Author: Majd Al-Ansari
Date: 06:38:56 01/29/06
Go up one level in this thread
I have a completely different result. I see quite a big difference and many
"won" games are now won instead of drawn. I have checked out quite a few games
and I will say that EGTB's greately improve endgame play for Rybka, and plug a
lot of holes. EGTB are especially important if the other side has them. Not
having them will leak a lot of points. Still there is quite a ways to go for
Rybka when endgames are concerned. It still plays some endings horribly. But
the gaps are getting smaller and smaller and I am very interested to see how
Rybka will be with beta 14 (EG knowledge added).
On January 29, 2006 at 08:19:23, Vasik Rajlich wrote:
>Normally I don't think it's appropriate for me to post my test results, but in
>the case of testing EGTB access I'll make an exception.
>
>Rybka Beta 13 (3 & 4 & selected 5 man TBs) vs Rybka Beta 12, G/3" on 3 64-bit
>AMD64 machines:
>
>+419 =710 -410
>
>This is really nothing new - in most cases, by the time tablebases give scores
>which differ significantly from the normal evaluation, it is too late for search
>to do anything about it.
>
>Barry ran a match from Sune & Joe's endgame suite on his machine, which has also
>(some) 6 man TBs. The game count is small, so no conclusions can really be
>drawn. The data is appended below.
>
>One interesting (but rare) case where the 6-man TBs made the decisive difference
>was:
>
>[D] 2k5/7R/4rp2/P1K4P/8/8/8/8 w - - 0 14
>
>Here, 6-man TBs are needed for Rybka to understand that 14. h6 throws away the
>win.
>
>This full game, and the full match data, is all below.
>
>Best regards,
>Vas
>
>
>
>[Event "Rybka v Hiarcs endgame, Blitz:15'"]
>[Site "ROSWELL-4500"]
>[Date "2006.01.25"]
>[Round "3"]
>[White "Rybka 1.01 Beta 12"]
>[Black "Hiarcs 10"]
>[Result "1/2-1/2"]
>[Annotator "1.62;1.71"]
>[SetUp "1"]
>[FEN "8/5pk1/r5pp/P7/3R3P/6P1/5PK1/8 w - - 0 1"]
>[PlyCount "97"]
>[TimeControl "900"]
>
>{AMD Opteron(tm) Processor 250 2391 MHz W=21.2 ply; 586kN/s B=15.4 ply;
>149kN/s; 148,737 TBAs} 1. Rd5 {Both last book move 1.62/20 45} Kf6 {1.71/16 62}
>2. Kf3 {(Rb5) 1.68/20 4} h5 {(Ke6) 1.71/15 24} 3. Ke4 {(Rb5) 1.87/19 29} Re6+ {
>(Ke6) 1.77/15 26} 4. Kd4 {(Kd3) 2.21/18 32} Ra6 {1.93/16 52} 5. f3 {2.52/21 0}
>Ke6 {(Ke7) 1.98/15 32} 6. g4 {(Rb5) 2.66/20 41} hxg4 {1.97/14 27} 7. fxg4 {
>(Re5+) 2.80/20 1} f6 {(Ra8) 1.95/14 32} 8. Rb5 {2.98/18 20} Kf7 {
>(Ke7) 2.12/14 12} 9. Kc5 {3.48/19 19} Re6 {2.34/13 4} 10. Rb7+ {3.53/16 5} Ke8
>{3.81/15 148} 11. Rh7 {(Rb6) 4.42/23 0} Kd8 {(Kf8) 3.54/15 95} 12. h5 {
