PDA

View Full Version : ACF September 2004 Ratings



Bill Gletsos
30-08-2004, 11:50 PM
The ACF September 2004 rating lists have been sent to the webmaster and State Rating Officers.

This rating list incorporates the changes outlined in ACF Bulletin #276.
These changes resulted in a recalculation of all players ratings prior to the processing of the September 2004 ratings period. As such players who played no games in the September rating period may have different ratings to those previously published. This change may also effect players eligibility and position on the Top Improvers list as their previously published rating may have changed.

For the September 2004 rating period there were 137 Tournaments rated and 13717 games of which 8270 were in the ACF Classic rating system and 5447 were in the ACF Rapid rating system.


Top Players
2641!! 40 NSW Rogers, Ian [GM]
2508! 0 NSW Wallace, John-Paul [IM]
2494!! 11 VIC Johansen, Darryl K [GM]
2477!! 0 NSW Lane, Gary W [IM]
2448!! 26 QLD Solomon, Stephen J [IM]
2431!! 39 VIC Smerdon, David C [IM]
2400!! 12 NSW Zhao, Zong-Yuan [IM]
2400!! 14 SA Chapman, Mark [IM]
2399!! 36 VIC Froehlich, Peter [IM]
2395! 10 SA Tao, Trevor
2387!! 12 VIC Bjelobrk, Igor
2373! 0 VIC Teichmann, Erik [FM]
2368!! 29 VIC West, Guy [IM]
2353!! 0 VIC Depasquale, Chris J [FM]
2351!! 7 VIC Sandler, Leonid [IM]
2348! 10 VIC Jordan, Bill [FM]
2334!! 26 NSW Canfell, Gregory J [FM]
2328!! 52 NSW Xie, George
2317! 8 NSW Drummond, Matthew
2302!! 7 NSW Reilly, Tim [FM]
2296! 5 NSW Smirnov, Vladimir
2293! 10 VIC Baron, Michael [FM]
2283!! 36 VIC Rujevic, Mirko [IM]
2281!! 1 NSW Seberry, Ralph B
2278! 0 VIC Saw, Geoffrey [FM]
2275! 6 QLD Caoili, Arianne B [WFM]
2257! 7 NSW Scott, Ronald
2249!! 20 VIC Levi, Eddy L [FM]
2240! 7 QLD Stephson, David J
2237! 7 NSW Feldman, Vladimir [IM]
2232!! 5 NSW Tan, Justin
2231! 0 VIC Hamilton, Douglas G [FM]
2229! 0 QLD Sorokina, Anastasia [WIM]
2226!! 8 NSW Flatow, A (Fred) [FM]
2223!! 15 WA Boyd, Tristan
2219!! 21 SA Zaric, Srboljub
2211! 7 NSW Kabir, Ruhul
2205!! 0 VIC Rashid, Abdulwahab [FM]
2203!! 0 NSW Rej, Tomek
2198!! 5 NSW Tindall, Brett [FM]
2197! 7 NSW Berezina - Feldman, Irina [IM]
2194! 8 NSW Salter, Michael
2192!! 16 NSW Goris, Robert
2188! 5 NSW Hirschhorn, Jeremy K
2188! 0 WA Painter, Julian
2173! 0 VIC Chow, Samuel
2173! 9 VIC Anderson, Tim
2170!! 15 NSW Bird, Andrew
2161!! 15 WA Barber, Haydn J [FM]
2158! 0 VIC Partsi, Dimitry
2158!! 9 WA Byrne, Stewart J
2150! 0 VIC Aghamalyan, Armen
2149!! 9 VIC Hacche, David J
2147! 0 VIC Woodhams, Michael V
2146! 0 NSW Dwyer, Danny [FM]
2141!! 14 NSW Ayvazyan, Armen
2139!! 9 SA Cowley, Robert G (Bob)
2138! 0 VIC Booth, Stewart
2131!! 29 NSW Charles, Gareth
2126!! 20 NSW Agulto, Edgardo

Bill Gletsos
30-08-2004, 11:51 PM
Top Females
2275! 6 QLD Caoili, Arianne B [WFM]
2229! 0 QLD Sorokina, Anastasia [WIM]
2197! 7 NSW Berezina - Feldman, Irina [IM]
2067!! 9 NSW Eriksson, Ingela
2052! 0 SA Nguyen, Giang
1995!! 0 NSW Dekic, Biljana [WIM]
1985!! 0 NSW Moylan, Laura A [WIM]
1960! 0 NSW Lip, Catherine [WFM]
1927! 0 NSW Klimenko, Veronica [WFM]
1799!! 7 NSW Lane, Nancy L [WIM]
1796!! 20 VIC Szuveges, Narelle S [WFM]
1781!! 6 VIC Lee, Michelle
1779! 0 VIC Zivanovic, Andjelija
1774!! 6 NSW Huddleston, Heather
1762!! 43 NSW Song, Angela
1694!! 14 WA Payne, Sophie
1668! 0 NSW Shields, Sylvia
1650!! 41 ACT Oliver, Shannon
1642!! 22 WA Maris, Natalie A
1635!! 38 NSW Reid, Vaness

Bill Gletsos
30-08-2004, 11:51 PM
Top Under 20
2431!! 19 VIC Smerdon, David C
2400!! 17 NSW Zhao, Zong-Yuan
2328!! 18 NSW Xie, George
2232!! 18 NSW Tan, Justin
2223!! 19 WA Boyd, Tristan
2203!! 17 NSW Rej, Tomek
2173! 16 VIC Chow, Samuel
2100!! 15 NSW Yu, Ronald
2091! 18 NSW Tian, Kuan-Kuan
2075!! 16 ACT Jovanovic, Peter
2064! 18 SA Yang, Song
2052! 18 SA Nguyen, Giang
2044!! 19 QLD Humphrey, Jonathan
2037!! 14 ACT Wei, Michael
2031!! 18 VIC Nemeth, Janos
2028!! 17 VIC Bourmistrov, Denis
2015!! 17 NSW Zvedeniouk, Ilia
2006! 19 VIC Soloveychik, Sasha
1991!! 15 NSW Lubarsky, Kostia
1985!! 14 VIC Stojic, Dusan

Bill Gletsos
30-08-2004, 11:52 PM
Top Under 16
2100!! 15 NSW Yu, Ronald
2037!! 14 ACT Wei, Michael
1991!! 15 NSW Lubarsky, Kostia
1985!! 14 VIC Stojic, Dusan
1948!! 15 ACT Oliver, Gareth
1941!! 12 ACT Ikeda, Junta
1936!! 9 NSW Song, Raymond
1920!! 13 SA Obst, James
1897!! 15 NSW Suttor, Vincent
1895!! 15 VIC Jager, Jesse
1894!! 12 QLD Ly, Moulthun
1878!! 14 QLD Barnard, Casey T
1851!! 14 NSW Cronan, James
1849!! 13 VIC Lugo, Ruperto
1837!! 13 VIC Wallis, Christopher
1825!! 11 NSW Illingworth, Max
1803!! 15 NSW Morris, Michael
1798!! 13 QLD Lazarus, Benjamin
1796! 15 VIC Lindberg, Douglas
1781!! 13 VIC Lee, Michelle

Bill Gletsos
30-08-2004, 11:53 PM
Top Under 12
1936!! 9 NSW Song, Raymond
1825!! 11 NSW Illingworth, Max
1762!! 11 NSW Song, Angela
1723!! 10 VIC Lin, Zhigen Wilson
1576!! 11 NSW Harris, Benjamin
1563!! 11 QLD Anderson, Daniel C
1561!! 11 NSW Wu, Edwin
1552!! 9 VIC Morris, James
1548!! 10 NSW Miranda, Adrian
1506!! 10 NSW Xu, William
1496!! 11 WA Choong, Yita
1496! 11 QLD Xu, Mingda
1492!! 9 QLD Finke, Kelvin
1458!! 11 ACT Brown, Andrew
1320!! 11 VIC Yu, Derek
1298!! 10 VIC Dalton, Samuel
1295!! 11 ACT Ung, Thomas
1253! 10 VIC Tang, Jason
1246!! 10 QLD Grigg-Tondeleir, Samuel
1223!! 9 VIC Schon, Eugene

Bill Gletsos
30-08-2004, 11:53 PM
Top Females Under 20
2275! 17 QLD Caoili, Arianne B
2052! 18 SA Nguyen, Giang
1960! 17 NSW Lip, Catherine
1781!! 13 VIC Lee, Michelle
1779! 19 VIC Zivanovic, Andjelija
1774!! 14 NSW Huddleston, Heather
1762!! 11 NSW Song, Angela
1694!! 18 WA Payne, Sophie
1668! 19 NSW Shields, Sylvia
1650!! 17 ACT Oliver, Shannon
1635!! 14 NSW Reid, Vaness
1604!! 14 NSW Harris, Rebecca
1523!! 14 QLD Jule, Alexandra
1452! 16 VIC Richter, Sally-Anne
1412!! 13 QLD Evans, Amy L
1341!! 13 VIC Hickman, Casey
1331! 18 WA Cassidy, Kelly
1274!! 14 ACT Ikeda, Miona
1266! 14 ACT Eldridge-Smith, Veronique
1229! 14 QLD Lyons, Kieran C

Bill Gletsos
30-08-2004, 11:54 PM
Top Seniors
2231! 63 VIC Hamilton, Douglas G
2226!! 67 NSW Flatow, A (Fred)
2219!! 64 SA Zaric, Srboljub
2074!! 66 NSW Ghenzer, Charles
2062!! 77 NSW Viner, Phillip J
2010!! 61 NSW Capilitan, Romeo
1959! 70 TAS Pavicic, Mile
1959!! 66 NSW Benson, Lachlan
1924! 72 NSW Creech, Ken E
1908!! 76 WA Leonhardt, Wolfgang
1906! 68 NSW Hutchings, Frank P
1875! 70 NSW Jens, Henk W
1873!! 61 NSW Winter, George
1850! 63 NSW Pilja, Djuro
1845!! 71 NSW Lazaridis, Julius
1840!! 62 NSW Gross, Mike C
1835! 69 TAS Gibbs, Glen B
1833! 69 QLD Mehltreter, Otto
1832! 67 NSW Bautista, Elpidio
1805! 65 ACT Morton, Richard

Bill Gletsos
30-08-2004, 11:54 PM
Top Improvers
1837 185 VIC Wallis, Christopher
1401 180 NSW Lam, David
803 157 QLD Curtis, Shaun
990 153 QLD Pedersen-Lee, Lachlan
1002 150 ACT Guo, Emma
1313 141 VIC Kostrzewa, Jake
1816 118 NSW Huynh, Arthur
705 113 QLD Webb, Christina
1232 113 WA Slack-Smith, Blair
858 107 ACT Baines, Casey
1401 106 VIC Stewart, Clint
1864 104 QLD Kimura, Toshi
1638 103 NSW Tse, Jeffrey
1875 100 NSW Jens, Henk W
1561 99 NSW Wu, Edwin
1023 98 WA Arnold, Matthew
1965 97 QLD Napier, Darren A
948 97 VIC Schreenan, Michael
906 97 NSW Nguyen, Dominic
757 94 QLD Medland, Robert
1536 94 NSW Hutchinson, Jason
1523 94 QLD Jule, Alexandra
1924 93 QLD Wongwichit, Phachara
1503 91 NSW Boyce, Jamie
1387 90 VIC Neymanis, Eric
1941 90 ACT Ikeda, Junta
849 88 SA Eustace, Sophie
1229 87 QLD Gordon, Graham
1458 86 ACT Brown, Andrew
1619 84 NSW Roberts, Jamie

Bill Gletsos
30-08-2004, 11:55 PM
Busiest Players
1626 78 NSW Keuning, Anthony V
1458 66 ACT Brown, Andrew
1877 55 NSW Mendes da Costa, Alex
1834 54 ACT Grcic, Milan
1557 52 NSW Greenwood, Norman
2328 52 NSW Xie, George
1787 51 ACT Ali, Mosaddeque
1567 48 ACT Guo-Yuthok, Sherab
1002 48 ACT Guo, Emma
1936 47 NSW Song, Raymond
1548 44 NSW Miranda, Adrian
1762 43 NSW Song, Angela
1377 42 ACT Beltrami, Matthew
1172 42 ACT Chow, Justin
1915 42 VIC Voon, Richard
1650 41 ACT Oliver, Shannon
951 41 ACT Xing, Edward
1152 41 QLD Russell, Luthien
2641 40 NSW Rogers, Ian
1056 39 ACT Nguyen, Van
1409 39 VIC Renzies, Elliott
2431 39 VIC Smerdon, David C
1941 38 ACT Ikeda, Junta
2095 38 NSW Bolens, Johny
1635 38 NSW Reid, Vaness
1523 38 QLD Jule, Alexandra
1905 36 VIC Raine, Marcus
2283 36 VIC Rujevic, Mirko
719 36 VIC Van Dijk, Marieke
1164 36 VIC Cartledge, Dwayne
2399 36 VIC Froehlich, Peter
1328 36 QLD Mills, Ross
2020 35 VIC Pecori, Ascaro
1167 35 QLD Fisher, Jayden
1016 34 ACT Beltrami, Michael
1166 34 ACT Oliver, Tamzin L
1973 34 NSW Castor, David
1825 34 NSW Illingworth, Max
1831 34 VIC Belletty, Malcolm
1488 34 VIC Stokie, William

