damage.
You can bluster all you like.
Your intimidation tactics wont work on me mate. ;)

jenni
19-04-2004, 08:11 PM
It was her first rating on the list back in Aug 1999 when she was -97. Of course she was not yet 8 years old at that stage.


Correct. ;)

Didn't she go down from that in the next ratings list? I seem to remember a 3 digit figure! She wasn't terribly interested in chess at that point and was pretty bad.

jenni
19-04-2004, 08:13 PM
She was not the first to do so.
In fact I remember the first time a player appeared on the old ACF Elo system with a negative rating. Since it was an ACT junior I spoke to Jenni Oliver at the time about it and asked her was it even remotely possible that this rating was accurate. Jenni's response was something like "yes, that rating is reasonably accurate. They had scored something like 0.5 from around 27 games against an avergae rating of 500-600.


Under Glicko2 there should be no need to artificially play with the RD since the Volatility factor should handle it.

I can't remember who the player was, but I remember the conversation and the player was pretty bad and definitely hundreds of points below the kids he was playing at Norths. I think the 1/2 point was from a stalemate.

Bill Gletsos
19-04-2004, 08:33 PM
I can't remember who the player was, but I remember the conversation and the player was pretty bad and definitely hundreds of points below the kids he was playing at Norths. I think the 1/2 point was from a stalemate.
I'll remind you next time we speak on the phone. ;)

Bill Gletsos
19-04-2004, 08:36 PM
I'm sorry I read the original message incorrectly.. I was not thinking of an RD floor but a rating floor, of around 100-300 rather than 0, to at least perhaps cut down the supposed problems with junior improvement. The QLD Junior rating system has a floor of about 500 I think (it's been changed around a couple of times).
We have a rating floor.
Its zero. ;)

Bill Gletsos
19-04-2004, 08:41 PM
Didn't she go down from that in the next ratings list? I seem to remember a 3 digit figure! She wasn't terribly interested in chess at that point and was pretty bad.
Nope in the next list (Dec 99) she went up 5 points.
And the 150 point uplift in April 2000 got her postive, although she would have gone positive anyway because she improved by 97 points on her own.

Alan Shore
19-04-2004, 08:45 PM
We have a rating floor.
Its zero. ;)

I know, that's why I said rather than 0. Are you adamant the 0 floor is the best or would you accept the possibility of a new rating floor if it would help?

Bill Gletsos
19-04-2004, 08:49 PM
I know, that's why I said rather than 0. Are you adamant the 0 floor is the best or would you accept the possibility of a new rating floor if it would help?
From a mathematical/statistical viewpoint there is no reason it shouldnt be zero.
From an ego viewpoint, maybe there is. Of course in that case you would still maintain the rating internally but just not publish ratings below xxx.

Cat
19-04-2004, 10:05 PM
As eagerly awaited as the first sightings of swallows in Spring, are the first few balls of a Test Match, or the first attempt by Matt to arc-up Bill in 2004.
The first ball was indeed head-high, but sadly it is down leg-side and Bill watches it whistle by, and takes the opportunity to sledge the wicket-keeper as well.

I'm afraid your cricketing metaphor was premature, as I'm still at the wicket - looks like MS has been on strike and scoring freely against a laboured attack bowling far too short with far too many no balls and wides. BG has only been supported sporadically by his main spinner KB and we're nearing a declaration, fully confident that Bill's team will be made to follow-on.

Cat
19-04-2004, 10:13 PM
how can any system 'keep up' when provided with insufficent and incomplete data?

Junior participation rates are not really a significant part of the cause of the problem on the Gold Coast Garvin, in that most of our juniors are reliably rated whereas it's the adult population that largely has the unstable ratings. Hence in open competition the ratings of the smaller number of adults tend to move towards the larger, more stable junior group, a reverse of Glickman's supposition.

Cat
19-04-2004, 10:32 PM
I think the 'problem' of underrated juniors is simply one of when players are most likely to improve - when they have time to study the game and have minimal distractions while playing their games ... children fit very neatly into this category.
As Bill says, just because someone is a junior does not mean that they will automatically improve, but the opposite also applies. Typically improvement happens when a player gets some coaching, plays more events (simply exposure to the game and learning from mistakes can be very beneficial) and is generally motivated to do well.
I think some of these factors may also apply to university students, as they too tend to have some time on their hands to study chess, and typically do not have quite the level of distarctions that others.
Once you get past university level, other factors tend to play a more important role in life, which takes away from any potential chess improvement. Things like work, family/relationships, financial needs, etc become prioritised, so that when one gets to the chess board, these things remain in the back of the mind, so that the focus can not be 100% chess.
One also has to keep in mind the 'dropout factor' of older kids - yes they might improve significantly, but if they stop playing chess when they hit 16, what good does it do to the rating pool as a whole if they are given extra points for being a junior if there is no chance of these points being redistributed amongst the rating pool?
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.
As for the kids mentioned, I think Jessica Kinder's game has improved enormously since the Australian Clubs Teams Championship! :D ;)

Kerry I think you're missing the point. There is something innately biological furthering child development in all arenas, something that is not present in University students or enthusiastic adults.

For example, one would not expect a 12 yr old to match up against a team of 15 year old in Rugby, he's simply not physically capable. Give him another 18 months and he's a different child. His physical prowess at 12 is a poor predictor of his performance at 15. The intellect develops in much the same way. There are no parallels to find at Unversity, at 25 or at any other time of life. Early puberty is unique.

Tennis tends to disregard aged performances of their protegees, recognising only relatively current results are reliably indicative of performance.

Of course juniors progress at different rates, but again you are entirely missing the point. One has to consider the overall dynamics of the population pool. The adult pool is stable, one improver usually offset by another who is declining. The junior pool is quite different, as a cohort displaying completely different dynamics. My observations from the results at the Gold Coast Classic and elsewhere is that as a cohort, they improve at around 50-60 points /year. It requires a different set of dynamics to properly describe this quite different population.

Is 50/60 points /year really that surprising? It would suggest a junior rated around 800 age 8 will be on average rated around 1300-1400 age 18.

Again, I have not suggested these children recieve these rating points as a gratuity, but rather the dynamics of the system reflects this natural dynamic trend to ensure through performance their progress is more accurately reflected.

Cat
19-04-2004, 10:35 PM
As David Richards says, there is so much information that can be used in ratings calculations, but it is not used because such theorectical factors based on meta analysis - factors that Bill denies exist.


Yes the problem in a nutshell!

Cat
19-04-2004, 10:46 PM
Factors that David cannot adequately quantify.

Well I've just quantified a biological trend haven't I? Interestingly though, when I examined the data from the Gold Coast Classic I could find no correlation between rating reliability and predictive performance, over the 4 year I analysed. This was true of all rating groups.

One could view RD as the error bars on a graph. In the Glicko system the error bars have simply been drawn far too large, in the interests of creating a dynamic engine. No one is arguing that observational data increases reliability of the recorded result, but rather that this is hugely overemphasised.

Cat
19-04-2004, 11:03 PM
[=Kevin Bonham]Do Resistance members qualify?

QUOTE]

Fair call


[QUOTE]You're the one claiming to have the information. Produce a proposed improvement to the system that can be tested and if it's tested and works I'll listen

From what I've said above I'd suggested;

1. Skewed variance for juniors permitting an average inflationary effect within the junior population of around 50-60 points /year assuming average participation rates.

2. A reduced effective k for higher rated adults.

3. Ratings to be nationally standardised by cross-regional sampling.

4. FIDE style entry into the rating system, that is ACF rating only given when performance reaches entry standard (say 500-600), to prevent deflationary drag and avoid the awarding of insulting ratings.

PHAT
19-04-2004, 11:08 PM
The idea that a longer game must predict chess skill better follows if you assume the purpose of ratings is to predict performance in some kind of hypothetical condition where the players can take as long as they like ...


Kevin, I just do not get this quote. What boundry condition are you trying to claim applies with, "...in some kind of hypothetical condition where the players can take as long as they like ..."?

Bill Gletsos
19-04-2004, 11:11 PM
Junior participation rates are not really a significant part of the cause of the problem on the Gold Coast Garvin, in that most of our juniors are reliably rated whereas it's the adult population that largely has the unstable ratings. Hence in open competition the ratings of the smaller number of adults tend to move towards the larger, more stable junior group, a reverse of Glickman's supposition.
This seems to contradict what you said previously. You said :

Juniors on the Gold Coast generally compete in Junior competition far more often than they do in open competition - FACT
Adults are more likely to be playing another adult at the ADULT club - FACT.

