Page 1 of 2
T58 passing LD2
Posted: Sat Nov 30, 2019 12:18 am
by dkappe
According to the testers on discord, the T58 run has passed LD2 in strength. My own endgame tests put 58432 in third place behind LE and 42850. Looks like there’s more playing strength to be squeezed out of even the smaller nets.
Re: T58 passing LD2
Posted: Sat Nov 30, 2019 3:04 pm
by Eduard
On my GTX 1050 Ti, 1m+2s:
58432 vs LD2 +10 =33 -7
Re: T58 passing LD2
Posted: Sat Nov 30, 2019 5:48 pm
by Eduard
Quicktest on 1050 Ti, 1m+2s, GUI Fritz 17, Opening Hert500.pgn by Stefan Pohl:
58427 vs 42850 nach 20 Partien
+2 =12 -6
58447 vs 42850 nach 20 Partien
+6 =11 -3
Re: T58 passing LD2
Posted: Mon Dec 02, 2019 8:21 am
by AdminX
In my test results LD2 wins. Using 2x GTX1060, I gave each net their own GPU. Time control was 5'3 No opening books were used.
Code: Select all
Engine Match
Lc0 v0.23 58432 [3] - Lc0 v0.23 LD2 [3] 22.5 - 27.5 (50: +0, =45, -5) 45.00% Perf = -34 Elo
( 45.0% TP = -34 Elo 68%->[-80,-29] 95%->[-129,-24] 99.7%->[-184,-19] )
50 games played
Re: T58 passing LD2
Posted: Mon Dec 02, 2019 1:29 pm
by Eduard
Btw, my 50 games are played with Hert500 by Stefan Pohl. It seems that newest T58 nets are even stronger.
Re: T58 passing LD2
Posted: Mon Dec 02, 2019 1:40 pm
by Eduard
New strong setting for T58 on slow GPU - - >with LogitQ=true!!
[OPTIONS]
UCI_Chess960=true
Threads=2
MinibatchSize=64
MaxPrefetch=4
LogitQ=true
CPuct=2.700000
FpuValue=0.550000
PolicyTemperature=1.530000
MoveOverheadMs=75000
Slowmover=0.700000
I tested it live on playchess with RTX 1050 Ti. Timecontrol 16+0:
[pgn][Event "Wertungspartie, 16+0"]
[Site "Maschinenraum"]
[Date "2019.12.02"]
[Round "?"]
[White "Stockfish 32 Cores, 60.000 kns"]
[Black "Lc0 v0.23.0-rc2+git T58 58462 tuned 1050Ti"]
[Result "1/2-1/2"]
[ECO "A28"]
[WhiteElo "2483"]
[BlackElo "2427"]
[PlyCount "157"]
[EventDate "2019.12.02"]
[SourceTitle "playchess.com"]
1. c4 e5 2. Nc3 Nf6 3. Nf3 Nc6 4. e4 Be7 5. d4 exd4 6. Nxd4 O-O 7. Nc2 a5 8. f3
d6 9. Be2 Nd7 10. O-O Nc5 11. Ne3 Bd7 12. Rb1 Bg5 13. Re1 f5 14. Nxf5 Bxf5 15.
exf5 Bxc1 16. Rxc1 Rxf5 17. b3 Qh4 18. g3 Qd4+ 19. Qxd4 Nxd4 20. f4 Kf8 21. Bg4
Rf7 22. Rcd1 Nc6 23. h4 Re8 24. Rxe8+ Kxe8 25. Kg2 Nb4 26. Rd2 Kd8 27. h5 Rf8
28. Re2 c6 29. Rd2 Kc7 30. Re2 Kd8 31. Rd2 Kc7 32. a3 Nbd3 33. Na2 Ne1+ 34. Kf2
Nxb3 35. Kxe1 Nxd2 36. Kxd2 b5 37. cxb5 cxb5 38. Nc1 Rb8 39. Bd1 d5 40. Ne2 a4
41. Kc3 Re8 42. Kd2 Rb8 43. Kc3 Re8 44. Kd3 Rb8 45. Bc2 Kd6 46. Kc3 Rc8+ 47.
Kb2 Re8 48. Bd3 Kc5 49. Kc3 Re3 50. g4 Rh3 51. Nd4 h6 52. g5 hxg5 53. fxg5 Rxh5
54. g6 Rh3 55. Ne6+ Kb6 56. Nxg7 b4+ 57. axb4 a3 58. Nf5 a2 59. Kb2 Rxd3 60.
Kxa2 Rd2+ 61. Kb3 Rg2 62. Kc3 Rxg6 63. Ne7 Rd6 64. Nc8+ Kc7 65. Nxd6 Kxd6 66.
Kb3 d4 67. Kc4 d3 68. Kxd3 Kc6 69. Kc4 Kb6 70. b5 Ka7 71. Kc5 Kb7 72. b6 Kb8
73. Kc4 Ka8 74. Kb4 Kb7 75. Kb5 Kb8 76. Ka5 Kb7 77. Kb5 Kb8 78. Ka5 Kb7 79. Kb5
1/2-1/2
[/pgn]
Re: T58 passing LD2
Posted: Mon Dec 02, 2019 1:51 pm
by AdminX
Eduard wrote: ↑Mon Dec 02, 2019 1:29 pm
Btw, my 50 games are played with Hert500 by Stefan Pohl. It seems that newest T58 nets are even stronger.
Hi Eduard,
Yes, I saw that in testing results on Discord. I look forward to tesing with the new setting you posted.
Re: T58 passing LD2
Posted: Mon Dec 02, 2019 2:09 pm
by Eduard
Have fun! But you know, this
MoveOverhead and Slowmover Settings are only for playing on playchess. Not for tests on own Pc.
Re: T58 passing LD2
Posted: Mon Dec 02, 2019 5:44 pm
by Eduard
PolicyTemperature=1.520000 seems a little better.
Re: T58 passing LD2
Posted: Tue Dec 03, 2019 12:11 am
by carldaman
Thanks Eduard, but on a really slow GPU (AMD Radeon card) it begins to troll as early as the middlegame (!?)
So many suboptimal moves played in a crushing positions...