Bill Gletsos
31-08-2004, 12:00 AM
Normal Tournaments Rated
ACT
16/07/2004 44 9 2004 ACT Juniors
25/07/2004 86 7 2004 ANU Open
27/08/2004 10 9 2004 Belconnen First Division
27/08/2004 10 9 2004 Belconnen Premier
27/08/2004 10 9 2004 Belconnen Second Division
27/08/2004 10 9 2004 Belconnen Third Division
18/06/2004 50 3 2004 Ginninderra Cup Finals
27/08/2004 13 5 2004 Voldemort Goblet
29/08/2004 26 6 ACT U14 Championship 2004
16/06/2004 15 7 Cold Comfort Cup
28/05/2004 27 7 Ginninderra Cup qualifiers group 1
28/05/2004 27 7 Ginninderra Cup qualifiers group 2
16/07/2004 13 4 Swanson Sprint
26/07/2004 29 7 brindabella snows
NSW
11/07/2004 46 6 2004 Fairfield Winter Cup
04/08/2004 63 11 2004 NSW Grade Matches U1600
21/07/2004 48 9 2004 NSW Grade Matches U1800
22/07/2004 52 9 2004 NSW Grade Matches U2000
14/06/2004 52 7 2004 NSW Open Champs
14/06/2004 30 7 2004 NSW Open Under 1600
01/08/2004 50 7 2004 NSWCA August Weekender
29/06/2004 19 10 2004 WGONG COLLIES KO
20/06/2004 16 7 CCLCCC District Championship 2004
04/07/2004 8 7 CJS Purdy Div 2
04/07/2004 8 7 CJS Purdy Div 3
04/07/2004 4 6 CJS Purdy Div 4
20/05/2004 40 15 Central Coast InterClub Teams 2004
08/08/2004 29 7 Coal City Open 2004
29/06/2004 9 9 Coffs Harbour 1hr
27/06/2004 17 7 Common Man 2004
10/07/2004 7 7 Ettalong Winter Round Robin 2004
14/05/2004 4 3 Gosford April Gammas 2004
05/07/2004 10 9 Hakoah Cup 2004 Open
05/07/2004 22 9 Hakoah Cup 2004 Under 1700
20/06/2004 12 7 Mid North Coast Open
27/07/2004 29 10 NSW 2004 Grade Matches Open
27/07/2004 39 10 NSW 2004 Grade Matches U1400
16/07/2004 44 10 NSW JUNIOR CHAMPIONSHIP - UNDER 12
16/07/2004 39 9 NSW JUNIOR CHAMPIONSHIP - UNDER 18
04/07/2004 8 7 NSW Masters
02/05/2004 2 1 NSW Masters Miscellaneous
14/06/2004 8 7 NSW Womens Champs 2004
21/07/2004 9 9 Ryde Eastwood Prelim A
21/07/2004 9 9 Ryde Eastwood Prelim B
21/07/2004 9 9 Ryde Eastwood Prelim C
12/04/2004 32 7 Sydney Easter Cup
QLD
29/07/2004 22 5 2004 BCC Winter Swiss
04/07/2004 94 7 2004 Caloundra Open
14/06/2004 25 7 2004 Darling Downs Open
27/06/2004 35 7 2004 Gold Coast Open
27/06/2004 65 7 2004 Gold Coast Under 1600
03/06/2004 25 5 BCC Autumn Swiss 2004
05/05/2004 16 8 Bruce Harris Trophy 2004
25/07/2004 8 7 Gold Coast Women's and Girls 2004
17/05/2004 16 1 Gold Coast v Bullwinkle 170504
15/06/2004 21 8 Junior Masters 270404
01/08/2004 14 7 Mackay Open
08/08/2004 17 7 North Qld Open 2004
02/07/2004 26 8 Qld Under 14,16,18 2004
25/07/2004 6 5 Qld Vets and Disabled 2004
18/07/2004 60 7 Queensland Open 2004
15/03/2004 10 6 Ruth Coxhill Reserves 2004
15/06/2004 23 5 Townsville Swiss April
03/08/2004 24 7 Townsville Swiss June 2004
16/06/2004 25 7 Viv Greenelsh Memorial
SA
13/06/2004 14 6 2004 QUEEN'S BIRTHDAY
22/06/2004 99 9 2004 SACA Grade matches
16/07/2004 19 8 SA Junior Championships
11/07/2004 89 7 University Open 2004
TAS
01/06/2004 7 7 Burnie Club Champs A 2004
01/06/2004 7 7 Burnie Club Champs B 2004
28/07/2004 7 7 Robert Isted Memorial 2004
14/06/2004 24 6 Tasmanian Open
VIC
02/06/2004 18 6 Albury Rating Tournament
14/08/2004 12 11 Australian Masters
12/08/2004 14 13 Ballarat Club Championship
12/08/2004 7 16 Ballarat Club Championship Reserves
16/04/2004 100 9 Box Hill Club Championship
27/04/2004 10 9 Box Hill FIDE Rated Tournament (April)
04/05/2004 30 9 Box Hill Winter Swiss
28/06/2004 32 9 City of Melbourne Open
30/06/2004 50 7 Dandenong Autumn Open
10/07/2004 22 7 Elwood Round Robin
16/06/2004 25 7 Hobsons Bay Cup
25/04/2004 12 5 MCC Anzac Day Weekender
07/04/2004 6 10 MCC Raipalis Memorial (Group B)
05/05/2004 10 9 Melbourne Autumn Round Robin
05/04/2004 42 9 Melbourne Club Championship
04/04/2004 6 10 Preliminary Qualifier for Australian Masters
13/06/2004 97 7 Victorian Open
WA
07/06/2004 18 6 2004 Foundation Day Tournament
29/08/2004 46 9 2004 West Australian State Championship
15/06/2004 12 7 Fremantle Open
24/06/2004 50 7 Metro Open 2004
23/06/2004 12 6 Midland Club Championship

Bill Gletsos
31-08-2004, 12:01 AM
Rapid Tournaments Rated
ACT
13/07/2004 41 17 2004 ACT Junior Reserves
01/08/2004 47 9 2004 ACT Under 10 Championship
06/06/2004 18 5 2004 June Dev Squad
25/06/2004 24 8 2004 NJCC Term 2 Rapid
14/07/2004 15 7 Canberra Rapid
25/06/2004 25 7 Curtin Primary Term 2, 2004
24/06/2004 23 8 Souths Term 2 2004
10/07/2004 12 7 Street Chess 10 July 2004
12/06/2004 8 7 Street Chess 12 June 2004
19/06/2004 6 7 Street Chess 19 June 2004
26/06/2004 10 7 Street Chess 26 June 2004
03/07/2004 14 7 Street Chess 3 July 2004
09/08/2004 15 6 Tuggeranong Allegro
NSW
17/08/2004 21 9 2004 St. George Allegro
25/07/2004 24 7 Laurieton July One Day 2004
27/06/2004 26 7 Laurieton One Day Chess 27 June 2004
25/08/2004 62 9 Metropolitan Leagues Rapid Play 2004
15/08/2004 44 7 NSWCA Rapid 2004
QLD
11/06/2004 55 6 2004 CQ South High Schools Championships
01/08/2004 23 7 2004 Fastrain Under 10 Tournament
01/08/2004 40 7 2004 Fastrain Under 12 Tournament
01/08/2004 21 7 2004 Fastrain Under 14 Tournament
01/08/2004 22 7 2004 Fastrain Under 18 Tournament
01/08/2004 64 7 2004 Qld Girls Primary Interschools Teams Ch
01/08/2004 16 3 2004 Qld Girls Secondary Interschool Teams Ch
11/05/2004 108 8 GC Primary Rd 2/2004 North A
12/05/2004 133 8 GC Primary Rd 2/2004 South A
25/07/2004 25 13 Green and Gold Cup 2004
30/07/2004 39 6 IGS End of Season Tournament
08/04/2004 24 6 Northside Junior Championship 2004
02/07/2004 24 8 Qld Under 10 2004
02/07/2004 25 8 Qld Under 12 2004
29/06/2004 23 9 Qld Under 8 2004
29/05/2004 137 8 Surfers Paradise Premier Rapid 29 May 04
14/08/2004 57 9 Tweed City Junior Tournament 140804
SA
08/07/2004 6 10 Girls 1-day Tournament
08/07/2004 25 11 Open 1-day Tournament
16/07/2004 24 12 SA Under 12 Championships
06/07/2004 22 6 Winter Rapid 2004
VIC
06/07/2004 15 5 Box Hill Allegro (July)
15/08/2004 60 7 Rookies & Queens Cup (August)
18/07/2004 46 7 Rookies & Queens Cup (July)

Bill Gletsos
31-08-2004, 12:35 AM
Some regular BB posters ratings:

Ian Rogers - 2641!!
Greg Canfell - 2334!!
jeffrei - 2278!
Tristan Boyd - 2223!!
Gareth Charles - 2131!!
Ronald Yu - 2100!!
Brian Jones - 2056!!
bobby1972 - 2020!!
Kerry Stead - 2004!!
paulb - 2003
Ian Rout - 1918!!
Jason Chan - 1982!!
rob - 1975!!
Goughfather - 1973!! (2100! rapid)
Kevin Bonham - 1943!!
firegoat7 - 1919!
Shaun Press - 1837!
David Richards - 1792!Barry Cox - 1767!!
George Howard - 1715
Amiel Rosario - 1640!!
Candy Cane - 1635!!
starter - 1623!!
Bruce Dickinson - 1591! (1857 rapid)
antichrist - 1492!
Paul Sike - 1468!!
Scott Colliver - 1394
Trent Parker - 1311!!
Matt Sweeney - 1308!!
Garvin Gray - 1166!!

Trent Parker
31-08-2004, 01:22 AM
Thanks bill.
hmm a wee bit disappointin' for miself. Atleast its goin' in the right direction!

Kevin Bonham
31-08-2004, 04:45 AM
David Richards - 1972!
Kevin Bonham - 1943!!
firegoat7 - 1919!

If keeping this sort of company won't inspire me to recover the 96 ratings points I've chucked in my last 12 games, I'm not sure what will. :eek:

Amazes me the rate at which David keeps going up as an adult :clap: :clap: , and yet he complains about the system not responding to rapidly improving juniors.

Alan Shore
31-08-2004, 05:20 AM
Thanks for the info Bill.. surpising I went up that much despite drawing with little girls but it's heading back to where it once was at least after being cursed by Glicko before. But that's Glicko for you, what a great rating system.

David Richards > Kevin Bonham ... :owned:

hahahahahaha... that will keep me amused for the next three months :D

PHAT
31-08-2004, 06:51 AM
1308!! Up 55 for a PB. :) Going in the right direction - Was 1038 about three years ago. Now to crack 1400 :eek:

TP, 3 pts ain't gonna save you - UR so dead. :evil:

Rincewind
31-08-2004, 08:30 AM
Barry Cox - 1767!!

55 point drop but totally deserving after a couple of pretty ordinary results against substatially lower rated opposition. (Although not Matt this time, thank goodness).

I should be back up to the 1800 mark again in December though as I've been performing well in the current club tourny. ;)

Trent Parker
31-08-2004, 09:14 AM
1308!! Up 55 for a PB. :) Going in the right direction - Was 1038 about three years ago. Now to crack 1400 :eek:

I can tell u that at the fairfield summer cup in 2001 U were1070 (or thereabouts) and at the May w/ender u were 1078. (based on my achives :)




TP, 3 pts ain't gonna save you - UR so dead. :evil:
I'm underrated mate! .......underrated! :lol: :lol: :lol: :lol: :evil: :evil:

Bill Gletsos
31-08-2004, 10:07 AM
If keeping this sort of company won't inspire me to recover the 96 ratings points I've chucked in my last 12 games, I'm not sure what will. :eek:

Amazes me the rate at which David keeps going up as an adult :clap: :clap: , and yet he complains about the system not responding to rapidly improving juniors.
DR's rating was a typing error on my part. It should have been 1792!!.
I've corrected it in the other post.

Bill Gletsos
31-08-2004, 10:09 AM
David Richards > Kevin Bonham ... :owned:

hahahahahaha... that will keep me amused for the next three months :D
Actually it wont as I just mis-typed it. :doh:

pax
31-08-2004, 10:28 AM
I'm a little puzzled over the removal of points from inactive players. Is it the case that the system itself is deflating, therefore removal of points from inactive players is necessary to retain parity with active players?

You claim that Glicko will quickly determine the new rating of an inactive player. If this is indeed the case, the removal of points seems unnecessary, unless there is substantial deflation going on which is (clearly) not affecting inactive players.

The way I see it, inactive players who subsequently become active again will take a double hit. One they lose some arbitrary number of points determined by the ratings officer. Second, when they return to activity they will probably take a hit due to a temporary fall in performance and large RD. When they return to their true strength after (say) a couple of tournaments, their RD has probably gone down again and it takes much longer to get back.

By way of disclaimer, I am an inactive player, and have been affected by the rating change. I'm not particularly worried for my own sake, just curious as to the rationale. I do think the argument that inactive players will sit back deliberately to watch their rating go up is somewhat ludicrous.

The relevant section from Bulletin #276 is quoted below.


Clearly the aim of the rating system is to determine the relative strength of the players. However with players who have not participated for some time it is generally the case that they will decrease in strength. By continually providing players with these bonus points to attempt to keep the ACF and FIDE ratings in some sort of synchronisation, inactive players are reaching heights that would in reality be totally unattainable if the player was actually playing. Therefore players who may consider returning are not doing so because they do not wish to see their new "inflated" ratings drop. This is clearly not in the ACF's interests. We want ex players to participate, not sit back and just have their ratings increase through inactivity.

Therefore from a practical viewpoint, the implementation of the motion regarding the 150 points and the decision regarding the 70 points, it is hoped to entice some of these players back to being active players and again play competitive chess.

The statistical impact of deducting these points from these inactive players is minimal given we use the Glicko2 system which can quickly determine their true playing strength upon their return.

arosar
31-08-2004, 10:34 AM
Hey Bill, can you explain to me how my rapid rating is suddenly 1800+?? This is not fair man.

AR

Goughfather
31-08-2004, 10:43 AM
Hey Bill, can you explain to me how my rapid rating is suddenly 1800+?? This is not fair man.

AR

You played well in the Interleagues Rapid, did you not?

I think that my new Rapid rating is amusing, considering that it was 1608 just thirteen months ago. I couldn't even get into the top half of the draw for the Interleagues Rapid last year! It's increased 492 points in the meanwhile, so I guess, no more rating prizes for me.

Bill Gletsos
31-08-2004, 11:01 AM
Hey Bill, can you explain to me how my rapid rating is suddenly 1800+?? This is not fair man.

AR
You played too well. ;)

Trent Parker
31-08-2004, 11:12 AM
I imagine the RD for rapidplay would be much higher than in standard. Hence more volatility

Bill Gletsos
31-08-2004, 11:18 AM
I'm a little puzzled over the removal of points from inactive players. Is it the case that the system itself is deflating, therefore removal of points from inactive players is necessary to retain parity with active players?

You claim that Glicko will quickly determine the new rating of an inactive player. If this is indeed the case, the removal of points seems unnecessary, unless there is substantial deflation going on which is (clearly) not affecting inactive players.
There is no evidence of deflation and from a purely statistical viewpoint there is no real need to take the points off the inactive players.
However there are other reasons for removing the points as described in bulletin #276 and doing so has virtually no statistical impact.

Note that the only points taken from players were points we added to them in the first place. We did not deduct any actual rating points earned from tournament results from players.

If players are really inactive and they never return to playing then taking the points off them has no real effect. For those that eventually do return then Glicko2 will fairly quickly get them to their correct rating.