If as you said to starter:

I'm afraid your cricketing metaphor was premature, as I'm still at the wicket
The umpires finger has gone up.
Unfortunately for you, you have been caught at slips.

Bill Gletsos
19-04-2004, 11:11 PM
Kerry I think you're missing the point.
Why is it Kerry and apparently everyone else is missing the point.
It seems if anyone has the wrong point it is you.



There is something innately biological furthering child development in all arenas, something that is not present in University students or enthusiastic adults.

For example, one would not expect a 12 yr old to match up against a team of 15 year old in Rugby, he's simply not physically capable. Give him another 18 months and he's a different child. His physical prowess at 12 is a poor predictor of his performance at 15. The intellect develops in much the same way. There are no parallels to find at Unversity, at 25 or at any other time of life. Early puberty is unique.

Tennis tends to disregard aged performances of their protegees, recognising only relatively current results are reliably indicative of performance.

Of course juniors progress at different rates, but again you are entirely missing the point. One has to consider the overall dynamics of the population pool. The adult pool is stable, one improver usually offset by another who is declining.
An over generlisation not based in fact.



The junior pool is quite different, as a cohort displaying completely different dynamics. My observations from the results at the Gold Coast Classic and elsewhere is that as a cohort, they improve at around 50-60 points /year. It requires a different set of dynamics to properly describe this quite different population.
This is just plain and simple rubbish.
The any given junior over a 12 mth period may improve, may not improve and some even decline.


Is 50/60 points /year really that surprising? It would suggest a junior rated around 800 age 8 will be on average rated around 1300-1400 age 18.
Its not surprising, its just pure and simple rubbish.


Again, I have not suggested these children recieve these rating points as a gratuity, but rather the dynamics of the system reflects this natural dynamic trend to ensure through performance their progress is more accurately reflected.
There is no way to predict when a junior will make a sudden leap in skill.
The only indicator that a junior has improved is their results, not their age.

Bill Gletsos
19-04-2004, 11:12 PM
Yes the problem in a nutshell!
Looks like two nuts in the one shell.

Bill Gletsos
19-04-2004, 11:19 PM
From what I've said above I'd suggested;

1. Skewed variance for juniors permitting an average inflationary effect within the junior population of around 50-60 points /year assuming average participation rates.
No basis in fact.


2. A reduced effective k for higher rated adults.
This was implemented last year for players over 2200.


3. Ratings to be nationally standardised by cross-regional sampling.
Not necessarily justifiable.
Too many variables.
The benefits probably dont warrant the complexity.


4. FIDE style entry into the rating system, that is ACF rating only given when performance reaches entry standard (say 500-600), to prevent deflationary drag and avoid the awarding of insulting ratings.
One of your dumber ideas.
Just when FIDE is moving to lowering their rating cutoffs you want to implement one where none exists.
What deflationary drag.
If the US can gave ratings down to 100 why cant we.

Cat
19-04-2004, 11:22 PM
Bill Gletsos - This seems to contradict what you said previously.

The 2 statements are not mutually exclusive. GG seemed to believe lack of junior participation in ACF rated tournaments was the major problem on the Gold Coast. While we could be doing better, the participation rates are not that bad. As I say juniors encounter juniors more often than they do adults, but they still play adults.


If as you said to starter:

The umpires finger has gone up.
Unfortunately for you, you have been caught at slips.

I'm afraid it was another no ball :snooty:

Bill Gletsos
19-04-2004, 11:29 PM
The 2 statements are not mutually exclusive. GG seemed to believe lack of junior participation in ACF rated tournaments was the major problem on the Gold Coast. While we could be doing better, the participation rates are not that bad. As I say juniors encounter juniors more often than they do adults, but they still play adults.
But thats not what you said.
Your statement:

Hence in open competition the ratings of the smaller number of adults tend to move towards the larger, more stable junior group, a reverse of Glickman's supposition.
implies that the adults are more likely to play juniors than adults, however you said:

Juniors on the Gold Coast generally compete in Junior competition far more often than they do in open competition - FACT
Adults are more likely to be playing another adult at the ADULT club - FACT.
Hence they are totally contradictory.



I'm afraid it was another no ball :snooty:
Your eyesights failed and your hearing.
No no ball call was made.
The umpires sent you to the pavillion.

Cat
19-04-2004, 11:30 PM
The any given junior over a 12 mth period may improve, may not improve and some even decline.

Of course, I'm talking about a population dynamic, not an individual.


There is no way to predict when a junior will make a sudden leap in skill.

No, but one can predict a population dynamic from repeated observation.



The only indicator that a junior has improved is their results,

And therein lies the crux of the problem. The data is historical and continually and significantly lags development.

Alan Shore
19-04-2004, 11:30 PM
1.No basis in fact.


2.This was implemented last year for players over 2200.


3.Not necessarily justifiable.
Too many variables.
The benefits probably dont warrant the complexity.


4.One of your dumber ideas.
Just when FIDE is moving to lowering their rating cutoffs you want to implement one where none exists.
What deflationary drag.
If the US can gave ratings down to 100 why cant we.

I am enjoying the cricket match here. Bill had bowled three successive maidens I believe with the first three, yet David might have scored some runs off the final over of the spell. David believed it may prevent deflationary drag and avoid the awarding of insulting ratings. The latter is a subjective thing and I don't see how you could be insulted by a number now that it's going to be a positive integer, hehe. As for the former, well, perhaps this has some merit. Obviously the FIDE cutoffs are incomparable with those of the ACF since implementing a floor of 200 would not be cutting out significant proportions of the chess playing community, unlike the FIDE threshold (which I believe should come down even more). I would suggest, if you perform at below a 200 rating, you are not good enough to deserve a rating and your results should be dismissed until you prove you can play at 200+ strength. This would be at least a step in reducing this (possible?) deflation at that playing level.

PHAT
19-04-2004, 11:31 PM
This seems to contradict what you said previously. You said :


If as you said to starter:

The umpires finger has gone up.
Unfortunately for you, you have been caught at slips.

DR cannot be given out on a no ball. There is no condradiction in the two quotes. You are making up bullshit to divert the discussion. Lets get back on track.

Senario: You enter a guessing competition where you have to guess which player is best. The only information you can be given is the the result of either their last
bullit, or
blitz, or
rapid, or
90m+30s,

The result of which game do you choose to be told?

Cat
19-04-2004, 11:38 PM
This was implemented last year for players over 2200.

Yes, and in response to Kevins question I'd extend it. It was a good idea Bill.


Not necessarily justifiable.
Too many variables.
The benefits probably dont warrant the complexity.

Well you won't know until you find out, will you?


One of your dumber ideas.

And I suppose giving a child a negative rating or the rating of a chimpanzee is a smart thing to do, do you Bill?


What deflationary drag.

What happens when Elizabeth with a rating of -50 beats Sam with a rating of 550? What happens to Sams rating, Bill?


If the US can gave ratings down to 100 why cant we.

Ah, the Alexander Downer argument!

Rincewind
19-04-2004, 11:41 PM
From a mathematical/statistical viewpoint there is no reason it shouldnt be zero.
From an ego viewpoint, maybe there is. Of course in that case you would still maintain the rating internally but just not publish ratings below xxx.

Froma mathematical point of view there isprobably no need for a floor at all. However, having one has practical advantages and introduces very few errors into the system consider the number and activity of impacted players.

Bill Gletsos
19-04-2004, 11:43 PM
DR cannot be given out on a no ball. There is no condradiction in the two quotes. You are making up bullshit to divert the discussion. Lets get back on track.
Of course there is a difference.
As you yourself said earlier :

If you cannot concur on this simple fact, you have some fairly specific type brain damage. :owned:

Senario: You enter a guessing competition where you have to guess which player is best. The only information you can be given is the the result of either their last
bullit, or
blitz, or
rapid, or
90m+30s,

The result of which game do you choose to be told?[/QUOTE]
Irrelevant and misleading.
If I want a blitz rating I'd use the blitz game.
If I wanted the rapid I'd use the rapid game.
If I wanted a normal I'd uzse the normal game

However if I had a game at G/60 and one at G/90 I'd use both.

