The Winner: nn-ef1c0cd86e12.nnue - Congrats to Lonfom169! Note: All EvalFiles are as last release per Author, and available: NNs
# PLAYER : RATING POINTS GAMES (%) 1 ef1c0cd86e12 : 3743.0 418.0 800 52.3% 2 8084373356c9 : 3738.7 411.5 800 51.4% 3 4f56ecfca5b7 : 3738.4 411.5 800 51.4% 4 13406b1dcbe0 : 3735.5 407.0 800 50.9% 5 e80cf49f78ea : 3733.7 405.0 800 50.6% 6 a27a526755f0 : 3691.0 347.0 800 43.4%
Some Notes The overall draw percentage (based on 2400 games) 91% All games of the above ranking are played as Ponder OFF But later I switched to Ponder ON , TC 1m+1s, see below:-------------------------------------------------------
Some more results, for anyone interested:
1) Test with two different Evalfiles - Ponder ON, 1m+1s:
1 4f56ecfca5b7 +23/-22/=1053 50.05% 549.5/1098 2 ef1c0cd86e12 +22/-23/=1053 49.95% 548.5/1098
Some Notes: It seems, both are almost identical in Elo performance And as usual, the overall draw percentage is high: 96%-------------------------------------------------------
2) Testing two same Evalfiles - Ponder ON, 1m+1s:
1 ef1c0cd86e12 +16/-14/=738 50.13% 385.0/768 2 ef1c0cd86e12_2 +14/-16/=738 49.87% 383.0/768
Some Notes: The Target is to see the influence of draw percentage As expected, the draw percentage is again high: 96%-------------------------------------------------------
More Details, All Evalfiles are played by same engine: SugaR AI 2.5 BMI2 Exc.a27a526755f0 (former Champ) played by Cfish 110421 It seems, almost all NN files are almost identical in Elo strength.. Exc. a27a526755f0, which 40-50+ Elo weaker, but it's old-dated I realized to test it again...just out of curiosity and for comparing And no doubt that the latest EvalFiles have been improved a lot!
Note also that, Running matches under these conditions are not my favorite I preferred Bullet time control...mainly to reduce the draw perc. But as we see, even with fast time controls etc., no BIG sense.. Sure, for less draws, perc. I have to use very weak openings But for nowadays, I have no such plans, for more Details
GAMES
# PLAYER : RATING POINTS GAMES (%) 1 ef1c0cd86e12 : 3743.0 418.0 800 52.3% 2 8084373356c9 : 3738.7 411.5 800 51.4% 3 4f56ecfca5b7 : 3738.4 411.5 800 51.4% 4 13406b1dcbe0 : 3735.5 407.0 800 50.9% 5 e80cf49f78ea : 3733.7 405.0 800 50.6% 6 a27a526755f0 : 3691.0 347.0 800 43.4%
Some Notes The overall draw percentage (based on 2400 games) 91% All games of the above ranking are played as Ponder OFF But later I switched to Ponder ON , TC 1m+1s, see below:-------------------------------------------------------
Some more results, for anyone interested:
1) Test with two different Evalfiles - Ponder ON, 1m+1s:
1 4f56ecfca5b7 +23/-22/=1053 50.05% 549.5/1098 2 ef1c0cd86e12 +22/-23/=1053 49.95% 548.5/1098
Some Notes: It seems, both are almost identical in Elo performance And as usual, the overall draw percentage is high: 96%-------------------------------------------------------
2) Testing two same Evalfiles - Ponder ON, 1m+1s:
1 ef1c0cd86e12 +16/-14/=738 50.13% 385.0/768 2 ef1c0cd86e12_2 +14/-16/=738 49.87% 383.0/768
Some Notes: The Target is to see the influence of draw percentage As expected, the draw percentage is again high: 96%-------------------------------------------------------
More Details, All Evalfiles are played by same engine: SugaR AI 2.5 BMI2 Exc.a27a526755f0 (former Champ) played by Cfish 110421 It seems, almost all NN files are almost identical in Elo strength.. Exc. a27a526755f0, which 40-50+ Elo weaker, but it's old-dated I realized to test it again...just out of curiosity and for comparing And no doubt that the latest EvalFiles have been improved a lot!
Note also that, Running matches under these conditions are not my favorite I preferred Bullet time control...mainly to reduce the draw perc. But as we see, even with fast time controls etc., no BIG sense.. Sure, for less draws, perc. I have to use very weak openings But for nowadays, I have no such plans, for more Details
GAMES