The way I see it, inactive players who subsequently become active again will take a double hit. One they lose some arbitrary number of points determined by the ratings officer. Second, when they return to activity they will probably take a hit due to a temporary fall in performance and large RD.
Actually that was far more likley to occur if they had been given the 150 and 70 point uplifts.


When they return to their true strength after (say) a couple of tournaments, their RD has probably gone down again and it takes much longer to get back.
The volatility factor of Glicko2 should handle it satisfactorily.


By way of disclaimer, I am an inactive player, and have been affected by the rating change. I'm not particularly worried for my own sake, just curious as to the rationale. I do think the argument that inactive players will sit back deliberately to watch their rating go up is somewhat ludicrous.
I have had it stated to me directly and also from other sources that some players are deliberately not playing for exactly that reason.

Bill Gletsos
31-08-2004, 11:20 AM
I imagine the RD for rapidplay would be much higher than in standard.
Only in so far as where the player is playing few rapid rated games.
For players who play many rapid games this would not be the case.

Paul S
31-08-2004, 11:25 AM
Had a cursory scan of the ratings on the ACF website (well done, Paul B for putting the new ratings up so quickly :) :clap: ) of players that I am familiar with. It seems to me that the average Rapid Rating is about 100 points higher than the average normal rating (perhaps a 100 point deflation needs to be applied to Rapid ratings to bring them into line with Normal ratings?). The rapid ratings seem to be less reliable than the Normal ratings, although this is probably due to (in NSW) considerably fewer Rapidplay comps having been rated (as opposed to Normal games).

Still, I like my Rapid Rating - its 202 points higher than my Normal rating (1670 versus 1468)! ;)

Trent Parker
31-08-2004, 11:30 AM
The rapid ratings seem to be less reliable than the Normal ratings, although this is probably due to (in NSW) considerably fewer Rapidplay comps having been rated (as opposed to Normal games).

This was my reasoning for my comment above.

pax
31-08-2004, 11:44 AM
There is no evidence of deflation and from a purely statistical viewpoint there is no real need to take the points off the inactive players.
However there are other reasons for removing the points as described in bulletin #276 and doing so has virtually no statistical impact.

Note that the only points taken from players were points we added to them in the first place. We did not deduct any actual rating points earned from tournament results from players.

That's a vacuous argument. If you add points to everybody *except* some subgroup, then you have deducted from that group relative to the whole system. The only value of any rating system is relative, so whether points are earned or added is irrelevant.




Actually that was far more likley to occur if they had been given the 150 and 70 point uplifts.

Actually, you're right - this is true. Glicko behaves quite anomalously when RD is high. Example: If I perform at 1700 for a rating period with high RD, I will end up with a higher rating if my initial rating was much lower than 1700 (say 1400) than if my initial rating was much higher (say 2000).

On this point can I ask another question. In the ACF implementation, is there a check to ensure that this doesn't happen? E.g If I have a rating of 2000??, and I perform at 1600 over 20 games it is clearly better to reinitialise my rating at 1600 than to use the Glicko calculation (which could be as bad as 1300). It's an extreme case clearly, but seems to me to be worth checking.



The volatility factor of Glicko2 should handle it satisfactorily.


Perhaps. I don't have the ability to carry out Glicko II calculations, so I wouldn't know for sure.



I have had it stated to me directly and also from other sources that some players are deliberately not playing for exactly that reason.

The number of such players must be exceedingly small. If they care more about their rating than whether they play chess or not, then their priorities are extremely skewed.

Rhubarb
31-08-2004, 12:09 PM
I'm a little puzzled over the removal of points from inactive players. Is it the case that the system itself is deflating, therefore removal of points from inactive players is necessary to retain parity with active players?

Hi Pax, when I first joined this BB six months ago, I was the single biggest critic of inactive players being given upwards adjustments totalling 220 points along with the active players, because of the (in my view) very obvious deflation in the system.

As I have come to realise after a long debate with Bill, however, this had little to do with the Glicko system and was due to inherent deflation in the old system, a hangover that continued to exist well after Glicko was implemented. A further complication was the well-known inflation of the FIDE system, and the obvious desire to retain some kind of meaningful parity with FIDE ratings now that more than 100 Australians are FIDE-rated. Further still, there came the refinement of the system, Glicko II, which was to be expected in a pioneering system.

At the time, Bill did not like my suggestion that inactive players should be stripped of upward adjustments, presumably because of the wish to maintain the integrity of relative differences due to known results.

In the latest list, however, Bill has made the practical, extremely difficult and highly laudable decision to remove 0-220 points from inactive players depending on how inactive they are, and back-calculated all results from the first adjustment 4 years ago.

I'm not sure how many inactive players, if any, will now erroneously think that they are underrated, but if they have genuine grievances they can just start playing again and the Glicko will be sure to put them at their true strength in double-quick time.

Disclaimer: All of the above is my own opinion amd I'm sure Bill can explain everything more accurately and in far greater detail.

Regards,
Greg

EDIT: My own fault for getting distracted and taking half an hour to post: the thread's moved on considerably since I started typing. :doh:

Bill Gletsos
31-08-2004, 12:11 PM
That's a vacuous argument. If you add points to everybody *except* some subgroup, then you have deducted from that group relative to the whole system. The only value of any rating system is relative, so whether points are earned or added is irrelevant.
I agree and diasagree with those comments at the same time. ;)
If the players never return then it makes no difference whatsoever. its the same as if they died and were removed from the list entirely.
For those that do return, then Glicko2 will quickly get them to their current strength. This would be true whether they had the additional points added or not. However if the system was Elo then you couldnt possibly do this as it would take far too long to get the player to his correct rating.



Actually, you're right - this is true. Glicko behaves quite anomalously when RD is high. Example: If I perform at 1700 for a rating period with high RD, I will end up with a higher rating if my initial rating was much lower than 1700 (say 1400) than if my initial rating was much higher (say 2000).
Not totally true. clearly this depends on the number of games played in the rating period.


On this point can I ask another question. In the ACF implementation, is there a check to ensure that this doesn't happen? E.g If I have a rating of 2000??, and I perform at 1600 over 20 games it is clearly better to reinitialise my rating at 1600 than to use the Glicko calculation (which could be as bad as 1300). It's an extreme case clearly, but seems to me to be worth checking.
Under the ACF implementation no players rating can drop in a rating period below their performance rating for the period and similarly no players rating can increase above their performance rating for the rating period.



Perhaps. I don't have the ability to carry out Glicko II calculations, so I wouldn't know for sure.
We have evidence of players with quite steady ratings over a number of periods with relatively low RD's suddenly increasing their rating in the next period by over 200 points.


The number of such players must be exceedingly small. If they care more about their rating than whether they play chess or not, then their priorities are extremely skewed.
I would agree that their priorites are skewed.

arosar
31-08-2004, 12:11 PM
FMD! Another ratings thread.

Starter mate, watch Bill's post count.

AR

Bill Gletsos
31-08-2004, 12:20 PM
At the time, Bill did not like my suggestion that inactive players should be stripped of upward adjustments, presumably because of the wish to maintain the integrity of relative differences due to known results.
In fact if the stripping of the points would have led in anyway to creating statistical inaccuracies we wouldnt have done it. If we still ran Elo it would definitely not have happened.


In the latest list, however, Bill has made the practical, extremely difficult and highly laudable decision to remove 0-220 points from inactive players depending on how inactive they are, and back-calculated all results from the first adjustment 4 years ago.
Yes it involved going back and determing activity on a year by year basis from old master files back to 1992.



EDIT: My own fault for getting distracted and taking half an hour to post: the thread's moved on considerably since I started typing. :doh:
That will teach you. ;)

Bill Gletsos
31-08-2004, 12:21 PM
FMD! Another ratings thread.

Starter mate, watch Bill's post count.

AR
Ha ha.
And of course you just had to post that to increase yours, didnt you. ;)

Rhubarb
31-08-2004, 01:00 PM
There is no evidence of deflation and from a purely statistical viewpoint there is no real need to take the points off the inactive players.

Bill, I don't understand why you're still so intransigent on this point. I'm not criticising the Glicko or your administration of the ratings. Are you just avoiding the question by a technicality by suggesting that it was compaction, not deflation, or are you just being politic on behalf of your predecessors?

Can you not just admit that there was either deflation or compaction in the rating system before you took over?

ursogr8
31-08-2004, 01:06 PM
FMD! Another ratings thread.

Starter mate, watch Bill's post count.

AR

Amiel
I know. I know.
But I can't help but watch. :rolleyes:
I have tried all sorts of 'give-up-the habit' tricks but none have overcome the obsession. In the end I just have to watch.


Fortunately Bill posts are never allowed to be part of the punters club's 'daily-differential'. There are strict criteria on posters who can be tracked as part of the punters club's 'daily-differential'. Bill's posts are not allowed, for a well publicised reason. :uhoh:

starter

Bill Gletsos
31-08-2004, 01:11 PM
Bill, I don't understand why you're still so intransigent on this point. I'm not criticising the Glicko or your administration of the ratings. Are you just avoiding the question by a technicality by suggesting that it was compaction, not deflation, or are you just being politic on behalf of your predecessors?
Ah Greg mate, we are misunderstanding each other.
My comments regarding deflation above and in response to pax were regarding the Glicko/Glicko2 system.

With regards, to the old Elo system, it was deflationary in some places, inflationary in others depending on various factors.

A couple of quick examples would be:
players rated at the bottom end of the list were deflating due to the 336 cutoff rule which meant if they didnt score 20% they lost rating points even if in reality they performed to their true performance rating. Likewise players at the very top of the list gained rating points if they scored in excess of 80%.
Also most players who gained bonus points simply overshot their rating. This resulted in them losing most of the bonus points in the following period.

Bill Gletsos
31-08-2004, 01:12 PM
Fortunately Bill posts are never allowed to be part of the punters club's 'daily-differential'. There are strict criteria on posters who can be tracked as part of the punters club's 'daily-differential'. Bill's posts are not allowed, for a well publicised reason. :uhoh:
I must have missed it.
What reason?

pax
31-08-2004, 01:18 PM
I agree and diasagree with those comments at the same time. ;)
If the players never return then it makes no difference whatsoever. its the same as if they died and were removed from the list entirely.

Agreed. If they don't play again, they don't matter.



For those that do return, then Glicko2 will quickly get them to their current strength. This would be true whether they had the additional points added or not. However if the system was Elo then you couldnt possibly do this as it would take far too long to get the player to his correct rating.


I do agree that Elo would be much worse in this respect. As for how quickly GlickoII recovers the rating, well I'll just have to use myself as a test case!



Under the ACF implementation no players rating can drop in a rating period below their performance rating for the period and similarly no players rating can increase above their performance rating for the rating period.


Glad to hear it.

Rhubarb
31-08-2004, 01:52 PM
Ah Greg mate, we are misunderstanding each other.

No I think we understand each other perfectly well. :)

Well done, Bill, and I hope the system now settles into a paragon of modern rating systems.

Oepty
31-08-2004, 02:06 PM
Thank you Bill for the ratings. Great to see them out early.
As to me personally I think my fall in rating was intirely deserved and it is about what I expected.
One little thing though, why is Caoili still listed as being a overseas player?
Scott

Bill Gletsos
31-08-2004, 02:16 PM
Thank you Bill for the ratings. Great to see them out early.
As to me personally I think my fall in rating was intirely deserved and it is about what I expected.
One little thing though, why is Caoili still listed as being a overseas player?
Scott
Thats simple.
I forgot to change her to QLD. :doh:

bobby1972
31-08-2004, 02:17 PM
another rating thread he he,love this system love it ,i cant believe it was nice to me again ,i can see clearly now :clap:

Ian Rout
31-08-2004, 02:19 PM
I think that independent of deflation or compaction it is a well-observed phenomenon (and not just in chess) that most players who don't play decline in strength. It's not universally true, they may be studying openings and endings or playing on the internet but on the balance of probability (and a rating is just a statistical exercise) an inactive player is weaker than the same player when they were playing. Whether it's tied to the 220 shift or whatever this has to be factored into the rating system somewhere.

If anything it's better to err on the side of understating inactive players' ratings. If they never play again it doesn't do any harm, unlike the damage of having people who don't play proudly sitting at the top of rating lists, and if they do play again the system will soon boost them back up.

Rincewind
31-08-2004, 03:39 PM
I can see that if the "better to understate" argument is used to liberally then this will be a deflationary effect without any balancing inflationary effect. So if implemented with no other doctoring of the system then the overall rating system would suffer a potential deflation.

However, if we got it right (if the rating reduction led to a more accurate rating, more often than not) than it could be a good thing.

At this stage we have idea what the function of rating decline vs period of inactivity looks like. Also players like Wohl and Speck who have extended periods of inactivity, when they are just inactive from the Aust scene, might take issue with their CF rating falling off when their FIDE rating isn't.

Bill Gletsos
31-08-2004, 03:49 PM
I can see that if the "better to understate" argument is used to liberally then this will be a deflationary effect without any balancing inflationary effect. So if implemented with no other doctoring of the system then the overall rating system would suffer a potential deflation.
I didnt take iot that Ian was suggesting we deduct points from active players, but that he was just supporting the idea of not giving them points.

The relevance of this is that in future in line with the ACF Council motion of last year we will be doing ACF/FIDe alignments every March list. Players who were inactive since the previous uplift will not receive any future uplift.


At this stage we have idea what the function of rating decline vs period of inactivity looks like. Also players like Wohl and Speck who have extended periods of inactivity, when they are just inactive from the Aust scene, might take issue with their CF rating falling off when their FIDE rating isn't.
I suspect Wohl is significantly overrated on the ACF list at 2561 compared to his FIDE of 2379 whilst Specks ACF rating is only 6 points behind his FIDE.

bobby1972
31-08-2004, 03:53 PM
hey bill how about another 150 freebies please he he

Bill Gletsos
31-08-2004, 03:59 PM
Just off the top of my head and without doing any sort of checking I would expect the next uplift to only be around 5-10 points.

So dont get too excited.

Garvinator
31-08-2004, 04:03 PM
Just off the top of my head and without doing any sort of checking I would expect the next uplift to only be around 5-10 points.

So dont get too excited.
so for someone sitting on 2149, that five points could be rather handy ;)

Bill Gletsos
31-08-2004, 04:06 PM
so for someone sitting on 2149, that five points could be rather handy ;)
Yes, but it will only be handy towards the end of 2005. :hand:

Rincewind
31-08-2004, 04:11 PM
I suspect Wohl is significantly overrated on the ACF list at 2561 compared to his FIDE of 2379 whilst Specks ACF rating is only 6 points behind his FIDE.