Cat
19-04-2004, 11:47 PM
I am enjoying the cricket match here. Bill had bowled three successive maidens I believe with the first three, yet David might have scored some runs off the final over of the spell. David believed it may prevent deflationary drag and avoid the awarding of insulting ratings. The latter is a subjective thing and I don't see how you could be insulted by a number now that it's going to be a positive integer, hehe. As for the former, well, perhaps this has some merit. Obviously the FIDE cutoffs are incomparable with those of the ACF since implementing a floor of 200 would not be cutting out significant proportions of the chess playing community, unlike the FIDE threshold (which I believe should come down even more). I would suggest, if you perform at below a 200 rating, you are not good enough to deserve a rating and your results should be dismissed until you prove you can play at 200+ strength. This would be at least a step in reducing this (possible?) deflation at that playing level.

So BD comes on at the Pavillion End and promptly bowls a series of wides. But in the last ball of his over comes in with a well pitched delivery which Richards dollies to BG at cover, only for the hapless chap to fumble - so ends the over, Richards perillously close to his double century in a 350 partnership with Sweeney.

Bill Gletsos
19-04-2004, 11:49 PM
Froma mathematical point of view there isprobably no need for a floor at all. However, having one has practical advantages and introduces very few errors into the system consider the number and activity of impacted players.
On the March 2003 active list dropping everyone below 500 would effect 174 players.

Rincewind
19-04-2004, 11:53 PM
On the March 2003 active list dropping everyone below 500 would effect 174 players.

I was talking about the present floor. How many players has it affected? My guess would be a handful.

Alan Shore
19-04-2004, 11:59 PM
So BD comes on at the Pavillion End and promptly bowls a series of wides. But in the last ball of his over comes in with a well pitched delivery which Richards dollies to BG at cover, only for the hapless chap to fumble - so ends the over, Richards perillously close to his double century in a 350 partnership with Sweeney.

Being a part-time bowler who's just scored a 222* in the first innings, I'd be happy to almost claim your wicket David :p

Cat
20-04-2004, 12:02 AM
Being a part-time bowler who's just scored a 222* in the first innings, I'd be happy to almost claim your wicket David :p

I guess that's my signal for a declaration. Good-night - the wicket's looking a little worn, I can't wait to bowl to you tomorrow.

Bill Gletsos
20-04-2004, 12:03 AM
Yes, and in response to Kevins question I'd extend it. It was a good idea Bill.
No justification.
In fact testing showed that it worsened the predictive accuracy.
If course I stated this previously in an earlier thread.
Guess you were asleep at the wheel at the time.



Well you won't know until you find out, will you?
Maybe, maybe not.


And I suppose giving a child a negative rating or the rating of a chimpanzee is a smart thing to do, do you Bill?
You really are a goose.
Jenni made it clear in an earlier post that we dont allocate negative ratings.
Under the old ACF Elo some players had negative ratings which were published.
Under the early Glicko lists we did still allocate negative ratings but we never published them.

As for chimps if a child plays like a chimp then I see no problem in them getting a similar rating. After all its based on results.
Also there are some damn smart chimps. Just look at those launched into space.

In fact in Planet of the Apes some were doctors. :owned:


What happens when Elizabeth with a rating of -50 beats Sam with a rating of 550? What happens to Sams rating, Bill?
Well given we dont allocate negative ratings Elizabeth couldnt have one no could she.
But lets say we did.
Sam's rating would go down.
He did after all lose to a lower rated oppient.



Ah, the Alexander Downer argument!
Nice try but no cigar.
As I said FIDE are dropping their rating floors. progressively.
The only reason they wont drop it in one go is that they could not handle the work load.

PHAT
20-04-2004, 12:05 AM
Irrelevant and misleading.
If I want a blitz rating I'd use the blitz game.
If I wanted the rapid I'd use the rapid game.
If I wanted a normal I'd uzse the normal game

However if I had a game ay G/60 and one at G/90 I'd use both.

You have not answered the f.....g question

If the senario is irrelevant you can still answer it.
If the senario is misleading you can still answer it.

It is a hypothetical "Guessing Comp".
Which game do you pick to help you guess which player is better. Come on diddums I am sure you can choose. If you are too frightened or dull to enter, just squib again. Of course, others may see your piking as an indication of panic.

BTW I could just as easily given the choice of a 15min, 20, 25 and 30 rapid. The principle is the same. :rolleyes:

Bill Gletsos
20-04-2004, 12:09 AM
I was talking about the present floor. How many players has it affected? My guess would be a handful.
In March 2004 it was 35 in the normal list, 20 in the rapid.
However in Dec 2003 it was 9 in the normal list and 334 in the rapid.

Bill Gletsos
20-04-2004, 12:10 AM
So BD comes on at the Pavillion End and promptly bowls a series of wides. But in the last ball of his over comes in with a well pitched delivery which Richards dollies to BG at cover, only for the hapless chap to fumble - so ends the over, Richards perillously close to his double century in a 350 partnership with Sweeney.
You really are having yourself on.

You are an even bigger goose than I imagined.

Kevin Bonham
20-04-2004, 12:11 AM
Kevin, I just do not get this quote. What boundry condition are you trying to claim applies with, "...in some kind of hypothetical condition where the players can take as long as they like ..."?

Well, you're arguing that more time means more evidence of chess skill (I'm not sure if you were arguing that because you thought more time meant more moves as well). But games played at extremely long time limits would probably provide less reliable evidence of performance at average time limits than games played at those time limits. You seemed to be pushing an idea of "chess skill" that's dissassociated from "tournament chess skill".

Bill Gletsos
20-04-2004, 12:12 AM
You have not answered the f.....g question
I answered the question.
You just did not like my answer.

Thats your problem not mine, diddums. :hand:

PHAT
20-04-2004, 12:16 AM
I answered the question.
You just did not like my answer.


No, you made a responce to the question but you have not answered it. If you do not answer it, all of us will assume you are snookered.

Alan Shore
20-04-2004, 12:21 AM
In March 2004 it was 35 in the normal list, 20 in the rapid.
However in Dec 2003 it was 9 in the normal list and 334 in the rapid.

334!! wow.. a true Mark Taylor/Don Bradman effort.

I don't think David noticed my 222* was supposed to correspond to my post count.. ah well :)

PHAT
20-04-2004, 12:24 AM
I'm not sure if you were arguing that because you thought more time meant more moves as well.

No, more time increases the reliability of the result, independant of the number of moves.


But games played at extremely long time limits would probably provide less reliable evidence of performance at average time limits than games played at those time limits.

I am not and have not spoken of "extremely long time limits". If I was, then I would agree that it would not be predicting tournament skill. It would be like comparing Corr to OTB.

Now, would you like to try your hand at the "Guessing Comp" or ar you as craven as Bill.

Bill Gletsos
20-04-2004, 12:25 AM
334!! wow.. a true Mark Taylor/Don Bradman effort.

I don't think David noticed my 222* was supposed to correspond to my post count.. ah well :)
Well as I noted earlier he is not very observant. ;)

Kevin Bonham
20-04-2004, 12:28 AM
Tennis tends to disregard aged performances of their protegees, recognising only relatively current results are reliably indicative of performance.

Tennis performance varies far more rapidly by age, and probably also by other factors like fitness, coaching and "form" than chess.

Rincewind
20-04-2004, 12:29 AM
Senario: You enter a guessing competition where you have to guess which player is best. The only information you can be given is the the result of either their last
bullit, or
blitz, or
rapid, or
90m+30s,

The result of which game do you choose to be told?

Best at what?

Bill Gletsos
20-04-2004, 12:30 AM
No, you made a responce to the question but you have not answered it. If you do not answer it, all of us will assume you are snookered.
No based on your lack of information in the question, I answered it.
If you cannot frame a logical ly relevant question that is not my problem.
You asked who is better.
You did not state what type of game you wanted to know who was better at.

Clearly anyone who was going to use a single game as their only data point would use the one that best answered the question they were trying to answer.

Under those conditions I answered the question.

Bill Gletsos
20-04-2004, 12:30 AM
Best at what?
Exactly.

PHAT
20-04-2004, 12:32 AM
Best at what?

Best at chess, Einstein. Get with the program.

Rincewind
20-04-2004, 12:33 AM
Best at chess, Einstein. Get with the program.

Define "chess"?

Bill Gletsos
20-04-2004, 12:35 AM
Best at chess, Einstein. Get with the program.
Einstein is with the program.
Its the quizmaster who has been smelling his socks and lost the plot. ;)

Alan Shore
20-04-2004, 12:35 AM
Haha, true, Lightning, Blitz, Rapid and Standard are still all 'chess'.

Kevin Bonham
20-04-2004, 12:38 AM
Yes, and in response to Kevins question I'd extend it. It was a good idea Bill.