>(Rh6) 5.36/18 24} gxh5 {3.30/13 24} 13. gxh5 {5.36/21 0} Kc8 {(f5) 12.02/16 126
>} 14. h6 {(Rf7) 5.36/18 6} f5 {2.70/11 8} 15. Rh8+ {(Kd5) 5.36/21 29} Kb7 {
>2.00/12 8} 16. h7 {(Kd5) 5.36/22 28} Rh6 {0.02/14 9} 17. a6+ {(Kd5) 5.36/23 11}
>Ka7 {0.02/15 4} 18. Kd4 {(Kc4) 5.36/24 40} f4 {(Rh4+) 0.02/17 6} 19. Ke4 {
>5.36/23 60} f3 {(Rh4) 0.02/19 0} 20. Kxf3 {5.36/21 90} Rh3+ {(Rxa6) 0.00/1 0}
>21. Kf2 {0.00/1 0} Rh4 {0.00/1 0} 22. Kg1 {0.00/1 0} Rh6 {0.00/1 0} 23. Kg2 {
>0.00/1 0} Rh4 {0.00/1 0} 24. Kg3 {0.00/1 0} Rh1 {0.00/1 0} 25. Kf4 {0.00/1 0}
>Rh4+ {0.00/1 0} 26. Ke5 {0.00/1 0} Rh6 {0.00/1 0} 27. Ke4 {0.00/1 0} Rh2 {
>0.00/1 0} 28. Kf3 {0.00/1 0} Rh4 {0.00/1 0} 29. Kg3 {0.00/1 0} Rh6 {0.00/1 0}
>30. Kg2 {0.00/1 0} Rh4 {0.00/1 0} 31. Kf2 {0.00/1 0} Rh2+ {0.00/1 0} 32. Kf3 {
>0.00/1 0} Rh5 {0.00/1 0} 33. Ke4 {0.00/1 0} Rh1 {0.00/1 0} 34. Kd5 {0.00/1 0}
>Rh3 {0.00/1 0} 35. Ke6 {0.00/1 0} Rh5 {0.00/1 0} 36. Kf6 {0.00/1 0} Rh2 {
>0.00/1 0} 37. Ke6 {0.00/1 0} Rh4 {0.00/1 0} 38. Kf5 {0.00/1 0} Rh2 {0.00/1 0}
>39. Kf4 {0.00/1 0} Rh5 {0.00/1 0} 40. Ke4 {0.00/1 0} Rh1 {0.00/1 0} 41. Kf5 {
>0.00/1 0} Rf1+ {0.00/1 0} 42. Kg4 {0.00/1 0} Rh1 {0.00/1 0} 43. Kg5 {0.00/1 0}
>Rg1+ {0.00/1 0} 44. Kf4 {0.00/1 0} Rh1 {0.00/1 0} 45. Ke5 {0.00/1 0} Re1+ {
>0.00/1 0} 46. Kd6 {0.00/1 0} Rh1 {0.00/1 0} 47. Ke5 {0.00/1 0} Rh4 {0.00/1 0}
>48. Kf6 {0.00/1 0} Rh1 {0.00/1 0} 49. Ke5 {Draw accepted 0.00/1 0} 1/2-1/2
>
>
>
>
>******************************************************************************************
>
>
>4 matches Rybka Beta 12-x64 and 13-x64 v Hiarcs 10
>
>1) Rybka Beta 12 +10/-9/=21 51.25% 20.5/40 elo +8 3 min + 0
> Hiarcs 10 +9/-10/=21 48.75% 19.5/40
>
>2) Rybka Beta 13 +11/-9/=20 52.25% 21.0/40 elo +17 3 min + 0
> Hiarcs 10 +9/-11/=20 47.75% 19.0/40
>
>3) Rybka Beta 12 +12/-11/=17 51.25% 20.5/40 elo +8 15 min + 0
> Hiarcs 10 +11/-12/=17 48.75% 19.5/40
>
>4) Rybka Beta 13 +14/-10/=16 55.00% 22/40 elo +34 15 min + 0
> Hiarcs 10 +10/-14/=16 45.00% 18/40
>
>
>*****************************************************************************************
>
>Engines -
>
>Rybka 1.01 Beta 12-x64 - defaults
> Beta 13-x64 - defaults
>
>Hiarcs 10 - position learning turned off
>
>Book - perfect v9.0.ctg
>System - Windows x64, 3gb RAM, dual AMD Opteron @2.4ghz
>EGTB- all common 3-4-5, some 6
>GUI - Fritz 9 (date Dec 20)
>Hash - 128mb (3min) each engine
> 512mb(15min) each engine
>Book Learning - Off
>Book variety slider - maximum
>Draw - Never, Resign - Late
>Permanent brain - On - (1 cpu per engine)
This page took 0.01 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.