I was using Aleks and Nick as examples of people having periods of "inactivity"; not people who's ACF rating actually was lower than their FIDE rating.

Garvinator
31-08-2004, 04:12 PM
Yes, but it will only be handy towards the end of 2005. :hand:
i didnt say when it would be handy :P just said that it would be ;)

bobby1972
31-08-2004, 04:14 PM
hey barry have you updated your site ,its great to know how one is travelling between rating list ,

Bill Gletsos
31-08-2004, 04:16 PM
i didnt say when it would be handy :P just said that it would be ;)
The more interesting scenario would be if there turned out not to be an uplift but a down shift of 5-10 points. ;)

Garvinator
31-08-2004, 04:22 PM
The more interesting scenario would be if there turned out not to be an uplift but a down shift of 5-10 points. ;)
or if someone moved a motion and it was passed to increase from 2150 to 2200 :uhoh:

auriga
31-08-2004, 04:22 PM
Top Players
2641!! 40 NSW Rogers, Ian [GM]
2508! 0 NSW Wallace, John-Paul [IM]
2494!! 11 VIC Johansen, Darryl K [GM]
2477!! 0 NSW Lane, Gary W [IM]
2448!! 26 QLD Solomon, Stephen J [IM]


with the rating info such as this,
do you think it's useful for a column
with the change from the previous period.
ie. would show which of the top players are on the way up (or visa versa)

i guess this maybe hard depending on the software program, etc.

in fact, another useful column would be high/low
that showed the historical highest and lowest rating for that person.

i don't actually track this for myself but would be helpful to know
(ie. tell me my best ever rating)

Ian Rout
31-08-2004, 04:48 PM
I didnt take iot that Ian was suggesting we deduct points from active players, but that he was just supporting the idea of not giving them points.

I was suggesting (not an original thought) that there needs to be some mechanism to cope with the likelihood that an inactive player has declined, rather than leave them at their previous rating. In this case depriving inactives of some or all of the 220 adjustment has that effect.

The legal system says that it is better for ten rusty players to be over-rated than for one inactive but capable player to be under-rated, but I don't think this is a valid principle for the rating system.

Rincewind
31-08-2004, 05:25 PM
hey barry have you updated your site ,its great to know how one is travelling between rating list ,

Not yet. I usually try to get it update within the week of the new list being released. Tonight is a chess club night so expect it to be done sometime Wednesday at the earliest.

However, you can still use the site if you enter your ratings manually. It's just is less convenient to use it that way. ;)

(I've already done this to reassure myself that I've already reclaimed around 40 of the points I lost this period). :D

Bill Gletsos
31-08-2004, 06:43 PM
I have now corrected Caoili from OS to QLD so that she appears in the Top lists.

Recherché
31-08-2004, 07:16 PM
It says on the website that !! means a very reliable rating. I don't really understand how someone who has jumped 185 points (to take Christopher Wallace as an example) can be considered to have a reliable rating. Wouldn't it be assumed, in a situation like that, that the ratings system is still catching up to the rapid improvement of the player?

It seems like only those ratings which are (relatively) steady could really qualify as "very reliable", in terms of being accurate predictors of playing performance.

Unlike some others, I don't really have a problem with ratings being "innacurate" for quickly improving (or declining, though I would think that would be rarer) players, or those players "robbing" high rated players of their hard won ratings points - it's just the system at work. However I do find the reliability marks a bit confusing.

I assume the number of games submitted (and the number played in the past) has some bearing on the !! mark, what else has an effect? :)

Incidentally, my own rating went up 171 points, and I've got a !! as well. I'm not sure why I didn't make the improvers list. I've got the necessary number of past games, and I had a ! mark on my last rating. Perhaps my memory is in error with regards to the ! mark from June.

Bill Gletsos
31-08-2004, 07:38 PM
It says on the website that !! means a very reliable rating. I don't really understand how someone who has jumped 185 points (to take Christopher Wallace as an example) can be considered to have a reliable rating. Wouldn't it be assumed, in a situation like that, that the ratings system is still catching up to the rapid improvement of the player?
No.
The !! simply reflects his low RD. Wallis's rating is very reliable based on his many results over a wide rating range within the period.


It seems like only those ratings which are (relatively) steady could really qualify as "very reliable", in terms of being accurate predictors of playing performance.
Incorrect.
It all depends on the players RD which is reflected by the symbols.


Unlike some others, I don't really have a problem with ratings being "innacurate" for quickly improving (or declining, though I would think that would be rarer) players, or those players "robbing" high rated players of their hard won ratings points - it's just the system at work. However I do find the reliability marks a bit confusing.
Tha is because you are making the assumption that if his rating suddenly changed it cannot be reliable.
A players rating coukld be reliable and non volatile, reliable and volatile, unreliable and non volatile, unreliable and volatile amongst many other permutations.


I assume the number of games submitted (and the number played in the past) has some bearing on the !! mark, what else has an effect? :)
The ratings of your opponents and your results. :hmm:


Incidentally, my own rating went up 171 points, and I've got a !! as well. I'm not sure why I didn't make the improvers list. I've got the necessary number of past games, and I had a ! mark on my last rating. Perhaps my memory is in error with regards to the ! mark from June.
Since I have no idea who you are I cannot check. If you really want to know then PM me with your name and I'll check it out.

Bill Gletsos
31-08-2004, 07:40 PM
For those that may PM me asking for a list of their rating record for the past few periods, this isnt going to happen.

If you honestly believe there is an error, then I'm happy to check it out for you.
If you just want the record to look at then I'm sorry but I'm not about to do it otherwise everyman and his dog will request it.

arosar
31-08-2004, 08:00 PM
Bill, why don't just write an FAQ and post it everywhere?

AR

Bill Gletsos
31-08-2004, 08:36 PM
Bill, why don't just write an FAQ and post it everywhere?
That wouldnt solve the problem of those wanting their ratings record.

Bill Gletsos
31-08-2004, 09:11 PM
A minor error has been found in the NSW U1400 Grade matches.
A corrected set of normal reports has been sent to the State Rating Officer's and the webmaster. Also Caoili is now listed as QLD.
The affected listings at the front of this thread have been corrected.

Cat
31-08-2004, 10:14 PM
There have been so many corrections to the rating system in recent times it's been hard to know what real effect it's been having on the fundamental dynamics, certainly at our Gold Coast Club.

This quarter the changes have been fairly minimal, so I thought I'd take the opportunity to assess the changes over the last quarter.

Firstly, I wondered what effect Bill's latest correction would have on the ratings. I counted 37 players who were inactive over the last quarter and they recieved an average rating increase of about 18pts. This correction was retrospective, returning to the start of the Glicko period, so is a one-off correction.

Next I looked at the juniors. I counted 48 juniors, who played an average of 17 games in the quarter recieved an average rating incease of 36 pts or about 2pts/game. However, given that the average correction was around 18pts for the quarter, the increase from the Glicko system was (36-18) =18,
or 1.05pts/game.

Then I looked at the adults. I counted 28 adults, who played an average of 15 games in the quarter recieved an average rating decrease of 12 pts or about just under 1pt/game. However, given that the average correction was around 18pts for the quarter, the decrease from the Glicko system was (12 +18) =30,
or 2pts/game.

So in other words in regular competition on the Gold Coast, a junior will on average gain 1 rating point/game and the average adult will loose 2 rating points /game under the Glicko 2 system, at least for the last quarter.

In other words, the Glicko system is describing a population where the adults are declining twice as quickly as the juniors are improving.

I'm contacting the council tomorrow, I seriously suspect the lack flouride in our water is having an effect on our adult population which has hitherto been unrecognised. Alternatively, maybe BSE has arrived on the Gold Coast, this could indeed be a major public health issue.

Of course one conclusion could be that despite all the window-dressing, bluster and posturing, the Glicko system is still failing to come to grips with the pooling problems I have described previously. Naa, couldn't be, it must be the BSE!

Bill Gletsos
01-09-2004, 12:11 AM
There have been so many corrections to the rating system in recent times it's been hard to know what real effect it's been having on the fundamental dynamics, certainly at our Gold Coast Club.

This quarter the changes have been fairly minimal, so I thought I'd take the opportunity to assess the changes over the last quarter.

Firstly, I wondered what effect Bill's latest correction would have on the ratings. I counted 37 players who were inactive over the last quarter and they recieved an average rating increase of about 18pts. This correction was retrospective, returning to the start of the Glicko period, so is a one-off correction.
They received no increase at all.
The differences in ratings are simply due to the change in rating method due to the introduction of the changes described in ACF Bulletin #276.
It certainly doesnt amount to a one off correction in any way shape or form.
As such the less said aboutyour 18 point average the better. Its total crap.


Next I looked at the juniors. I counted 48 juniors, who played an average of 17 games in the quarter recieved an average rating incease of 36 pts or about 2pts/game. However, given that the average correction was around 18pts for the quarter, the increase from the Glicko system was (36-18) =18,
or 1.05pts/game.

Then I looked at the adults. I counted 28 adults, who played an average of 15 games in the quarter recieved an average rating decrease of 12 pts or about just under 1pt/game. However, given that the average correction was around 18pts for the quarter, the decrease from the Glicko system was (12 +18) =30,
or 2pts/game.

So in other words in regular competition on the Gold Coast, a junior will on average gain 1 rating point/game and the average adult will loose 2 rating points /game under the Glicko 2 system, at least for the last quarter.
Given that the June ratings were recalculated from scratch from the December 2000 period thru to the June 2004 period, you cannot use the previously published rating figures for June 2004 to make any comparision to the Sept 2004 figures.
As for the use of avergae figures this is total crap.
A simple example would be two juniors. One not improving and one improving significantly.
The non improving junior plays 15 games and his rating change is an increase of 5 points. The other junior also played 15 games but improved by 145 points. The net effect is that the average increase per junior was 75 points. The average increase per game was 5 points.
The net value of this information?
100% worthless crap.
Just like your data above, except yours is even more worthless since its based on mixing ratings from 2 different calculation methods.



In other words, the Glicko system is describing a population where the adults are declining twice as quickly as the juniors are improving.
Given your data above is flawed then you have no means of determining if this is true or not.


I'm contacting the council tomorrow, I seriously suspect the lack flouride in our water is having an effect on our adult population which has hitherto been unrecognised. Alternatively, maybe BSE has arrived on the Gold Coast, this could indeed be a major public health issue.
Stick to medicine doc because ou sure as hell know crap about ratings.
And basing your figures on a mixture of current sept 2004 and outdated June 2004 figures just shows how poor your stats skill is.
Of course your calculations back in April 2003 were based on equally useless data when you used a mixture of Elo, Glicko and QLD junior ratings along with incorrect performance ratings to make a number of claims.


Of course one conclusion could be that despite all the window-dressing, bluster and posturing, the Glicko system is still failing to come to grips with the pooling problems I have described previously. Naa, couldn't be, it must be the BSE!
Your a waste of time.
Stick to medicine.

Bill Gletsos
01-09-2004, 12:27 AM
You know DR you never cease to amaze me. :rolleyes:

I cannot believe you could be so utterly stupid as to attempt a comparison of the Sept 2004 ratings with the previously published June 2004 figures. :whatthe:

Given my comments on the ACF ratings page and also in post #1 of this thread where I said:
"These changes resulted in a recalculation of all players ratings prior to the processing of the September 2004 ratings period".
it should be blatantly obvious that any comparison isnt worth squat.

In fact you should be embarrassed to have even have given it a seconds thought.

You really are a fool and a complete waste of time. :hand:

PHAT
01-09-2004, 01:12 AM
I cannot believe you could be so utterly stupid as to attempt a comparison of the Sept 2004 ratings with the previously published June 2004 figures. :whatthe:

Given my comments on the ACF ratings page and also in post #1 of this thread where I said:
"These changes resulted in a recalculation of all players ratings prior to the processing of the September 2004 ratings period".
it should be blatantly obvious that any comparison isnt worth squat.


Unfortunately, know one can checkup on anything to do with inflation/deflation/compaction/pooling/"speciation"/trend analysis/volatility et cetera because you are the only phkr with access to all the orginal numbers. Who knows what you do with the, additions/subtractions/dampening factors/volatility factors/targeted adjustments et cetera. Untouchable in your fortress of secracy and piecemeal explanations, you rule the ratings like a dictator.

Is it any wonder so many of the rank and file just give up trying to understand how/why their rating went up/down so little/much. They do not trust the system so they must simply have faith, or FO.

Bill Gletsos
01-09-2004, 01:28 AM
Unfortunately, know one can checkup on anything to do with inflation/deflation/compaction/pooling/"speciation"/trend analysis/volatility et cetera because you are the only phkr with access to all the orginal numbers. Who knows what you do with the, additions/subtractions/dampening factors/volatility factors/targeted adjustments et cetera. Untouchable in your fortress of secracy and piecemeal explanations, you rule the ratings like a dictator.
You are as big a fool as your tag team partner DR.
No one but a clown like you would ever accuse me of piecemeal explanations.


Is it any wonder so many of the rank and file just give up trying to understand how/why their rating went up/down so little/much. They do not trust the system so they must simply have faith, or FO.
The majority of players never really understood the Elo system, so the current situation is no different.

Kevin Bonham
01-09-2004, 03:45 AM
DR's rating was a typing error on my part. It should have been 1792!!.

*breathes sigh of relief*

I was actually going to ask Dave since he was such a strong player now, why he hardly ever posted anything related to chess.

Not that 1792 is by any means bad.

Cat
01-09-2004, 08:31 AM
*breathes sigh of relief*

I was actually going to ask Dave since he was such a strong player now, why he hardly ever posted anything related to chess.

Not that 1792 is by any means bad.

Rest easy KB, I think I'm pretty close to plateau now!

Cat
01-09-2004, 08:53 AM
You know DR you never cease to amaze me. :rolleyes:

I cannot believe you could be so utterly stupid as to attempt a comparison of the Sept 2004 ratings with the previously published June 2004 figures. :whatthe:

Given my comments on the ACF ratings page and also in post #1 of this thread where I said:
"These changes resulted in a recalculation of all players ratings prior to the processing of the September 2004 ratings period".
it should be blatantly obvious that any comparison isnt worth squat.

In fact you should be embarrassed to have even have given it a seconds thought.