Lower effective k factor over too much of the system would certainly result in a decline in predictive efficiency. That's one of the points of having a more active system. Even with the top player tweak there was a small decline.

I'm also not convinced 500 is a good ratings floor. Too inflationary. It's best to put ratings floors at a level that virtually no one is below. Zero is good. Maybe 100 but no higher.

As for the cricket match, even if we did follow up, we could still knock up 400 in quick time and then my infrequent spin should still be good for 5-20 in the final innings if the other bowlers don't run through you first.


And I suppose giving a child a negative rating or the rating of a chimpanzee is a smart thing to do, do you Bill?

That anecdote dealt with an example of a program that supposedly mimicked a chimpanzee's play. Such a program would not get a rating in Australia. If Chessmaster 9000 gives it a rating above zero then Chessmaster's rating is flawed. A DOP including a strictly random move program (if it was random) in Australia should have seeded it with rating zero.

PHAT
20-04-2004, 12:40 AM
No based on your lack of information in the question, I answered it.
If you cannot frame a logical ly relevant question that is not my problem.
You asked who is better.
You did not state what type of game you wanted to know who was better at.


1. We are always making choices with incomplete knowledge - the lack of information is absolutley normal.
2. The question is unfaulted.
3. I did state the type of game - the game of chess.

Come on, show us what you're packin'. Answer the question.

PHAT
20-04-2004, 12:43 AM
Its the quizmaster who has been smelling his socks and lost the plot. ;)

Answer the question - which game result do you choose to be told?

PHAT
20-04-2004, 12:47 AM
Define "chess"?
A board game for two players, each beginning with 16 pieces of six kinds that are moved according to individual rules, with the objective of checkmating the opposing king.

Any other stupid questions?

Bill Gletsos
20-04-2004, 12:47 AM
1. We are always making choices with incomplete knowledge - the lack of information is absolutley normal.
2. The question is unfaulted.
3. I did state the type of game - the game of chess.

Come on, show us what you're packin'. Answer the question.
Your question is simply stupid.
Lack of information is not normal.
One may not know the answer the the aim of any fact finding exercise is to adequately define the question.
In that regard you have failed.

When I say I want to know how good a chess player someone is, I want to know how good he is at a particular type of chess.
In which case I'll pick the game closest to that style.

If I am forced to pick not knowing which brand I want then I would pick at random as the question is flawed, hence a random choice is as good as any.

Rincewind
20-04-2004, 12:49 AM
1. We are always making choices with incomplete knowledge - the lack of information is absolutley normal.
2. The question is unfaulted.
3. I did state the type of game - the game of chess.

Come on, show us what you're packin'. Answer the question.

Just saying chess is to vague. You are assuming it is impossible for someone to better better at blitz and worse at standard than someone else at the same time. This is patently false.

So you are either prejudiced into thinking that one form of the game is somehow innately superior to the other, or else you must admit that your definition of just "chess" is too imprecise.

However, if you are going to stick with it then I would say it makes absolutely no difference which you pick. Toss a coin, they all give you the same information about "chess" as the term is a vague one.


However, this has little to do with the subject at hand as we know precisely the type of chess we want to predict with the rating systems.

Rincewind
20-04-2004, 12:53 AM
A board game for two players, each beginning with 16 pieces of six kinds that are moved according to individual rules, with the objective of checkmating the opposing king.

Any other stupid questions?

Yes, do players have a limited amount of time in which to play their moves?

PHAT
20-04-2004, 12:54 AM
Your question is simply stupid.
Lack of information is not normal.
One may not know the answer the the aim of any fact finding exercise is to adequately define the question.
In that regard you have failed.

When I say I want to know how good a chess player someone is, I want to know how good he is at a particular type of chess.
In which case I'll pick the game closest to that style.

If I am forced to pick not knowing which brand I want then I would pick at random as the question is flawed, hence a random choice is as good as any.

You are such an eel. Or maybe you need a warmup question to get you into a success state of mind.

Q. Who is better at chess, me or Ian Rogers.

Oo, oo, oo. I am soooo excited. Can you get his question right or don't you have enough information.

PHAT
20-04-2004, 12:57 AM
Yes, do players have a limited amount of time in which to play their moves?

If a limit is set, then yes. If not, no.

Any other stupid questions?

Rincewind
20-04-2004, 01:01 AM
If a limit is set, then yes. If not, no.

Any other stupid questions?

Yes,

Player A has a normal rating of 1700 and rapid rating of 1500

Player B has a normal rating of 1500 and rapid rating of 1700

RD's and volatilities are very low for both players (highly reliable ratings and consistent performers)

Which player is better at "chess"?

Bill Gletsos
20-04-2004, 01:03 AM
You are such an eel. Or maybe you need a warmup question to get you into a success state of mind.

Q. Who is better at chess, me or Ian Rogers.

Oo, oo, oo. I am soooo excited. Can you get his question right or don't you have enough information.
If you are just going to sit there, ask stupid questions and play silly games then I'm just going to sit back and let you play with yourself. :whistle:

PHAT
20-04-2004, 01:07 AM
Just saying chess is to vague. ...So you are either prejudiced into thinking that one form of the game is somehow innately superior to the other, or else you must admit that your definition of just "chess" is too imprecise.


As a mathematician, you know that modelling complex systems is tricky. We offen have to make assumptions and approximations in order to generate a solution. I do not care if you think "chess" is too vague or imprecise. The question is as it stands. If Bill even plucks up the courage to answer it, he may in fact find that longer games are worth more as a data point than shorter games.


However, if you are going to stick with it then I would say it makes absolutely no difference which you pick. Toss a coin, they all give you the same information about "chess" as the term is a vague one.


Now you are bullshitting to yourself. If the prize was a squillion, you would find some way to pick one game result ahead of the others.

Alan Shore
20-04-2004, 01:12 AM
Yes,

Player A has a normal rating of 1700 and rapid rating of 1500

Player B has a normal rating of 1500 and rapid rating of 1700

RD's and volatilities are very low for both players (highly reliable ratings and consistent performers)

Which player is better at "chess"?

Excellent comparision :clap:

For example, I fall into category B, a good friend of mine is in category A. We drew our last tournament game and play a lot of friendly games with quite even results.

Bill Gletsos
20-04-2004, 01:12 AM
As a mathematician, you know that modelling complex systems is tricky. We offen have to make assumptions and approximations in order to generate a solution. I do not care if you think "chess" is too vague or imprecise. The question is as it stands.
I see no reason why we should waste our time with you on this.


If Bill even plucks up the courage to answer it, he may in fact find that longer games are worth more as a data point than shorter games.
I answered it.



Now you are bullshitting to yourself. If the prize was a squillion, you would find some way to pick one game result ahead of the others.
The only person bullshitting here is you.
Barry picked a result. Like me he picked it randomly, because your question is imprecise and flawed.

PHAT
20-04-2004, 01:19 AM
Yes,

Player A has a normal rating of 1700 and rapid rating of 1500

Player B has a normal rating of 1500 and rapid rating of 1700

RD's and volatilities are very low for both players (highly reliable ratings and consistent performers)

Which player is better at "chess"?

On that info, for that question, I will contend that they are equil.

However!! if these players were given their two ratings (rapid and normal) based on just one game each, then I would say that player A was probably better. I justify this by saying the sample size of thinking is higher in the noraml rating and thus is more reliable.

PHAT
20-04-2004, 01:24 AM
Excellent comparision :clap:

No it is not, because it gives too much of the wrong information. It cannot illucidate the general principle of sample size effecting reliability. Now get you tounge out of Bazza's date.

Alan Shore
20-04-2004, 01:26 AM
On that info, for that question, I will contend that they are equil.

However!! if these players were given their two ratings (rapid and normal) based on just one game each, then I would say that player A was probably better. I justify this by saying the sample size of thinking is higher in the noraml rating and thus is more reliable.

Garbage, you can't make that assumption at all - all you can say is that Player A is better with longer time controls.

Rapid chess requires different skills, it requires you to think faster, know your position better at a glance without having aeons of time to calculate 10 moves in advance fritz-like variations and to utilise the resources you have, regarding time and position. To me it makes the game more exciting and more realistic in the sense of a battle. Therefore Matthew, anyone who devalues rapid chess against the interminable snail pace of a long time control hasn't got a clue.

Alan Shore
20-04-2004, 01:29 AM
I don't follow Matthew, are you saying Rapid ratings have insufficient sample size? Perhaps you can 'illucidate' (sic) further?

