Author: Axel Schumacher
Date: 15:00:28 04/29/03
Go up one level in this thread
On April 29, 2003 at 13:57:28, Drexel,Michael wrote: >On April 29, 2003 at 13:26:47, Axel Schumacher wrote: > >>On April 29, 2003 at 04:24:12, Drexel,Michael wrote: >> >>>On April 28, 2003 at 21:02:27, Axel Schumacher wrote: >>> >>>>On April 28, 2003 at 19:57:53, Omid David Tabibi wrote: >>>> >>>>>On April 28, 2003 at 01:16:09, Axel Schumacher wrote: >>>>> >>>>>>News: >>>>>> >>>>>>-over 200 new engine-versions >>>>>>- 60 new engines >>>>>>- new Chessmaster settings, logos >>>>>>- extra TOP 50 amateur ranking >>>>>>- short engine report >>>>>> >>>>>>Surak's Trophy: The Trophy is an ongoing engine vs. engine tournament, played on >>>>>>several computers (minimum hardware pentium II, 350mhz: maximum: Athlon 2100+). >>>>>>Games played so far: #72.460 >>>>>>GUI's: The tournament is played within the following GUI's: Fritz, Winboard, >>>>>>Arena, Shredder-Classic and Chessmaster. >>>>>>Number of Engine-versions: #671 >>>>>>Time-controls: Mixed. Mostly Blitz (5m+1s); a lot of Rapids (10m+1s) and some >>>>>>games with long time-control (40m/40moves) or tournament-time. >>>>>>Books: Remis-book (Draw-book) or Kurzbuch (Short-book). >>>>>> >>>>>>Tourney-Page: >>>>>>http://www.grailmaster.com/misc/chess/comp/compindex.html >>>>>>or >>>>>>Chess-Page: >>>>>>http://www.grailmaster.com/misc/chess/chess.html >>>>>>or >>>>>>Homepage: >>>>>>www.grailmaster.com >>>>>> >>>>>>Here a small excerpt of one of the rating-lists (Best-versions list): >>>>>> >>>>>>Place/Engine/Elo/Games >>>>>>--------------------------------------- >>>>>>1. Deep Fritz 7 2761 395 >>>>>>2. Chess Tiger 14.0 2732 2301 >>>>>>3. Hiarcs 8 2706 1024 >>>>>>4. CM9000 Grailmaster7 2700 219 (best CM-Setting) >>>>>>5. Ruffian 1.0.1 2688 1092 (best Amateur !) >>>>>>6. Deep Junior 6.0 2681 1613 >>>>>>7. Shredder 7.0 2674 354 >>>>>>8. Green Light Chess 3.00 2653 90 (!!! Takes the 2nd Amateur place.) >>>>>>9. SOS.3 for Arena 2642 325 >>>>>>10. Pepito 1.59UCI 2630 268 (greatly improved !) >>>>>>11. Delfi 4.1 2628 185 (Soon a Top-engine; I'm sure) >>>>>>12. Gandalf 4.32UCI 2627 576 >>>>>>13. Crafty 17.13 2626 786 (long time no improvement) >>>>>>14. List 504 2609 430 >>>>>>15. Zarkov 4.5e 2607 315 >>>>>>16. Little Goliath 2000 v3.9 2603 270 >>>>>>17. Smarthink 0.15b1 2601 116 (also a rising star) >>>>>>18. Nimzo2000b EN 2601 80 >>>>>>19. Aristarch 4.4 2598 502 >>>>>>20. Yace 0.99.50 2591 403 >>>>>>.... >>>>>>196. Pyotr Novice 2.6 1815 143 (I'm 100% against it :-) ) >>>>>> >>>>> >>>>>A great website! Keep up the good work! >>>>> >>>>>Just a question: how do you calculate the ratings? It seems that the ratings are >>>>>too optimistic (minimum being over 1800). >>>>> >>>>> >>>>>> >>>>>>Cheers >>>>>>Axel >>>> >>>> >>>>Hi Omid, >>>>it's sadly the old problem. I calculate with Fritz. Fritz gives good results IF >>>>the Elo-startvalue is within the pgn/cbh-file. I don't have this for all engines >>>>because it would take several hours to do so. Usually the elos for the good >>>>engines are very accurate, the weak engines receiving too much points. >>>>Unfortunatelt, I save the games in CB-format, therefore I can't use EloStat. It >>>>takes also up to a day for my computer to translate 72.000 games into pgn. But >>>>maybe I will do that in the next update. >>>> >>>>Axel >>> >>>I dont think your ratings are accurate at all, and I dont believe the >>>Grailmaster 7 setting is the best CM-setting. >>>Even the personality I have posted some time ago won a 100 games blitz-match >>>against it. >>> >>>52.5-47.5 (+33 =39 -28) >>> >>>5 min, AMD 2200+, ponder off, Remis.ctg, 3,4,5-men, 64MB hash, alternate colours >>> >>>My personality lost the same match with Fritz8.ctg against Chessmaster SKR >>> >>>48-52 (+29 =38 -33) >>> >>>I would suggest to test your personalities and programs similarly to the Kurt >>>Utzinger and Rolf Bühler tests. >>>Their tests are IMO exemplary. >>> >>>Michael >> >> >>First of all, test between the different CM personalities are quite worthless, >>they produce results which does not represent their real playing strenght. > >Correct of course > >In contrast to all other testers I match the engines against at least 50 other >>engines, which gives the most accurate results. With this method the results are >>extremely reproducible. I made a test with Nemjet 3.05: >>The first 100 games: Elo 2524. >>The second 100 games: Elo 2524. >>The third 100 games: Elo 2523. >> >>Other engines behaved the same, therefore my method works very well. > >How should I now against whom and how often you matched the different >Chessmaster settings? >I noticed that almost all Chessmaster personalities play very strong >against some top engines and very bad against others. >IMO accurate tests should be played with the same conditions for all programs: > >- same number of games against all top programs This is O.K. if you test only a few engines. However, I test now 700 engine versions! >- same hardware All CM personalities run at the same computer. >- same time control (this is not mandatory, however results of blitz All have the same time control. > games and results of longer time control games should be seperated) > >Michael > > > > > > > > > > > > > >> >>Axel >> >> >> >> >>Axel
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.