You really are a fool and a complete waste of time. :hand:

My comparisons never do impress you, nevertheless, there is a discernable pattern re-emerging. As Matt says, precise estimations are difficult, but the figures I have presented I have no doubt are close to ball park. Already some of the adults like Bruce Harris, Fred Rosenow, Bernie Saavedra & Nell van der Graff are almost back to their pre-uplift ratings 18 months ago.

Since Smerdon, Queensland has only had 1 player break through 2000 (Jonathon Humphreys) and he had to win the Australian Championships to do it. Justin Pengelly had a 2000 rating, but dropped to 1700 when he started playing on the Gold Coast, and has now clawed back up to around 1950, thanks largely to the uplifts. We have 3 players, Toshi, Matt & Ptachara, 2 of whom hold 2000+ FIDE ratings, who have moved up & down in the 1700-1950 range over the last few years, despite fantastic performances. The periodic form slumps, which all players experience, have held their ratings back its true, but that's not the whole story. There are a number of other 2000+ FIDE rated players in Queensland who a trapped in the 1700-1950 rating bracket.

The Gold Coast cannot afford to see the adult ratings decimated again. This would lead to disaffection amongst the adults, and stagnation for the juniors. It's hard to sell this product.

If this is indeed a trend, and history tells us this has happened before, then within 12-18 months all the good work you have put in will have evaporated. I would suggest introducing floorings to the Gold Coast, maybe Queensland-wide, until the trend can be properly analysed and if indeed this trend does continue, a lasting solution found. If we sit on our hands, do nothing and wait, then the damage will already have been done.

Recherché
01-09-2004, 09:53 AM
No.
The !! simply reflects his low RD...
It all depends on the players RD...

I may be missing something obvious, but I don't know what RD is. :)


Since I have no idea who you are I cannot check. If you really want to know then PM me with your name and I'll check it out.

Done.

alexmdc
01-09-2004, 10:20 AM
For those that may PM me asking for a list of their rating record for the past few periods, this isnt going to happen.

If you honestly believe there is an error, then I'm happy to check it out for you.
If you just want the record to look at then I'm sorry but I'm not about to do it otherwise everyman and his dog will request it.

Why not? Surely this should be available somehow to look at.

I personally want to see how my previous ratings look after they were recalculated. Again.

arosar
01-09-2004, 10:22 AM
Why not?

Cos your membership fee doesn't cover this level of personalised service mate. Besides, he's a volunteer, if I may say so meself.

AR

Rincewind
01-09-2004, 10:26 AM
Alex,

You can estimate it by working from your June Rating and calculating your September rating based on your results in the published list of events. Then compare this to your published September rating. The difference is likely to be in the ballpark of your rating adjustment. However, this is very rough as there is no way to readily model everyone else's rating also moving around.

alexmdc
01-09-2004, 10:33 AM
Cos your membership fee doesn't cover this level of personalised service mate. Besides, he's a volunteer, if I may say so meself.

AR

I don't see why it can't be published for each active player over the rating period. This shouldn't be too much more work, since it's all automatically generated anyway.

Then those players who don't understand their rating changes would have more info to look at before they complain.

Rincewind
01-09-2004, 10:41 AM
I don't see why it can't be published for each active player over the rating period. This shouldn't be too much more work, since it's all automatically generated anyway.

Then those players who don't understand their rating changes would have more info to look at before they complain.

I suspect the ratings officers probably feel that publishing two sets of ratings for the same period will cause more confusion and lead to more work rather than less.

Alan Shore
01-09-2004, 10:43 AM
Yeah, the 'Best' rating would be novel but I guess it should be up to you to remember what it was. If it was there on the page I could say 'Look! I used to be good! I'm still 140 pts in arrears..

Recherché
01-09-2004, 10:55 AM
It's really not that difficult to keep your own records of your rating progression. Especially if you're keeping records of your games, which should contain your rating if you fill your sheets/database records out properly.

Bill Gletsos
01-09-2004, 11:11 AM
My comparisons never do impress you, nevertheless, there is a discernable pattern re-emerging.
Based on dodgy data.
Your usual modus operandi.


As Matt says, precise estimations are difficult, but the figures I have presented I have no doubt are close to ball park. Already some of the adults like Bruce Harris, Fred Rosenow, Bernie Saavedra & Nell van der Graff are almost back to their pre-uplift ratings 18 months ago.

Since Smerdon, Queensland has only had 1 player break through 2000 (Jonathon Humphreys) and he had to win the Australian Championships to do it. Justin Pengelly had a 2000 rating, but dropped to 1700 when he started playing on the Gold Coast, and has now clawed back up to around 1950, thanks largely to the uplifts. We have 3 players, Toshi, Matt & Ptachara, 2 of whom hold 2000+ FIDE ratings, who have moved up & down in the 1700-1950 range over the last few years, despite fantastic performances.
Thats just subjective belief.


The periodic form slumps, which all players experience, have held their ratings back its true, but that's not the whole story. There are a number of other 2000+ FIDE rated players in Queensland who a trapped in the 1700-1950 rating bracket.
Most players rated under 2100 on the FIDE list have bogus FIDE ratings so using them as an example of 2000+ strength is just plain stupid.


The Gold Coast cannot afford to see the adult ratings decimated again. This would lead to disaffection amongst the adults, and stagnation for the juniors. It's hard to sell this product.

If this is indeed a trend, and history tells us this has happened before, then within 12-18 months all the good work you have put in will have evaporated.

Actually history tells us no such thing.
When we investigated the GC situation last year there was no indication that the GC players were suffering from pooling problems.
When recalculating GC players ratings with only players outside the GC area there is no indication there was a problem. Some would have had higher ratings some would have had lower ratings. Of course where you have to be extremely careful with this scenario is that you are ignoring valid wins/losses from within the GC player group.

If we had been totally statistically correct we would have deducted points from a significant number of GC players. As it was all we did was artifically inflate the GC pool by not reducing players ratings below their pre-correction ratings.


I would suggest introducing floorings to the Gold Coast, maybe Queensland-wide, until the trend can be properly analysed and if indeed this trend does continue, a lasting solution found. If we sit on our hands, do nothing and wait, then the damage will already have been done.
This isnt going to happen.

bobby1972
01-09-2004, 11:18 AM
i love rating time.it is very very funny,a friend of mine who has retired from turny playing and had a rating that would get him into the oz ch,has been devastated by the 150 points they took of him.its like very bad imagine to look at your fat rating in the master list and then from nowhere bang your are well whats the word,150 down thats a lot.

Garvinator
01-09-2004, 12:44 PM
David,

Not only is your rating method completey flawed, but to give it any chance of being relevant to gc players, you would also have to take out every game the gc players has played against players not from the gc. You have not done this so your method has even less credibility than before :whistle:

Btw, what are you trying to achieve with this David?

auriga
01-09-2004, 01:19 PM
It's really not that difficult to keep your own records of your rating progression. Especially if you're keeping records of your games, which should contain your rating if you fill your sheets/database records out properly.

we could, but we're in the information age now,
so we shouldn't have to.
the data would be in the acf ratings database somewhere.
it could be exported and people could download.

ratings like games (ozbase games database) should
be kept and archived for prosperity.
of course, the more transparent the system is
the more scrutiny (and critism) it could attract.

Cat
01-09-2004, 02:18 PM
David,


Btw, what are you trying to achieve with this David?

Bill is saying our problem is that we don't play enough rated games, that if we'd play more rated games then the ratings would be more accurate. I've never accepted this to be the cause of our rating problems, but even if we accept this, we are clearly being posed a dilemma.

To rate all these games costs money, and resources are invariably limited. Bill is asking us to pour money into his rating product, telling us that our satisfaction with the rating system will improve because it will better represent junior improvement. All the evidence I have suggests that it will lead to adult rating deflation and that in 12-18 months we'll be back where we started.

What I'm asking Bill for is a warranty with his product against adult rating deflation, a guarantee that with his product the adult rating pool will maintain stability over the next 12 months and that deflation will be corrected on an ongoing basis.

Now I don't think this is too much to ask for. If I purchase any product in Australia, I have an expectation of a guarantee and knowledge I have protection from the consumer watchdog. Why is purchasing this rating product any different?

What I am interested is maintaining participation at the Gold Coast Chess Club. If we use a product that disaffects our consumers, then that is bad for our business. All I am asking for is some support to implement the ACF brand. Sell it to me, Bill!

Garvinator
01-09-2004, 02:44 PM
if this was a real concern for most members on the gc, why hasnt Graeme been the one taking up the fight relating to these matters. Graeme is the one with the most to lose with dissatisfaction.

Cat
01-09-2004, 03:04 PM
if this was a real concern for most members on the gc, why hasnt Graeme been the one taking up the fight relating to these matters. Graeme is the one with the most to lose with dissatisfaction.

This is a Gold Coast Chess Club matter- the Gold Coast Chess Club pays rating fees. I was Vice-President of the club when I raised the issue of ratings concerns 18/12 ago. It was agreed I would investigate and attempt to act as advocate to try to solve some of these issues. I continue to be concerned that problems remain and I guess this is something that will need to be brought up at the next club meeting, how we resolve the issue. I guess some kind of warranty would make the issue easier to sell.

Bill Gletsos
01-09-2004, 04:22 PM
Bill is saying our problem is that we don't play enough rated games, that if we'd play more rated games then the ratings would be more accurate. I've never accepted this to be the cause of our rating problems, but even if we accept this, we are clearly being posed a dilemma.
I dont care what you do or dont accept.
You were arguing for more accurate junior ratings in the normal system.
My answer was have your juniors play more normal rated games.
If you dont want to thats your problem not mine.
At least its clear from ChessMum's post in the Planned Rating Changes thread that not everyone is as silly as you in this regard.


To rate all these games costs money, and resources are invariably limited. Bill is asking us to pour money into his rating product, telling us that our satisfaction with the rating system will improve because it will better represent junior improvement. All the evidence I have suggests that it will lead to adult rating deflation and that in 12-18 months we'll be back where we started.
Ah so now your going to argue that having more junior rated games will lead to the adult ratings deflating.
If the juniors end up with more accurate ratings then the adults playing them should have more accurate results, because they will be playing juniors with more accurate ratings.
As for your so called evidence you have no evidence. You have a belief.
I have a belief you are a fool. Based on you views on the Does God Exist thread then my belief is just as valid as yours. In fact from my perspective my belief is significantly more valid than yours. :owned:



What I'm asking Bill for is a warranty with his product against adult rating deflation, a guarantee that with his product the adult rating pool will maintain stability over the next 12 months and that deflation will be corrected on an ongoing basis.

Now I don't think this is too much to ask for. If I purchase any product in Australia, I have an expectation of a guarantee and knowledge I have protection from the consumer watchdog. Why is purchasing this rating product any different?
You arent purchasing a product at all.
The ACF runs a rating system. If you want your games rated you have to pay an admin fee.
Note it isnt a rating fee. Its an administrative fee that just happens to be tied to game submitted for rating.


What I am interested is maintaining participation at the Gold Coast Chess Club. If we use a product that disaffects our consumers, then that is bad for our business. All I am asking for is some support to implement the ACF brand. Sell it to me, Bill!
Although I have absolutely no faith in your ability to diagnose any issues with the rating system, that does not mean Graham Saint and I wouldnt make any changes we felt were necessary, provided such changes are warranted based on our investigations.
Thats about the best guarantee you are likely to get from us. :hand:

Bill Gletsos
01-09-2004, 04:29 PM
This is a Gold Coast Chess Club matter- the Gold Coast Chess Club pays rating fees. I was Vice-President of the club when I raised the issue of ratings concerns 18/12 ago. It was agreed I would investigate and attempt to act as advocate to try to solve some of these issues.
Gold Coast is just fortunate that I dealt with Sue Kimura who is a much more reasonable person to deal with than you.


I continue to be concerned that problems remain and I guess this is something that will need to be brought up at the next club meeting, how we resolve the issue.
My concern would be that the others dont really see it as an issue but would simply let you carry on with it just to keep you out of their hair. :rolleyes:
In which case instead of wasting their time you are wasting mine. :wall:

Bill Gletsos
01-09-2004, 04:35 PM
I suspect the ratings officers probably feel that publishing two sets of ratings for the same period will cause more confusion and lead to more work rather than less.
He didnt just ask for that.
He wanted me to provide him with a complete rating history including all his tournments and results.
Given he has no means to determine if the calculations are correct then it is pretty much a waste of my time. I have better things to do like arguing with DR and Matt here on the BB. :lol:
As for him just wondering what his recalculated ratings may have been due to the implementation of the changes described in ACF Bulletin #276 it really serves him no purpose other than curisoity value.

bobby1972
01-09-2004, 04:37 PM
i would like to take this opportunity to say how much i love gliko system specially at rating time

bobby1972
01-09-2004, 04:47 PM
question for bill,could you tell us did all inactive(and for how long) players loose 150 points

Cat
01-09-2004, 05:11 PM
Although I have absolutely no faith in your ability to diagnose any issues with the rating system, that does not mean Graham Saint and I wouldnt make any changes we felt were necessary, provided such changes are warranted based on our investigations.
Thats about the best guarantee you are likely to get from us. :hand:

OK, that's good enough. I'll obviously track the adult ratings. If there is obvious deflation, I'll take it you'll react accordingly.

Alan Shore
01-09-2004, 05:17 PM
OK, that's good enough. I'll obviously track the adult ratings. If there is obvious deflation, I'll take it you'll react accordingly.

If only you had of played some more Open tournaments involving juniors Dave you could have illustrated your point first-hand.. instead you've increased your rating from skilled play against adults!


i would like to take this opportunity to say how much i love gliko system specially at rating time

Me too.. I'm now 9th in QLD for rapid out of all active players.

Garvinator
01-09-2004, 05:40 PM
Me too.. I'm now 9th in QLD for rapid out of all active players.that means youll put your ranking to a true test at the gold coast rapid championships on sunday 24 october right
:lol:

Bill Gletsos
01-09-2004, 06:29 PM
I may be missing something obvious, but I don't know what RD is. :)
RD is a fundamental part of the Glicko System. It represent a Rating Deviation. In simple terms it can be considered a measure of reliability. The lower it is the more reliable the rating. The sysbols used just equate to various RD value ranges.



Done.
Ok Your originally published June 2004 rating was 1233 and your Sept 2004 rating is now 1404 an apparent increase of 171. However your recalculated June 2004 rating was 1352 which therefore only represents an increase of 52 points. Thats why you are not on the top improvers list.