PHAT
20-04-2004, 01:29 AM
If you are just going to sit there, ask stupid questions and play silly games then I'm just going to sit back and let you play with yourself. :whistle:

Don't you know who is better at chess, me or Rogers.

Admit it, you cannot answer the question because it is devestating to your position. You have been done like a Sunday dinner. You know you have to answer 90m+30s but your ego cannot let you say it.

Rincewind
20-04-2004, 01:30 AM
On that info, for that question, I will contend that they are equil.

However!! if these players were given their two ratings (rapid and normal) based on just one game each, then I would say that player A was probably better. I justify this by saying the sample size of thinking is higher in the noraml rating and thus is more reliable.

You first sentence is an attempt at an each-way bet. However, your second sentence reveals you bias towards the longer for of the game.

What about if a vital part of the game is to beable to think in time pressure or calculate variations accurately once without the need to check three or four times? These are both skills you devalue because you are favouring a result in a game with a more leisurely time control. However, they are valid "chess" skills, particularly at faster time limits. They are also skills which normally accompany professional chess players.

PHAT
20-04-2004, 01:45 AM
... you can't make that assumption at all - all you can say is that Player A is better with longer time controls.


It is true that we could guess that A is better at longer games than B. However, that does not answer the question, who is better at "chess"? If these players had had only one rapid and one normal game, the single long game predicts rank in "chess" more reliably than the single rapid game.




Therefore Matthew, anyone who devalues rapid chess against the interminable snail pace of a long time control hasn't got a clue.

Sorry mate, but you have taken a wrong path. I am not making any claims about the goodness/badness of the various time controls. They are all chess, each style appealing to a different taste.

PHAT
20-04-2004, 02:00 AM
What about if a vital part of the game is to beable to think in time pressure or calculate variations accurately once without the need to check three or four times? These are both skills you devalue because you are favouring a result in a game with a more leisurely time control.

I am not and have not been degrading, denagreating or devaluing the shorter time controls. I am only saying that longer ones are better predictors.

Let's simplify ths even further. Question: What game result is the better predictor for 25 minute rapid games; a 24 minute rapid game or a 26 minute? Answer: Obviously the 26 minute because, it is based on 8% more play than the 24, for what is essentially the same game length.

From this example we can say that: As a general pinciple, longer chess games are better predictors than shorter chess games.



I can't believe that I have to explain this to you lot. :wall:

Alan Shore
20-04-2004, 02:07 AM
I am not and have not been degrading, denagreating or devaluing the shorter time controls. I am only saying that longer ones are better predictors.

Let's simplify ths even further. Question: What game result is the better predictor for 25 minute rapid games; a 24 minute rapid game or a 26 minute? Answer: Obviously the 26 minute because, it is based on 8% more play than the 24, for what is essentially the same game length.

From this example we can say that: As a general pinciple, longer chess games are better predictors than shorter chess games.



I can't believe that I have to explain this to you lot. :wall:

I disagree: by your logic, a 27 minute game would be a better predictor than a 24 minute one. Now do you see the problem?

Kevin Bonham
20-04-2004, 02:39 AM
Now, would you like to try your hand at the "Guessing Comp" or ar you as craven as Bill.

The only cowardice I see here is the way you throw the word "craven" around for those who decline to answer inadequately specified questions. :whistle: :hand:

Kevin Bonham
20-04-2004, 03:10 AM
It is true that we could guess that A is better at longer games than B. However, that does not answer the question, who is better at "chess"? If these players had had only one rapid and one normal game, the single long game predicts rank in "chess" more reliably than the single rapid game.

I dispute this. Indeed since you are making such a big silly show of how those who are not answering are craven and how yours is the one true answer, I am going to play your game and try to beat you at it. Try this for spin:


Senario: You enter a guessing competition where you have to guess which player is best. The only information you can be given is the the result of either their last
bullit, or
blitz, or
rapid, or
90m+30s,

The result of which game do you choose to be told?

I'm going with rapid.

Why? Because when you're just taking a single game you want the one that the stronger player generally is more likely to win (and let's ignore your imprecision and just agree to take this across a range of time limits, let's say the stronger player is stronger no matter what the time limit is).

I'm thinking the increased randomisation in the rapid is likely to be about balanced out by the extent to which the stronger player's natural advantages are amplified by the clock. All those things that the strong player knows how to do automatically, while the weak player needs to think carefully and still may not get it right, will come to the fore here. I may be overstating it but it doesn't matter, because there's something else you've overlooked.

If you want to predict who is better from one game, the last thing you want is a draw. And the faster the time limit the less the risk of that.

It may be that the correct answer is even not "rapid" but "blitz" for that reason.

Didn't even think about that one, did you? :naughty:

Over sufficiently many games, the slower time limit will probably give you the most accurate indication of the skill difference between the players. Asking that though, is not necessarily the same thing as asking what will give you the best indication of who is better.

Ian Rout
20-04-2004, 08:43 AM
I have some sympathy with Matthew's view, on the basis that the skills in playing "normal" chess can be adapted to faster chess by playing more pragmatically and using time better, and some of the skills needed to play fast chess can be learned where they are lacking. Hence the good normal player has the potential to be better at the two forms overall.

On the other hand some of the skills and techniques in fast chess cannot be productively transferred to normal chess. For instance gaining half a second per move from the speed of your arm swing, while a genuine skill, is of minimal use in a longer game. There are also many skills in fast chess that might be considered non-legitimate and are not (or should not be) transferrable at all, for example knocking pieces to the floor at a crucial time, putting pieces hanging over two squares and deciding later where it is, judging the right time to deliberately make an illegal or bad move that your opponnent will miss, keeping your finger on the button while the opponent is trying to press the clock, engaging in a constant stream of irritating banter and commentary with spectators, castling with a Rook from the next board etc.

PHAT
20-04-2004, 08:51 AM
The only cowardice I see here is the way you throw the word "craven" around for those who decline to answer inadequately specified questions.

You couldn't possibly have the stupidity to think and say that I am showing "ignoble fear in the face of danger or pain." Sometimes Kevin you just blurt the first stupid crappy hackneyed come-back line that comes into your head and it falls way short of the mark. I almost, but not quite, feel insulted that you have made such scant effort to insult me.

Cat
20-04-2004, 09:05 AM
No justification.
In fact testing showed that it worsened the predictive accuracy.
If course I stated this previously in an earlier thread.
Guess you were asleep at the wheel at the time.

Your own figures demonstrated that the lower Glicko dynamic yeilded virtually the same result as the ELO system. If it worsens predictive accuracy I would imagine it would be marginal, and as you have suggested previously, given its a relatively stable population, in time the differences would merge. Of course it depends as well as to how throughly you performed your analysis. Perhaps you could demonstrate your method?







Jenni made it clear in an earlier post that we dont allocate negative ratings.
Under the old ACF Elo some players had negative ratings which were published.
Under the early Glicko lists we did still allocate negative ratings but we never published them.

As for chimps if a child plays like a chimp then I see no problem in them getting a similar rating. After all its based on results.
Also there are some damn smart chimps. Just look at those launched into space.


Well given we dont allocate negative ratings Elizabeth couldnt have one no could she.
But lets say we did.
Sam's rating would go down.
He did after all lose to a lower rated oppient.


Are you seriously suggesting a rating of (say) 50 in a novice means anything at all? Its entirely meaningless, a complete abstraction serving absolutely no purpose whatsoever. Not to mention the discouraging effect it has on the child's morale. Bill, you've done reasonably well so far in controlling your tendancy to verbal abuse. I think everyone is tired of your tirades. Please continue to stick with the arguments.

Cat
20-04-2004, 09:17 AM
Lower effective k factor over too much of the system would certainly result in a decline in predictive efficiency. That's one of the points of having a more active system. Even with the top player tweak there was a small decline.

I'm also not convinced 500 is a good ratings floor. Too inflationary. It's best to put ratings floors at a level that virtually no one is below. Zero is good. Maybe 100 but no higher.

As for the cricket match, even if we did follow up, we could still knock up 400 in quick time and then my infrequent spin should still be good for 5-20 in the final innings if the other bowlers don't run through you first.



That anecdote dealt with an example of a program that supposedly mimicked a chimpanzee's play. Such a program would not get a rating in Australia. If Chessmaster 9000 gives it a rating above zero then Chessmaster's rating is flawed. A DOP including a strictly random move program (if it was random) in Australia should have seeded it with rating zero.

Lower effective-k would only apply to the higher rated adult population which is essentially stable and rating fluctuation is largely stochastic variation. You would need to redo your modelling applying the lower effecive-k to this poulation alone and then determine accuracy.