Recherché
01-09-2004, 06:48 PM
Ok Your originally published June 2004 rating was 1233 and your Sept 2004 rating is now 1404 an apparent increase of 171. However your recalculated June 2004 rating was 1352 which therefore on represents an increase of 52 points. Thats why you are not on the top improvers list.

Oh yes, of course. It also explains why my ratings increase was much higher than I had been expecting, based on previous increases and my results.

Thanks. :)

(and thanks for clarifying the RD also)

Bill Gletsos
01-09-2004, 07:05 PM
Oh yes, of course. It also explains why my ratings increase was much higher than I had been expecting, based on previous increases and my results.
Your previously published ratings had been
Dec 2003 829!
Mar 2004 1146!
Jun 2004 1223!

and the new ones

Dec 2003 1055!
Mar 2004 1298!
Jun 2004 1352!!

which equates to differences of 226, 152 and 129 respectively.

The major cause of your increased Dec 2003 rating was the higher unreliability given to your rapid rating which was used as a starting value.
This increased rating then carried over into the following periods.



Thanks. :)
No problem.

Cat
01-09-2004, 07:34 PM
If only you had of played some more Open tournaments involving juniors Dave you could have illustrated your point first-hand.. instead you've increased your rating from skilled play against adults!


I'm pleased you found my play skilled! I did play a couple of the juniors, including Finkeeeee for my rating increase, so it wasn't all against the wrinklies. But point taken, weekends are hard for me, I've got kids & work on monday - a whole weekend's a bit of a drudge. Besides, I'm trying to keep up my support for the Chess Centre.

Rincewind
05-09-2004, 05:07 PM
hey barry have you updated your site ,its great to know how one is travelling between rating list ,

Now updated!

http://www.bjcox.com/modules.php?name=Glicko_Calc

Alan Shore
05-09-2004, 05:15 PM
I noticed on Barry's site when I looked up name there is a D Sampson with a rating of 2020?? from NSW. Is this really a separate person or perhaps is this from when I played a tourn in NSW and had a performance rating around this figure, some three years ago?

I noticed some years ago a similar error was made, giving 'Brian Jones' from 'QLD' a separate rating.

Bill Gletsos
05-09-2004, 05:31 PM
I noticed on Barry's site when I looked up name there is a D Sampson with a rating of 2020?? from NSW. Is this really a separate person or perhaps is this from when I played a tourn in NSW and had a performance rating around this figure, some three years ago?
Nice guess but wrong especially considering the D Sampson on file last played prior to 1981.

Alan Shore
05-09-2004, 05:33 PM
Nice guess but wrong especially considering the D Sampson on file last played prior to 1981.

Ah ok.. that settles that. Pity though, it's a nice rating.. I saw it was one of the ones that got owned with the adjustments too, used to be 2170. I could have almost entered the Aus Champs with sufficient swindling.

pax
06-09-2004, 10:31 AM
Bill,

This question may have been answered elsewhere.

How exactly do you calculate the intermediate ratings used in your recent system modification? The Bulletin post does not state this specifically, or reference the specifics.

Pax

Bill Gletsos
06-09-2004, 12:05 PM
Bill,

This question may have been answered elsewhere.

How exactly do you calculate the intermediate ratings used in your recent system modification? The Bulletin post does not state this specifically, or reference the specifics.

Pax
I mentioned it briefly back in the "Planned Rating Changes" thread.

All we are doing is applying a first pass of the Glicko2 system to determine the intermediate rating.
Since testing showed that this rather simple change was found to increase predictive accuracy it was adopted.
It may well turn out that tweaking the calculation of the intermediate rating in some other manner would further enhance predictive accuracy but this aspect has not been tested yet.

I first mentioned the issue of intermediate ratiings back in June 2003 in the Rating Theory 1001 thread on the old BBS. However with regards the ACT junior problem, the 2200+ anomaly and the FIDE realignment investigation it got put on the back burner until this year.

I should point out that when we make a number of changes as described in bulletin #276 we actually test each one seperately to determine if there is any advantage to ge gaiined. Tests are then run to see if combining the changes still provides an advanatge and whether this advantage is better than what was currently in place as well as if the advantage of combing them is better than any individual change.

pax
06-09-2004, 12:27 PM
I mentioned it briefly back in the "Planned Rating Changes" thread.

All we are doing is applying a first pass of the Glicko2 system to determine the intermediate rating.
Since testing showed that this rather simple change was found to increase predictive accuracy it was adopted.
It may well turn out that tweaking the calculation of the intermediate rating in some other manner would further enhance predictive accuracy but this aspect has not been tested yet.

And that intermediate rating is used for all players?

You know, that sounds not so very far removed in principle from my suggestion of batching the ratings in smaller periods. Both are aiming to use more up to date information in calculating the next rating change. The difference is that you are using games for the whole period to work out the intermediate rating (which is then used during the whole period), whereas my suggestion is simply to update the rating more often during the period to keep the rating as current as possible.

I'm a bit surprised, then, that you expressed the opinion that my suggestion would have no practical effect.

A theoretical difficulty I have with the intermediate ratings, is that it makes the system non-causal. That is, my rating gain from a particular game now depends on future results by that opponent. This will also exaccerbate the difference that a particular result will have on a rating depending on whether it is rated at the end of one rating period or the beginning of the next.
Maybe it works in practise - we shall see.

Pax

PHAT
06-09-2004, 03:08 PM
You know, that sounds not so very far removed in principle from my suggestion of batching the ratings in smaller periods.

pax, I agree with you, but don't expect BG to to anything other than say you are wrong.

arosar
06-09-2004, 03:27 PM
I wonder how long it will be before BG gets frustrated with pax and starts calling him names?

(Sorry...sorry...couldn't help meself).

It's nice to see another person here by the way.

AR

pax
06-09-2004, 03:39 PM
I wonder how long it will be before BG gets frustrated with pax and starts calling him names?

(Sorry...sorry...couldn't help meself).

It's nice to see another person here by the way.

AR

Oh don't worry, I've already been called "stupid" and an "idiot". I haven't quite graduated to "fool", "clown" or "moron" yet, but I'm working on it ;)

Actually, I've been around since well before this BB started. I'm just a bit sporadic is all.

Pax

Bill Gletsos
06-09-2004, 03:40 PM
And that intermediate rating is used for all players?
Yes.
For the calculation of the player's rating his rating at the start of the period is still used but the opponents intermediate rating is used as the opponents rating rather than their rating at the start of the rating period.


You know, that sounds not so very far removed in principle from my suggestion of batching the ratings in smaller periods. Both are aiming to use more up to date information in calculating the next rating change. The difference is that you are using games for the whole period to work out the intermediate rating (which is then used during the whole period), whereas my suggestion is simply to update the rating more often during the period to keep the rating as current as possible.
Its similar.
One of the major problems with game by game rating is the following. Player A is 1500 and plays player B who is 1500. Player A wins.
Now all rating systems would increase A's rating and decrease B rating.
However with only the one game result you have no way of knowing if player A actually stronger than 1500 or is it that player B has declined and is weaker than 1500.


I'm a bit surprised, then, that you expressed the opinion that my suggestion would have no practical effect.
The situations although similar are not exactly the same.
You were just having shorter periods.
Your suggestion of a shorter period still was not using an intermediate rating for the calculation of ratings in the first monthly period.
I based my opinion on the fact that previous testing showed that game by game or a short period (weekly or monthly) is is not as good as a longer period.

Testing showed that using intermediate ratings for a 3 mth period were superior to just batching 3 mths.

Graham and I believe that intermediate ratings based on say monthly periods would be superior to straight monthly periods but inferior to the current setup.



A theoretical difficulty I have with the intermediate ratings, is that it makes the system non-causal. That is, my rating gain from a particular game now depends on future results by that opponent. This will also exaccerbate the difference that a particular result will have on a rating depending on whether it is rated at the end of one rating period or the beginning of the next.
Maybe it works in practise - we shall see.
As I said testing showed that use of the intermediate rating led to increased predictive accuracy.
If it hadnt we wouldnt have implemented it. ;)

Garvinator
06-09-2004, 03:55 PM
Oh don't worry, I've already been called "stupid" and an "idiot". I haven't quite graduated to "fool", "clown" or "moron" yet, but I'm working on it ;)

Actually, I've been around since well before this BB started. I'm just a bit sporadic is all.

Pax
you will have to work alot harder than that to get the esteemed and much sort after title of goose though :lol: :whistle:

pax
06-09-2004, 04:00 PM
One of the major problems with game by game rating is the following. Player A is 1500 and plays player B who is 1500. Player A wins.
Now all rating systems would increase A's rating and decrease B rating.
However with only the one game result you have no way of knowing if player A actually stronger than 1500 or is it that player B has declined and is weaker than 1500.

That's not a major problem by any stretch. The expectation of player A's true strength is 1500, as is B's prior to the game.

When they play a game, you have an observation which suggests that player A may be stronger than player B. It doesn't say player A is definitely stronger, or player B definitely weaker. Player B may even be stronger than A. What you do have is an extra bit of evidence which will add to previous evidence to say that the expectation of A's rating is a bit higher than 1500 and B a bit lower. It is still an expectation of an unknown value, so nothing is ever known with certainty.

Pax

Bill Gletsos
06-09-2004, 04:08 PM
Oh don't worry, I've already been called "stupid" and an "idiot". I haven't quite graduated to "fool", "clown" or "moron" yet, but I'm working on it ;)
Actually instead of calling you explicitly stupid I said "Try not to demonstrate stupidty. I dis however say "dont be a complete idiot.

As for fool and moron you will have to work a lot harder for those.
Try to avoid working on it. ;)

Bill Gletsos
06-09-2004, 04:12 PM
That's not a major problem by any stretch. The expectation of player A's true strength is 1500, as is B's prior to the game.

When they play a game, you have an observation which suggests that player A may be stronger than player B. It doesn't say player A is definitely stronger, or player B definitely weaker. Player B may even be stronger than A. What you do have is an extra bit of evidence which will add to previous evidence to say that the expectation of A's rating is a bit higher than 1500 and B a bit lower. It is still an expectation of an unknown value, so nothing is ever known with certainty.

Pax
I dont entirely agree.
You dont know from that single game if A really is a bit stronger than 1500 or not as it depends on B. You just know that the game suggest he is stronger than B.
The point is if you had batched that game with others you would have had a better indication of whether A is improving or B is declining.

Bill Gletsos
06-09-2004, 04:13 PM
you will have to work alot harder than that to get the esteemed and much sort after title of goose though :lol: :whistle:
:lol: :lol: :lol:

Kevin Bonham
06-09-2004, 11:24 PM
A theoretical difficulty I have with the intermediate ratings, is that it makes the system non-causal. That is, my rating gain from a particular game now depends on future results by that opponent.

I like this feature a lot actually. It means that if you get sconed by an opponent who is clearly underrated, chances are you will not suffer as much as losing to an accurately-rated opponent of the same rating.

pax
07-09-2004, 09:37 AM
I dont entirely agree.
You dont know from that single game if A really is a bit stronger than 1500 or not as it depends on B. You just know that the game suggest he is stronger than B.
The point is if you had batched that game with others you would have had a better indication of whether A is improving or B is declining.

I don't get this argument Bill. You don't know anything with 100% certainty about any set of results. If A beats 1500 players ten times in succession, it still is statistically possible (though with smaller probability) that A got lucky ten times, or that 10 players got weaker.

What you do have is an extra piece of statistical evidence. That is, an expectation based on all of your evidence that A's strength is higher than 1500 and B's lower. The rating is simply the expectation (or mean) of an uncertain probability distribution of the player's strength. The addition of a new piece of evidence must push that mean in one direction or another.

Whether you batch the calculations or not, A defeating B will result in A gaining points. There's no other way to interpret the result.

If you have a sequence of results, and you calculate ratings game by game the net effect should be that the rating of A converges to the true strength, as long as the games occur often enough with respect to the change with time of A's true strength, and that the results are unbiased.

Pax

pax
07-09-2004, 09:39 AM
I like this feature a lot actually. It means that if you get sconed by an opponent who is clearly underrated, chances are you will not suffer as much as losing to an accurately-rated opponent of the same rating.

It may have practical advantages. My objection is mainly one of inelegance.

Bill Gletsos
07-09-2004, 01:54 PM
I don't get this argument Bill. You don't know anything with 100% certainty about any set of results. If A beats 1500 players ten times in succession, it still is statistically possible (though with smaller probability) that A got lucky ten times, or that 10 players got weaker.
I never said you knew anything with a 100% certainty.
All you can surmise from the single game is that A is more likely to be stronger than B than he is to be weaker than B. With no additional data you cannot determine how much stronger he is or if he is stronger at all.


What you do have is an extra piece of statistical evidence. That is, an expectation based on all of your evidence that A's strength is higher than 1500 and B's lower. The rating is simply the expectation (or mean) of an uncertain probability distribution of the player's strength. The addition of a new piece of evidence must push that mean in one direction or another.
Whether you batch the calculations or not, A defeating B will result in A gaining points. There's no other way to interpret the result.
Its not a matter of A gaining points, its a matter of how many points.

Player A rated 1500 plays player B rated 1500 twice over 2 successive days.
If A beats B then B beats A.
Now to keep the maths simple lets use Elo with a K of 30.
After the first game A's rating is 1515 and B's 1485.
after the second A's is 1498.71 and B's 1501.29.
However what if the results had been in the order A lost to B then A beat B.
After the first game A's rating is 1485 and B's 1515.
After the second A's is 1501.29 and B's 1498.71.
Now after this game both Player A plays 20 players rated 1500 for 20 draws and Player B plays a different group of 20 players rated 1500 for 20 draws.
In the case where A won the first game against B, the additional 20 games will result in A's rating will approaching 1500 from below whilst B's rating will approach 1500 from above.
Where B won the first game the situation is reversed.
Now although the results are mathematically correct, the practicality is that a true indication of player A and player B's strength following there two games would be 1500.


Now lets look at this scenario.
Player A rated 1500 plays players 11 players all rated 1500 in successive rounds in a competition where the games are played daily. A results are win, loss followed by 9 draws.

If you rate the results game by game then after the first game, A will have a rating of 1515 and after the second a rating of 1499.35.
His rating for the remaining games will be 1499.38, 1499.41, 1499.44, 1499.46, 1499.48, 1499.52, 1499.54, 1499.56, 1499.58.
Whats more the first guy he drew with ended up with a rating of 1499.97 as does the second. The remaining guys end up with ratings of 1499.98.
However if you batched them Player A's rating would have been 1500, as would have the ratings of all those he drew with.


Finally let's look at the following scenario.
Player A rated 1500 plays a single grade match game in the rating period against player B also rated 1500. A wins.
B on the other hand is an active player who during the rating period has apart from his game against A has played another 20 games against players for a performance rating for those 20 games of 1400.

Player A however really is only 1500 strength.

With just the single game information A increases by 15 to 1515.

However if you had used intermediate ratings B's intermediate rating would have been 1400.96.
A's new rating therefore would have been 1510.84.

With only the single game to go on for A the 1511 rating is a better than the 1515.


If you have a sequence of results, and you calculate ratings game by game the net effect should be that the rating of A converges to the true strength, as long as the games occur often enough with respect to the change with time of A's true strength, and that the results are unbiased.
Sounds like a lot of if's in that statement. ;)

pax
07-09-2004, 02:50 PM
I never said you knew anything with a 100% certainty.
All you can surmise from the single game is that A is more likely to be stronger than B than he is to be weaker than B. With no additional data you cannot determine how much stronger he is or if he is stronger at all.


Yes, but your estimate of his strength must now show A to be stronger than 1500 based on the evidence. A simple application of Bayes rule shows this.



Player A rated 1500 plays player B rated 1500 twice over 2 successive days.
If A beats B then B beats A.
Now to keep the maths simple lets use Elo with a K of 30.
After the first game A's rating is 1515 and B's 1485.
after the second A's is 1498.71 and B's 1501.29.
However what if the results had been in the order A lost to B then A beat B.
After the first game A's rating is 1485 and B's 1515.
After the second A's is 1501.29 and B's 1498.71.


This is fine, if you assume that chess strength is time varying (which, of course it is). What you end up with is a rating that gives slightly more weight to recent results. I don't think there's any problem with that in principle (indeed, it's the whole reason I advocate shorter batches or single games).



Now after this game both Player A plays 20 players rated 1500 for 20 draws and Player B plays a different group of 20 players rated 1500 for 20 draws.
In the case where A won the first game against B, the additional 20 games will result in A's rating will approaching 1500 from below whilst B's rating will approach 1500 from above.
Where B won the first game the situation is reversed.
Now although the results are mathematically correct, the practicality is that a true indication of player A and player B's strength following there two games would be 1500.


Still no problem for me, as the more recent results will have slightly higher weight, which is fine. Both ratings approach 1500 as the results approach a consistent 1500.

Under batch rating calculations, the first game might occur on the last day of the rating period and the second game at the beginning of the next, with similar effect.



However if you batched them Player A's rating would have been 1500, as would have the ratings of all those he drew with.


Ok, and that is fine under the assumption that a player's true strength is constant.

Now there is some compromise to be had here. Obviously, a player's true strength will not change very much over a day. It clearly can change substantially over three months. Your intermediate rating system is designed to capture something of this time variation as is my shorter batch system.



Finally let's look at the following scenario.
Player A rated 1500 plays a single grade match game in the rating period against player B also rated 1500. A wins.
B on the other hand is an active player who during the rating period has apart from his game against A has played another 20 games against players for a performance rating for those 20 games of 1400.

Player A however really is only 1500 strength.

With just the single game information A increases by 15 to 1515.

However if you had used intermediate ratings B's intermediate rating would have been 1400.96.
A's new rating therefore would have been 1510.84.

With only the single game to go on for A the 1511 rating is a better than the 1515.


Ok, this is true, but only if you know what the true strengths are. Based only on the evidence, you might draw the conclusion that between game 1 and game 2, player B had a drop in strength of 100 points. Maybe he copped a blow to the head or something ;) The fact remains that at the time of the game A vs B, the evidence available estimated both players at 1500. I don't like the principle of basing rating changes on future results. After all, B might not play any more games in the period after the first - you still have to compute the rating change. I agree however, that the intermediate rating has the advantage of incorporating a time varying element within a rating period.



Sounds like a lot of if's in that statement. ;)

As you well know, many assumptions are required to make any guarantees about convergence in any rating system (just look at the assumptions and approximations in Glickman's papers).

arosar
07-09-2004, 03:10 PM
Hey Bill . . . stop being so abusive Bill. It's not very elegant.

You know Bill, you could just about write an anthology of your ratings related posts by now. It could be the bible for all the Particles out there - both discovered and undiscovered.

AR

Bill Gletsos
07-09-2004, 03:20 PM
Hey Bill . . . stop being so abusive Bill. It's not very elegant.
DR keeps recycling the same old crap.
He deserves what he gets.

Bill Gletsos
07-09-2004, 03:26 PM
Yes, but your estimate of his strength must now show A to be stronger than 1500 based on the evidence. A simple application of Bayes rule shows this.
I dont think I was disagreeing with you on this point.
A should be stronger than B even if only by a minisule amount.


This is fine, if you assume that chess strength is time varying (which, of course it is). What you end up with is a rating that gives slightly more weight to recent results. I don't think there's any problem with that in principle (indeed, it's the whole reason I advocate shorter batches or single games).
Except that testing shows that single games or monthly periods is inferior to the current 3mth period.



Still no problem for me, as the more recent results will have slightly higher weight, which is fine. Both ratings approach 1500 as the results approach a consistent 1500.

Under batch rating calculations, the first game might occur on the last day of the rating period and the second game at the beginning of the next, with similar effect.
True.




Ok, and that is fine under the assumption that a player's true strength is constant.

Now there is some compromise to be had here. Obviously, a player's true strength will not change very much over a day. It clearly can change substantially over three months. Your intermediate rating system is designed to capture something of this time variation as is my shorter batch system.
Its like I noted in post #112.



Ok, this is true, but only if you know what the true strengths are. Based only on the evidence, you might draw the conclusion that between game 1 and game 2, player B had a drop in strength of 100 points. Maybe he copped a blow to the head or something ;)
Possible but extremely highly improbable. ;)


The fact remains that at the time of the game A vs B, the evidence available estimated both players at 1500. I don't like the principle of basing rating changes on future results. After all, B might not play any more games in the period after the first - you still have to compute the rating change.
Of course you have to calculate a rating but if B had played only the one game you are no worse off. However if you have the additional information you should use it as testing shows it leads to better predictive accuracy.
After all if it didnt you wouldnt do it.


I agree however, that the intermediate rating has the advantage of incorporating a time varying element within a rating period.
Its main advantage is that it leads to better predictive accuracy.


As you well know, many assumptions are required to make any guarantees about convergence in any rating system (just look at the assumptions and approximations in Glickman's papers).
The reason for my ;)

pax
07-09-2004, 03:33 PM
Except that testing shows that single games or monthly periods is inferior to the current 3mth period.


I'll take your word for it.

Bill Gletsos
07-09-2004, 03:46 PM
I'll take your word for it.
Thanks. :)

Believe me, if testing had shown it had any real advantage then we would have found a way to try and implement it while still maintaining the current rating periods. One possible way to have done this would have been to have broken the processing up into say monthly batches.
e.g The last rating period was March this one is June.
Using the March ratings first process all events that ended prior to 1st March as a period.
Then process all events ending 1st April-30th April as a period.
Finally process all events ending 1st May-31st May.

ursogr8
09-09-2004, 08:08 AM
You arent purchasing a product at all.
The ACF runs a rating system. If you want your games rated you have to pay an admin fee.
Note it isnt a rating fee. Its an administrative fee that just happens to be tied to game submitted for rating.



Bill

I just came across this paragraph of yours, buried in a response to DR. (Usually, I sort of glaze over the posts between you two, as I said previously.........and yes I think you win hands down....but, I am not keeping score)
Anyhow........back to your para. ......... mate, you are just wrong. The rating is a product in the eyes of the buyer. It may not be in your eyes as a constructor. But the punters love parts of the product Bill, and are willing to pay for it.

The purpose of another thread is to find out what the buyers see as important in the product.

starter

ursogr8
17-09-2004, 10:35 PM
Bill

I have a techo. request from the local pairings chappie here.
When you initially release the ratings we have problems importing the file direct into SP.
First, we don't understand why the SP file is not released by you immediately. Is it hard to produce?

Second, what you do release we apparently have to pre-process before we can import to SP. It is a difficult pre-process because of the format you adopt. If you persist with this approach could you reformat so that it looks like

tab, id, tab, rating, tab, garbage col, tab, Name

The garbage col to contain things like 1g or !! or Unrated.

starter

Garvinator
17-09-2004, 11:10 PM
Bill

I have a techo. request from the local pairings chappie here.
When you initially release the ratings we have problems importing the file direct into SP.
First, we don't understand why the SP file is not released by you immediately. Is it hard to produce?

Second, what you do release we apparently have to pre-process before we can import to SP. It is a difficult pre-process because of the format you adopt. If you persist with this approach could you reformat so that it looks like

tab, id, tab, rating, tab, garbage col, tab, Name

The garbage col to contain things like 1g or !! or Unrated.

starter


starter, i have noticed some of these problems too. is this thread any use to you: swiss perfect database importing. http://www.chesschat.org/showthread.php?t=1275

I am wondering why the swiss perfect importing of september ratings isnt an option when the ratings are released?

Bill Gletsos
17-09-2004, 11:16 PM
starter, i have noticed some of these problems too. is this thread any use to you: swiss perfect database importing. http://www.chesschat.org/showthread.php?t=1275

I am wondering why the swiss perfect importing of september ratings isnt an option when the ratings are released?
The answer is simple.
The ratings are released on time for the start of the month.
Hiwever there may be errors reported in the first week or two whcih need to be corrected.
As I do not wish for there to be multiple SP ACF master trn files floaring around I dont issue the trn file till a couple od weeks later.

Generally those organisers who have an urgent need for it email me directly and I'll send them a copy on the proviso they do not distribute to others. That way when I send the copy to paulb for the ratings page, I can also send it to those that received the pre-release copy.

Bill Gletsos
17-09-2004, 11:17 PM
Bill

I have a techo. request from the local pairings chappie here.
When you initially release the ratings we have problems importing the file direct into SP.
First, we don't understand why the SP file is not released by you immediately. Is it hard to produce?

Second, what you do release we apparently have to pre-process before we can import to SP. It is a difficult pre-process because of the format you adopt. If you persist with this approach could you reformat so that it looks like

tab, id, tab, rating, tab, garbage col, tab, Name

The garbage col to contain things like 1g or !! or Unrated.

starter
The answer is no because I dont want people to use that file. It is missing data such as sex, dob etc.

Garvinator
17-09-2004, 11:23 PM
The answer is simple.
The ratings are released on time for the start of the month.
Hiwever there may be errors reported in the first week or two whcih need to be corrected.
As I do not wish for there to be multiple SP ACF master trn files floaring around I dont issue the trn file till a couple od weeks later.

Generally those organisers who have an urgent need for it email me directly and I'll send them a copy on the proviso they do not distribute to others. That way when I send the copy to paulb for the ratings page, I can also send it to those that received the pre-release copy.
okey dokey :) when is the september version being released?

Bill Gletsos
17-09-2004, 11:33 PM
okey dokey :) when is the september version being released?
This weekend.

Duff McKagan
19-09-2004, 03:01 PM
Quote:
Originally Posted by Bill Gletsos
Nice guess but wrong especially considering the D Sampson on file last played prior to 1981.

Ah ok.. that settles that. Pity though, it's a nice rating.. I saw it was one of the ones that got owned with the adjustments too, used to be 2170. I could have almost entered the Aus Champs with sufficient swindling.

Heh, well Bruce you mentioned that you were 140 points in arrears, you can have 140 of mine... in the last two periods i've gone up 147 for doing nothing but playing like a wet lettuce.

Incidentally Bruce left Samson in 1981 :)

Duff McKagan
19-09-2004, 03:25 PM
An unknown Russian enters Mount Buller Open championships. Now then, how do you fix this cock-up Bill, assuming these results are achieved?

Bill Gletsos
19-09-2004, 04:22 PM
The only one making a cock up is you. ;)
There is absolutely no reason why you should have assigned him a rating of 1800?? in the first place.

If he is truly unknown then he is also unrated.

You cannot calculate a rating for a player who in unrated and scores either 0% or 100%.

On top of that your example is fairly stupid and unrealistic. :hand:

Kevin Bonham
19-09-2004, 06:32 PM
You cannot calculate a rating for a player who in unrated and scores either 0% or 100%.

Is that really the case in Glicko? I thought a weak Tassie player got a rapid rating with a zero score a few years back.

Of course a rating calculated from scratch for Duff's example would be nothing like 4400 and probably not even over 3000.

Bill Gletsos
19-09-2004, 07:03 PM
Is that really the case in Glicko? I thought a weak Tassie player got a rapid rating with a zero score a few years back.
What I said is correct.
The rating system has never given a player who scored 0% or 100% a rating.
Remember however that a players rapid is seeded from their normal and vice versa.


Of course a rating calculated from scratch for Duff's example would be nothing like 4400 and probably not even over 3000.
Based on Duff's example the player must be rated at least 3556.

Kevin Bonham
19-09-2004, 09:31 PM
What I said is correct.
The rating system has never given a player who scored 0% or 100% a rating.
Remember however that a players rapid is seeded from their normal and vice versa.

Sounds like the player may have had some previously played games our Ratings Officer wasn't aware of when we were discussing that case.
I'll see if I can find out who it was.


Based on Duff's example the player must be rated at least 3556.

I am really surprised it is that high. That is 1128 points above the average rating of the opponents - which on the old ELO conversion tables would come out at well over 99%. Yes, I know Glicko doesn't batch in the same way but I still think 3556 is way too high an estimate for 11/11 against that field. About 3000 or maybe 3100 seems right to me.

There's really no easy mathematical solution to the problem of estimating a rating for a player with a perfect score or perfect zero for practical purposes, but I've often thought Glicko is just a shade generous in its treatment of 100% results. This looks like another possible case of this.

Bill Gletsos
19-09-2004, 10:21 PM
Sounds like the player may have had some previously played games our Ratings Officer wasn't aware of when we were discussing that case.
I'll see if I can find out who it was.
Ok.


I am really surprised it is that high. That is 1128 points above the average rating of the opponents - which on the old ELO conversion tables would come out at well over 99%. Yes, I know Glicko doesn't batch in the same way but I still think 3556 is way too high an estimate for 11/11 against that field. About 3000 or maybe 3100 seems right to me.
Remember I'm doing true performance ratings (the rating at which the players rating wont change) not approx like FIDE.
In fact anything above 3467 rounds to 3467.
Even with 3556 that still allows for a 0.299382 rating increase and 4451 leads to an increase of 0.001737.
Its not until the rating hists 4548 that the increase drops below 0.001.