The effective k of the junior population would essentially increase slightly under my suggestions.

Its hard to imagine that a rating floor of 500 would be excessively inflationary in an environment where the junior ratings are essentially deflated by 100-200 points already. One couldn't say what the net effect would be until you've done the modelling KB.

Now you did say you would apply this and so far no real response is forthcoming.

PHAT
20-04-2004, 09:58 AM
I'm going with rapid.

Why? Because when you're just taking a single game you want the one that the stronger player generally is more likely to win (and let's ignore your imprecision and just agree to take this across a range of time limits, let's say the stronger player is stronger no matter what the time limit is).



Are you saying that a player ~200 points above his opponent will win ~70% of normal games but ~80% of rapids against him?


I'm going with rapid...

If you want to predict who is better from one game, the last thing you want is a draw. And the faster the time limit the less the risk of that.

It may be that the correct answer is even not "rapid" but "blitz" for that reason.

Didn't even think about that one, did you? :naughty:



Touche. The draw outcome did not figure in my cogitations. OK, I will have to add the caveate that the result information cannot include a draw. However, if the draw result was allowed, we could conclude that the two players were of about equal strength but that dosen't help us differentiate between tham.



Ian Rout has, I think, made excellent points for accepting longer games as better indicators - differential skills utility - I won't even try to improve on them.

Garvinator
20-04-2004, 10:39 AM
There are also many skills in fast chess that might be considered non-legitimate and are not (or should not be) transferrable at all, for example knocking pieces to the floor at a crucial time, putting pieces hanging over two squares and deciding later where it is, judging the right time to deliberately make an illegal or bad move that your opponnent will miss, keeping your finger on the button while the opponent is trying to press the clock, engaging in a constant stream of irritating banter and commentary with spectators, castling with a Rook from the next board etc.

and the opponent should stop the clock and call for the arbiter for all of these scenarios.

Garvinator
20-04-2004, 10:40 AM
about the cricket analogies. Its richie benaud here from the channel 9 commentary team. I am sorry cricket fans but the second test has been rained out and we are showing you the replay of the first test.

Ian Rout
20-04-2004, 10:43 AM
and the opponent should stop the clock and call for the arbiter for all of these scenarios.

Quite right. But without the cheating it wouldn't be fun.

Garvinator
20-04-2004, 10:47 AM
Quite right. But without the cheating it wouldn't be fun.
ahh so that is why bruce likes blitz and rapid, so he can cheat :owned: :owned: :whistle: :lol:

Ian Rout
20-04-2004, 11:03 AM
ahh so that is why bruce likes blitz and rapid, so he can cheat :owned: :owned: :whistle: :lol:
I'm sure Bruce genuinely believes that fast chess is "better" in some sense. Ultimately it's in the eye of the beholder and Bruce is entitled to that view. Personally I think it is like arguing that backyard cricket is better than Test cricket or ODIs, or that lawnmower racing is better than F1.

But having said that, I know of few things more tedious than Schumacher and co vrooming round and round and round, I'd rather have a barbecue in the park and watch a good lawnmower race.

Bill Gletsos
20-04-2004, 11:07 AM
Your own figures demonstrated that the lower Glicko dynamic yeilded virtually the same result as the ELO system.
Again an overgeneralisation by you not based on facts.
Although the difference above 2200 between the systems is much less than under standard Glicko, this does not remain valid the lower you go down the rating list. Of course I explained all this previously, you just choose to forget/ignore it.


If it worsens predictive accuracy I would imagine it would be marginal, and as you have suggested previously, given its a relatively stable population, in time the differences would merge. Of course it depends as well as to how throughly you performed your analysis. Perhaps you could demonstrate your method?

You really are an idiot.
I explained all this previously over numwerous posts/threads.
I have no intention wasting my time revisting it.




Are you seriously suggesting a rating of (say) 50 in a novice means anything at all? Its entirely meaningless, a complete abstraction serving absolutely no purpose whatsoever.
If the sample size is sufficent then the rating of 50 is just as valid as one of 500, 1000 etc.


Not to mention the discouraging effect it has on the child's morale.
Perhaps the problem is with organisers who decide to rate not only intershool A and B divisions but also C and even D divisions.


Bill, you've done reasonably well so far in controlling your tendancy to verbal abuse. I think everyone is tired of your tirades. Please continue to stick with the arguments.
If you dont want verbal tirades then dont make posts that make you look like a goose.

Bill Gletsos
20-04-2004, 11:18 AM
Lower effective-k would only apply to the higher rated adult population which is essentially stable and rating fluctuation is largely stochastic variation. You would need to redo your modelling applying the lower effecive-k to this poulation alone and then determine accuracy
What data do you have to back up this statement.
Answer none.


The effective k of the junior population would essentially increase slightly under my suggestions.
Unnecessary.
A players rating should be based on their results.
If they show a marked improvement Glikco2's volatility will handle it.


Its hard to imagine that a rating floor of 500 would be excessively inflationary in an environment where the junior ratings are essentially deflated by 100-200 points already.
Their is no mathematical justification for an arbitrary floor.
However having a floor at zero so as not to have negative ratings is reasonable.
After all all ratings are relative.

You keep arguing as though all juniors are essentially the same.
You cannot seem to get it thru your skull that they are not.
They all improve at different rates.
the only way to measure this improvement is via their results.

Bill Gletsos
20-04-2004, 11:25 AM
Are you saying that a player ~200 points above his opponent will win ~70% of normal games but ~80% of rapids against him?

I dont think he is saying that at all.
What he is saying is that generally higher rated players normal and rapid ratings are similar, whereas generally as the time limit gets faster the weaker player gets weaker at a faster rate. Hence player A is rated 1800 normal and maybe 1750 rapid whilst player B is rated 1600 normal but only 1350-1500 rapid.


Ian Rout has, I think, made excellent points for accepting longer games as better indicators - differential skills utility - I won't even try to improve on them.
I dont think he was saying longer are better indicators.
Different skills that are relevant in blitz and rapid are also relevant in normal as the end of time controls approach.

Bill Gletsos
20-04-2004, 11:29 AM
and the opponent should stop the clock and call for the arbiter for all of these scenarios.
Unfortunately in the heat of battle its very easy to overlook the dubious tactics of ones opponents.
Naturally if the player sees it, he should do as you suggest.

Bill Gletsos
20-04-2004, 11:30 AM
about the cricket analogies. Its richie benaud here from the channel 9 commentary team. I am sorry cricket fans but the second test has been rained out and we are showing you the replay of the first test.
:lol: :lol:

Rincewind
20-04-2004, 12:08 PM
ahh so that is why bruce likes blitz and rapid, so he can cheat :owned: :owned: :whistle: :lol:

I think Bruce plays most of his blitz on line where most of Ian's cheats (no offense intended) don't work. However, playing online does open up a whole other set of possibilities. ;)

I have a good idea, why don't we have an interstate competition? Try to find the best blitz cheat from each state and territory and play off for the national blitz cheating championship? I'm sure everyone could nominate a few of their state/territory's finest.

Anyone want to get the ball rolling?

Bill Gletsos
20-04-2004, 12:12 PM
I note the British Chess Federation run a normal and rapid rating system.
All events with time limits greater than G/60 are treated as normal.
There is no weighting differentiation between games of say G/61 and G/90 or G/120 or variations with Fischer increments.

The USCF rates anything between G/10 and G/29 in their rapid system.
They did rate anything greater than G/30 in their normal system.
They have recently however started to rate games of G/30 to G/60 in both rating systems.
There is no weighting differentiation between the games based on the time limits.

In fact as far as I can determine no current national rating system weights the games differently based on time limits.

PHAT
20-04-2004, 01:44 PM
I dont think he [IR] was saying longer are better indicators. Different skills that are relevant in blitz and rapid are also relevant in normal as the end of time controls approach.

Question: What game result is the better predictor for 25 minute rapid games; a 24 minute rapid game or a 26 minute?

Answer: Obviously the 26 minute because, it is based on 8% more play than the 24 mijnute game, for what is essentially the same game length.

From this example we can say that: As a general pinciple, longer chess games are better predictors than shorter chess games.

Bill, do you agree with me or not - if not, how/why.

Bill Gletsos
20-04-2004, 01:49 PM
Question: What game result is the better predictor for 25 minute rapid games; a 24 minute rapid game or a 26 minute?

Answer: Obviously the 26 minute because, it is based on 8% more play than the 24 mijnute game, for what is essentially the same game length.
Disagree. One is not better than the other.


From this example we can say that: As a general pinciple, longer chess games are better predictors than shorter chess games.

Bill, do you agree with me or not - if not, how/why.
Disagree.
Various skills are used in a game of chess at various stages.
To try and argue that a G/90 is worse than a G/120 has no basis in fact.

PHAT
20-04-2004, 02:09 PM
A players rating should be based on their results.


You say one thing and do another. Utterly duplicitous.

*We all got 70 pts for nix, and few years back 0 to 150 were added to ratings.
*Chunks of the population are getting bonus points on account of their high junior content.
*The top end of the list have the value of their results discounted.
*Sub-routines multipy the value of games played by "volitile" players.

You have been restant to all these measures - chanting the mantra, "A players rating should be based on their results." Yet eventually you implemented them when it was obvious to you that it should be done. Funny thing is, that you have always been the one of the last to see how obviously needed they were.

One day, you will be implementing some of the ideas that DR has been promoting. The irony is, that it will be you as the Ratings Officer, who receive the credit.

PHAT
20-04-2004, 02:13 PM
Various skills are used in a game of chess at various stages.
To try and argue that a G/90 is worse than a G/120 has no basis in fact.


Is this supposed to be an answer to why you think a 26 minute game is not better predictor than a 24 minute game for a future 25 minute game. Wake up to yourself and answer the f.....g question - unless you are too gutless.

arosar
20-04-2004, 02:17 PM
Look, I'll see if I can answer the question. I can see you're getting frustrated. What's the question?

AR

PHAT
20-04-2004, 02:26 PM
Look, I'll see if I can answer the question. I can see you're getting frustrated. What's the question?

AR

The orginal Q was:
Senario: You enter a guessing competition where you have to guess which player is best. The only information you can be given is the the result of either their last
bullit, or
blitz, or
rapid, or
90m+30s,

The result of which game do you choose to be told?
*(Added constraint: no draws were had)
*(The definition of "chess" here is, "A board game for two players, each beginning with 16 pieces of six kinds that are moved according to individual rules, with the objective of checkmating the opposing king.")





Question #2: What game result is the better predictor for 25 minute rapid games; a 24 minute rapid game or a 26 minute?

Answer: Obviously the 26 minute because, it is based on 8% more play than the 24 mijnute game, for what is essentially the same game length.

From this example we can say that: As a general pinciple, longer chess games are better predictors than shorter chess games.

Bill, do you agree with me or not - if not, how/why.

Alan Shore
20-04-2004, 02:34 PM
ahh so that is why bruce likes blitz and rapid, so he can cheat :owned: :owned: :whistle: :lol:

You'd better watch yourself Gray, I am not a cheat. But I am a swindler and proud of it. :owned:

Bill Gletsos
20-04-2004, 02:38 PM
You say one thing and do another. Utterly duplicitous.
If anyone is being duplicitous it is you. Your comments are totally misleading.


*We all got 70 pts for nix, and few years back 0 to 150 were added to ratings.
Adding points to everyone makes no real difference to the ratings because ratings are relative.

*Chunks of the population are getting bonus points on account of their high junior content.
Again misleading.
The ACT junior problem was well known. It had nothing to do with any inherent problem with Glicko. It was caused back in the ACF Elo days.
In fact even the change to the Gold Coast players (less than 105 in total) was due to problems originating back under Elo.


*The top end of the list have the value of their results discounted.
The players still get their rating points based on actual results, not some nebuleous factor. The change to the weighting is justified based on perviosuly documented reasons.


*Sub-routines multipy the value of games played by "volitile" players.
The use of a volatility factor is integral to the Glicko2 system.
Anyway, the resultant rating changes are due to actual results, not based on age or sex etc.


*You have been restant to all these measures - chanting the mantra, "A players rating should be based on their results." Yet eventually you implemented them when it was obvious to you that it should be done. Funny thing is, that you have always been the one of the last to see how obviously needed they were.
Again misleading, if not an outright falsehood.
There was no arm twisting by the ACT to fix their junior problem, we just did it. Likewise with any of the other changes.
In fact all I ever demanded of others was actual proof of a problem and proof that any supposed correction was actually justified.


*One day, you will be implementing some of the ideas that DR has been promoting. The irony is, that it will be you as the Ratings Officer, who receive the credit.
I doubt it.
In fact no matter what changes are implemented people will always complain.
Its just nowadays with the internet its easier.

Alan Shore
20-04-2004, 02:39 PM
But having said that, I know of few things more tedious than Schumacher and co vrooming round and round and round, I'd rather have a barbecue in the park and watch a good lawnmower race.

That's how I feel, despite your strange analogy :)

Bill Gletsos
20-04-2004, 02:39 PM
Is this supposed to be an answer to why you think a 26 minute game is not better predictor than a 24 minute game for a future 25 minute game. Wake up to yourself and answer the f.....g question - unless you are too gutless.
Get with the program Einstein.
I said it was immaterial whether you used 24 or 26. Both were as good as each other.

Bill Gletsos
20-04-2004, 02:43 PM
Look, I'll see if I can answer the question. I can see you're getting frustrated. What's the question?

AR
Both Barry and I have answered his questions.
He just doesnt like our answers because they dont fit in with his predetermined result.

Alan Shore
20-04-2004, 02:46 PM
I think Bruce plays most of his blitz on line where most of Ian's cheats (no offense intended) don't work. However, playing online does open up a whole other set of possibilities. ;)

I have a good idea, why don't we have an interstate competition? Try to find the best blitz cheat from each state and territory and play off for the national blitz cheating championship? I'm sure everyone could nominate a few of their state/territory's finest.

Anyone want to get the ball rolling?

I'm not sure how you'd want this run, Barry! Of course if it was run on the net it wouldn't be possible to cheat.. furthermore the best thing about the dodginess OTB is its Machiavellian nature - you have to be subtle and if it's recognised as a 'cheats champs' then opps would be expecting it :doh:

PHAT
20-04-2004, 03:11 PM
I said it was immaterial whether you used 24 or 26. Both were as good as each other.

Do tell, why are they the same, or, why aren't they different?

Bill Gletsos
20-04-2004, 03:13 PM
Do tell, why are they the same, or, why aren't they different?
26 mins offers no better information than 24 mins.

Kevin Bonham
20-04-2004, 03:18 PM
Are you saying that a player ~200 points above his opponent will win ~70% of normal games but ~80% of rapids against him?

Ignoring the draw issue - I would be surprised if the difference was that large but I would not be surprised if the better player did say one or two percent better. I also wouldn't be surprised if they didn't.


Touche. The draw outcome did not figure in my cogitations. OK, I will have to add the caveate that the result information cannot include a draw.

If you are ignoring the draws (say making them replay) you are making the stronger player unnaturally likely to win the longer the time limit. Suppose A scores 70:30 against B no matter what the time limit. In blitz that might be +65%=10%-25% for a win-loss ratio of 2.6. In 90/+30 it might be +50%=40%-10% for a win-loss ratio of 5. So in this case the longer game where you take the first win as evidence is more likely to give you the better player, but the cost is that you are waiting more games to do so. Blitz will give you the better player 72% of the time within 1.111 games on average while G90/+30 takes 1.63-odd games to get it right 83% of the time. You're not comparing like with like.

Trent Parker
20-04-2004, 03:22 PM
resurgence of sweeney v gletsos?

Lucena
20-04-2004, 03:23 PM
resurgence of sweeney v gletsos?
the claws come out again eh?

Kevin Bonham
20-04-2004, 03:25 PM
You couldn't possibly have the stupidity to think and say that I am showing "ignoble fear in the face of danger or pain."

Pain, not danger. Danger implies the possibility of damage to your position not occurring. Would you say that Bill is showing ignoble fear in the face of danger or pain from your arguments? I suggest the only danger he is in here is one of wasting his time and the only pain would be from laughter.


Sometimes Kevin you just blurt the first stupid crappy hackneyed come-back line that comes into your head and it falls way short of the mark. I almost, but not quite, feel insulted that you have made such scant effort to insult me.

Nice to see your old habit of accusing others of behaving like you do as soon as you are clearly losing the debate hasn't left you.

You have to admit, the material I had to work with there was less than inspirational.

Kevin Bonham
20-04-2004, 03:32 PM
Answer: Obviously the 26 minute because, it is based on 8% more play than the 24 mijnute game, for what is essentially the same game length.