Also the original figure I gave and all the above are Elo (K=15) not Glicko.


There's really no easy mathematical solution to the problem of estimating a rating for a player with a perfect score or perfect zero for practical purposes, but I've often thought Glicko is just a shade generous in its treatment of 100% results. This looks like another possible case of this.
As I said my figure was Elo not Glicko.

Kevin Bonham
20-09-2004, 03:39 AM
Remember I'm doing true performance ratings (the rating at which the players rating wont change) not approx like FIDE.

Yes, I was aware of that. With the relatively small rating range of the example I would not expect that to make a huge difference.


Also the original figure I gave and all the above are Elo (K=15) not Glicko.

Hmmm, sounds like what I was interested in is an Elo property I had previously not noticed because Elo with k=15 is far too conservative for it to be picked up.

I reason something like this, theoretically: If a candidate scores x out of x, then there's some chance they're really thousands of points above the opposition, but it's far more likely they're someone who is just several hundred points better, expected to score a percentage in the high nineties and simply happened to get 100% in that case. Winning every game is clearly better than dropping half a point, but by how much?

For very rough quick-PR approximation purposes I like to substitute (x-b)/x for a 100% score or b/x for a 0% score, where b is some figure that must obviously be between 0 and 0.5. I had a go at modelling to find a value for b (using a sliding scale of probabilities for strengths matching expected scores between (x-0.5)/x and x/x), but had to approximate (as mentioned on the recreational maths thread I'm allergic to integral calculus). The approximate value of b that I get is 0.17, which means that I reckon (again using the FIDE approximations) a 100% performance in a tournament is about 130 ratings points better than dropping half a point. I think that approach is a lot more conservative than how such cases are actually handled - at least as far as your calculations based on Duff's example show.

ursogr8
25-10-2004, 01:12 PM
Box Hill Chess cLub has 180 members as at 25/10/4.
The distribution of player ratings (ACF September 2004 list) shows

% of Box Hill players with rating greater than
2147 =5%
1917 10
1849 15
1779 20
1679 25
1630 30
1586 35
1525 40
1443 45
1387 50
1342 55
1223 60
1192 65
1065 70
996 75
828 80
701 85

Bill, is a comparable distribution available from your data-base for Victoria? For Australia?

regards
starter

Bill Gletsos
26-10-2004, 09:57 AM
Not for such strange rating cutoffs.
See the percentage list on the ACF Ratings page for the divisions used.

ursogr8
26-10-2004, 10:04 AM
Not for such strange rating cutoffs.
See the percentage list on the ACF Ratings page for the divisions used.

Thanks Bill
I will re-cut my data to match what you have already provided via the ACF Ratings Page.
For a comparative exercise it doesn't matter whether there are
> equal rating spreads (the ACF data)
>> equal % spreads (my local data).
I will move towards yours.

tks
starter

ursogr8
28-10-2004, 05:26 PM
Thanks Bill
I will re-cut my data to match what you have already provided via the ACF Ratings Page.
For a comparative exercise it doesn't matter whether there are
> equal rating spreads (the ACF data)
>> equal % spreads (my local data).
I will move towards yours.

tks
starter

All
I am a bit disappointed in this (BHCC) data now that I have cut to match Bill's official ACF histogram of ratings.
The first disappointment is the higgledly-piggledly look when I paste EXCEL in here. First one to post how I improve the look gets a chocolate frog. :uhoh:

The main disappointment seems to come from the observation that Box Hill has 159 data-points but no players with a rating below 600. This skews our %'s to the top end and distorts the comparison. Ideas on how to remove this skew?


............................................BHCC % ACF %


2500 _ 2599 0.00 0.07
2400 _ 2499 1.26 0.35
2300 _ 2399 1.89 0.86
2200 _ 2299 2.52 1.69
2100 _ 2199 5.66 2.97
2000 _ 2099 9.43 6.05
1900 _ 1999 12.58 9.96
1800 _ 1899 20.13 15.46
1700 _ 1799 27.04 23.03
1600 _ 1699 37.11 31.81
1500 _ 1599 46.54 41.36
1400 _ 1499 55.35 50.83
1300 _ 1399 66.67 58.89
1200 _ 1299 71.70 65.63
1100 _ 1199 76.73 71.3
1000 _ 1099 84.28 76.94
900 _ 999 88.68 81.22
800 _ 899 90.57 85.24
700 _ 799 96.23 89.28
600 _ 699 100.00 91.98
500 _ 599 100.00 94.23
400 _ 499 100.00 96.33
300 _ 399 100.00 97.89
200 _ 299 100.00 99.1
100 _ 199 100.00 100


starter

Bill Gletsos
28-10-2004, 06:39 PM
All
I am a bit disappointed in this (BHCC) data now that I have cut to match Bill's official ACF histogram of ratings.
The first disappointment is the higgledly-piggledly look when I paste EXCEL in here. First one to post how I improve the look gets a chocolate frog. :uhoh:
Try putting it inside code tags.

ursogr8
28-10-2004, 07:46 PM
^^^
Thanks Bill
Worked a treat.

starter

Recherché
29-10-2004, 10:07 AM
The main disappointment seems to come from the observation that Box Hill has 159 data-points but no players with a rating below 600. This skews our %'s to the top end and distorts the comparison. Ideas on how to remove this skew?

What makes you describe it as a skew? I'm certainly not surprised that Box Hill is "stronger" than the national pool, given the quality of the tournaments and the coaching available there.

ursogr8
29-10-2004, 10:30 AM
What makes you describe it as a skew? I'm certainly not surprised that Box Hill is "stronger" than the national pool, given the quality of the tournaments and the coaching available there.

Well it seems to be a skew because one of our membership admittance criteria has had the effect of precluding players with a rating under 600. This factor then effectively increases the %s in the rating divisions above 600 (for BH) by an extra 8%. So, all our (BH) figures in the rating groups are inflated artificially relative to the comparable figure for the ACF.

starter

ps Incidentally, I started toying with the analysis in response to a few at BH saying that they were (rating-)depressed adults; as a consequence of under-rated juniors. At the moment, with the skew in the figures we look relatively over-rated instead, as a Club.

Recherché
29-10-2004, 12:30 PM
Well it seems to be a skew because one of our membership admittance criteria has had the effect of precluding players with a rating under 600. This factor then effectively increases the %s in the rating divisions above 600 (for BH) by an extra 8%. So, all our (BH) figures in the rating groups are inflated artificially relative to the comparable figure for the ACF.

I assume you're talking about the Whitehorse Junior Chess juniors who haven't started playing in the "normal" tournaments yet? Most of them are very young, what makes you think those sorts of players are routinely included in the ACF figures in other states? Many of them might well be only in junior or schools competitions that aren't ACF rated.


ps Incidentally, I started toying with the analysis in response to a few at BH saying that they were (rating-)depressed adults; as a consequence of under-rated juniors. At the moment, with the skew in the figures we look relatively over-rated instead, as a Club.

Well personally I think new, under-rated players are a natural part of the ratings system, be they juniors, or new adult players (such as myself). The ratings of everyone else should stabilise in relation to this cycle. It seems more natural to describe adults who are not subject to new players in this way as over-rated, instead.

Bill Gletsos
29-10-2004, 01:43 PM
Here are the ACF figures with all ratings below 600 removed as well as the VIC figures included.


............................................BHCC % ACF % VIC %



2600 _ 2699 0.00 0.04 0.00
2500 _ 2599 0.00 0.08 0.00
2400 _ 2499 1.26 0.38 0.66
2300 _ 2399 1.89 0.94 1.99
2200 _ 2299 2.52 1.84 2.99
2100 _ 2199 5.66 3.23 4.48
2000 _ 2099 9.43 6.58 7.13
1900 _ 1999 12.58 10.82 10.95
1800 _ 1899 20.13 16.80 18.24
1700 _ 1799 27.04 25.03 27.69
1600 _ 1699 37.11 34.57 37.31
1500 _ 1599 46.54 44.95 48.26
1400 _ 1499 55.35 55.24 57.88
1300 _ 1399 66.67 64.00 67.00
1200 _ 1299 71.70 71.33 74.63
1100 _ 1199 76.73 77.53 80.76
1000 _ 1099 84.28 83.65 86.73
900 _ 999 88.68 88.31 90.71
800 _ 899 90.57 92.67 93.86
700 _ 799 96.23 97.07 97.18
600 _ 699 100.00 100.00 100.00
500 _ 599 100.00 100.00 100.00
400 _ 499 100.00 100.00 100.00
300 _ 399 100.00 100.00 100.00
200 _ 299 100.00 100.00 100.00
100 _ 199 100.00 100.00 100.00

Kevin Bonham
29-10-2004, 02:13 PM
starter, I suggest the lack of sub-600s comes from a lack of very weak juniors playing mainly against each other.

A useful exercise for you, if you feel so inclined, could be to take out the ACF players who are below 600, and increase the rest of the ACF figures proportionally. That should give you a pretty good fix on where your players sit compared to serious tournament players in the rest of the country. It looks like there is relatively little difference once that is done.

ursogr8
29-10-2004, 03:41 PM
starter, I suggest the lack of sub-600s comes from a lack of very weak juniors playing mainly against each other.

No. The lack of U600 comes from the source I posted previously >
...our membership admittance criteria has had the effect of precluding players with a rating under 600.
Our U600 players are in another Club ...Whitehorse...with RAPID ratings only....until we admit them (via 4 criteria) to Box Hill


A useful exercise for you, if you feel so inclined, could be to take out the ACF players who are below 600, and increase the rest of the ACF figures proportionally. That should give you a pretty good fix on where your players sit compared to serious tournament players in the rest of the country. It looks like there is relatively little difference once that is done.
I have experimented with this..........although I adjusted BH figures rather than Bill's figures (which I left as sacrosanct :uhoh: ). Not yet ready for publication. But now I notice that Bill has adjusted his figures; I will pursue that angle too. I suspect after analysis I will agree with your conclusion.

starter

ursogr8
29-10-2004, 04:07 PM
Here are the ACF figures with all ratings below 600 removed as well as the VIC figures included.




Bill
Thanks for all the revision you did.
How did you resist drawing to the attention of DR?

starter

Bill Gletsos
30-10-2004, 12:06 AM
Bill
Thanks for all the revision you did.
I just dropped them out of the distribution.
There was no recalculation involved as far as ratings go to create to sort of 600 point floor.


How did you resist drawing to the attention of DR?
You dont always have to use a bazooka when a machine gun will do the job. :lol:

Also it helps to sometines keep the heavy artillery in reserve. ;)

Bill Gletsos
30-10-2004, 12:09 AM
I have experimented with this..........although I adjusted BH figures rather than Bill's figures (which I left as sacrosanct :uhoh: ).
Smart move there. ;)

Alan Shore
31-10-2004, 12:52 AM
I reckon we need some ACF Transfer/Suicide ratings.. that would kick ass... I could be #1 in AUS for suicide *offcially* ;)

ursogr8
31-10-2004, 08:10 AM
I assume you're talking about the Whitehorse Junior Chess juniors who haven't started playing in the "normal" tournaments yet?
I was referring to those who don't pass BH Memebership Sub-Committee's 4 entrance criteria.
But this probably equates to the group you mention.


Most of them are very young, what makes you think those sorts of players are routinely included in the ACF figures in other states?

Because there are non-zero figures globally.

starter

eclectic
31-10-2004, 10:27 PM
I was referring to those who don't pass BH Memebership Sub-Committee's 4 entrance criteria.
But this probably equates to the group you mention.



Because there are non-zero figures globally.

starter

starter,

mate ... ;)

you know how events submitted for ratings are meant to be in like a month before ?

(although bill does actually let results get sneaked in a few days before if he is in a particularly good mood)

well could you keep quiet for a month come midnight tonight so that you have something new to say for the upcoming december version of this thread?

regards,

eclectic

Kevin Bonham
31-10-2004, 10:51 PM
Our U600 players are in another Club ...Whitehorse...with RAPID ratings only....until we admit them (via 4 criteria) to Box Hill

Are you at liberty to divulge these criteria? I'd be interested, just out of curiosity really.

ursogr8
01-11-2004, 07:15 AM
Are you at liberty to divulge these criteria? I'd be interested, just out of curiosity really.
PM sent

ursogr8
01-11-2004, 07:24 AM
starter,

mate ... ;)
Yeesss ??? :uhoh:


you know how events submitted for ratings are meant to be in like a month before ?
Yeesss ??? :uhoh: :hmm:



(although bill does actually let results get sneaked in a few days before if he is in a particularly good mood)

Yeesss ??? :uhoh: :hmm: :eh:


well could you keep quiet for a month come midnight tonight so that you have something new to say for the upcoming december version of this thread

regards,

eclectic


:confused: :nervous: :shocked: :suprised: :huh:


:whatthe: :whatthe: :whatthe:

eclectic
I need you to translate what you mean.

starter

Bill Gletsos
01-11-2004, 06:02 PM
PM sent
Why not post them here so we can all see it.

Bill Gletsos
01-11-2004, 06:06 PM
BTW I notice that no Box Hill events have been submitted for rating by the CV ratings officer who submiitted all the events he had received so far.

Recherché
01-11-2004, 06:27 PM
^I hope it doesn't remain that way, if only the Interclub and none of the other tournaments count towards my December rating it's likely to be a sorry sight!

Bill Gletsos
01-11-2004, 06:30 PM
^I hope it doesn't remain that way, if only the Interclub and none of the other tournaments count towards my December rating it's likely to be a sorry sight!
So far I dont have the CV interclub either or for that matter the Vic Championship.
No doubt the CV ratings officer will be chasing the relevant parties.

ursogr8
01-11-2004, 07:25 PM
Why not post them here so we can all see it.
I will PM them to you if you wish to see them.

Bill Gletsos
01-11-2004, 11:12 PM
I will PM them to you if you wish to see them.
Ok, thanks.

My question still remains, why so secretive about them.

bobby1972
09-11-2004, 10:57 AM
Is Interclub Rated This Coming List

Recherché
09-11-2004, 11:45 AM
Is Interclub Rated This Coming List

As long as Chess Victoria submits the results in time, it should be.

Bill Gletsos
09-11-2004, 12:22 PM
The CV Ratings officer is very good.
I would expect him to submit the Interclub and Vic championship amongst others, provided of course they are actually sent to him in the first place.
If not, I suspect he would follow up on them anyway.

Recherché
09-11-2004, 12:45 PM
^ There we have it then.

Since the Interclub final-finals only finished a few weeks ago, it's probably still working its way through the system. :)