You're assuming "more play" in the G26 provides a better predictor of a G25 than "play under more time pressure" in the G24. Along the line of my previous comments, I am not convinced this is necessarily so.

Also again you would need to cheat by ruling out draws.

Bill Gletsos
20-04-2004, 03:35 PM
resurgence of sweeney v gletsos?
When it comes to rating debates its more like Gletsos & Bonham & Cox V Sweeney & Richards.

PHAT
20-04-2004, 03:37 PM
26 mins offers no better information than 24 mins.

This is both intuatively wrong and wrong in logic. You can realy arse it up when you try to defend the indefensible.

A judgement - result W/L/D - must on average, be more accurite if more time is spent on its determination. Hence 26 predicts a 25 better than 24 predicts.

Show your logic to disprove it.

Bill Gletsos
20-04-2004, 03:53 PM
This is both intuatively wrong and wrong in logic.
I disagree.


You can realy arse it up when you try to defend the indefensible.
The only person making an arse of themselves is you mate.


A judgement - result W/L/D - must on average, be more accurite if more time is spent on its determination. Hence 26 predicts a 25 better than 24 predicts.

Show your logic to disprove it.
No I dont have to disprove it.
You are the one making the claim.
You have to prove it.

PHAT
20-04-2004, 04:18 PM
In blitz that might be +65%=10%-25% for a win-loss ratio of 2.6.
In 90/+30 it might be +50%=40%-10% for a win-loss ratio of 5.
So in this case the longer game where you take the first win as evidence is more likely to give you the better player, but the cost is that you are waiting more games to do so.
Blitz will give you the better player 72% of the time within 1.111 games on average while
G90/+30 takes 1.63-odd games to get it right 83% of the time.

You're not comparing like with like.

"...but the cost is that you are waiting more games to do so..."
"You're not comparing like with like."

Unfortunately, the drawn game problem cannot be avoided, as far as I can see. Therefore, there is no way around not comparing likewith exactly like. Nevertheless I like the senario you have designed - not just because it supports my position, but because it explores.

Such a pitty BG is unable to muster his thoughts to present a coherent case to support his position.

Bill Gletsos
20-04-2004, 04:23 PM
"...but the cost is that you are waiting more games to do so..."
"You're not comparing like with like."

Unfortunately, the drawn game problem cannot be avoided, as far as I can see. Therefore, there is no way around not comparing likewith exactly like. Nevertheless I like the senario you have designed - not just because it supports my position, but because it explores.
How does Kevins scenario support your position.

Such a pitty BG is unable to muster his thoughts to present a coherent case to support his position.[/QUOTE]
You havent presented a case at all.
All you have been doing is asking incomplete questions.

PHAT
20-04-2004, 04:29 PM
You're assuming "more play" in the G26 provides a better predictor of a G25 than "play under more time pressure" in the G24. Along the line of my previous comments, I am not convinced this is necessarily so.


Welllllllll, I am not 100% convinced I am right either. However, it comes down to, do you think the psychological effects of having 24 or 26 minutes is different enough to effect the out come. I think not. Certainly, there is a difference between G4 and G6, while there would be no measurable difference between G124 and G126.

I would like to bring all this back to the beginning. Should a G120 game be given a weighting above a G60?

Kevin Bonham
20-04-2004, 04:34 PM
Unfortunately, the drawn game problem cannot be avoided, as far as I can see. Therefore, there is no way around not comparing likewith exactly like.

The way "around" it is to stop fudging and admit that you are sunk here and that there is no way of turning your question based around one single game into something that provides a useful foothold for what you are trying to say on this thread.


Nevertheless I like the senario you have designed - not just because it supports my position, but because it explores.

How does it support your position?

Kevin Bonham
20-04-2004, 04:40 PM
Welllllllll, I am not 100% convinced I am right either. However, it comes down to, do you think the psychological effects of having 24 or 26 minutes is different enough to effect the out come. I think not. Certainly, there is a difference between G4 and G6, while there would be no measurable difference between G124 and G126.

One could say the same about the claim that 26 minutes provides significantly more thinking time than 24 and hence significantly more evidence of playing strength. If 26 is not significantly more than 24 in my context I don't see why it should be in yours.


I would like to bring all this back to the beginning. Should a G120 game be given a weighting above a G60?

I don't see why. You haven't answered my point about doing so potentially causing more conservative junior ratings either.

PHAT
20-04-2004, 04:41 PM
You havent presented a case at all.
All you have been doing is asking incomplete questions.

When you have some relevant comment, or insight, or novel idea to contribute, return to the discussion. Until then, shut-up. I am pretty sure I posted the following observation some time ago but it is still relevant - you should concider what the difference is between a technician and an investigator.

PHAT
20-04-2004, 04:57 PM
The way "around" it is to stop fudging and admit that you are sunk here and that there is no way of turning your question based around one single game into something that provides a useful foothold for what you are trying to say on this thread.


Sunk?!?!?! I am walking on water here.

The single game experiment provides a simplified model for examining the relationship (if any) between the length of a game and its preditive power.


HEY! I just thought, by replacing the single game with an google of games, the draw problem disappears - we need only look at W:D:L ratios.


How does it support your position?
It shows that for a finite number of games, and ignoring drawn games, longer games are more predictive. ie It is supportive under given conditions.

PHAT
20-04-2004, 05:02 PM
If 26 is not significantly more than 24 in my context I don't see why it should be in yours.


I don't think I said that.


You haven't answered my point about doing so potentially causing more conservative junior ratings either.

Please repeat it.

Cat
20-04-2004, 06:36 PM
What data do you have to back up this statement.
Answer none.



A players rating should be based on their results.
If they show a marked improvement Glikco2's volatility will handle it.


Their is no mathematical justification for an arbitrary floor.
However having a floor at zero so as not to have negative ratings is reasonable. If the sample size is sufficent then the rating of 50 is just as valid as one of 500, 1000 etc.


You keep arguing as though all juniors are essentially the same.
You cannot seem to get it thru your skull that they are not.
They all improve at different rates.
the only way to measure this improvement is via their results.

Again you miss the point entirely Bill. Your ignorance of this science is no excuse for persistent abusive behaviour. Nobody is suggesting all juniors are the same, what is being discussed are the characteristics of a population - their dynamics. Through observational enquiry one can analyse population dynamics to determine important trends within that population. My own experience in this area comes from my background in human genetics, a sub-field called 'population genetics' devoted to the study of gene expression in different human populations. I am well acquainted with the essential considerations in determining a mathematical discription of a given population dynamic.

Consider the following scenarios;

Group A is a cohort of 50 10 yr old children. The mean rating for that cohort is 1000.

Group B is a cohort of 50 adults, the mean rating for the cohort is 1500.

Now consider the mean ratings of these cohorts 5 years later. In both cohorts participation is maintained.

In group B, the mean rating is likely to be unchanged. In essence the Glicko dynamics predicts this as it is not inflationary. If the cohort was an entirely closed group then the mean would remain approximately unchanged. In any event the chances of a mean rating increase is equal to a mean rating decrease.

Now consider group A. Again the Glicko system is essentially predicting the mean rating will remain unchanged, as again if the cohort was a closed group the mean rating would change little or the chances of an increase would be the same as a decrease. The question arises is this realistic? Which is the more realistic scenario for the cohort, which is now 15yrs old;

1. The mean rating performance would still be 1000, despite training, playing, practising and 5 yrs of biological development.

2. The mean rating performance of the cohort would be (say) 750.

3. The mean rating performance would be (say) 1250.

Clearly the 3rd option is more likely than option 1, and much more likely than option 2. There is a developmental bias which will move the population rating mean forward, and this is bourne out by experience, observational data including my own analysis of juniors on the Gold Coast, where I have found the average rating change amongst active juniors is 50-60 points/ year.

Therefore we have 2 populations, cohort a & cohort b with different dynamic characteristics.

That juniors are being systematically being underrated in the rating system is clearly being voiced by Paul Sikes, who started this thread and Kerry Stead who says that underrated juniors are a 'fact of life'. 'A fact of like'- so endemic has acceptance of junior underrating become, players are coming to believe that such a situation is inevitable, that there are no solutions.

Its not enough to say that Glicko 2 will handle the problem, that the problems are a hang-over from ELO 3 years ago. What is required is acceptance that a problem exists, that a solution can be found, a realisation that underrated juniors are not 'a fact of life', and a determination to understand the root cause of the problem.

Bill Gletsos
20-04-2004, 06:48 PM
How does it support your position?
I'm pleased to see I'm not the only one asking this question.