beram wrote:Could be because of your used SALC openingsbook. Perhaps latest SFdevs could be improved especially in the field of opposite castling situations.
Just my wild guess.
I doubt that. If the SALC-set is the reason, then why only against Houdini and Komodo a very big gain and against Gull and Equinox not?!? And in my singlecore Stockfish-testruns I use SALC-positions, too. So why is there no big Elo-gain against Komodo and Houdini?
Really strange...
And in the meantime, Stockfish 141112 climbed again over 67% score in my Endless RoundRobin...So it doesnt seem, that the result gets "more normal" with more games. Next week, I will update the Endless RoundRobin on my website.
Stay tuned!
beram wrote:Could be because of your used SALC openingsbook. Perhaps latest SFdevs could be improved especially in the field of opposite castling situations.
Just my wild guess.
I doubt that. If the SALC-set is the reason, then why only against Houdini and Komodo a very big gain and against Gull and Equinox not?!? And in my singlecore Stockfish-testruns I use SALC-positions, too. So why is there no big Elo-gain against Komodo and Houdini?
Really strange...
And in the meantime, Stockfish 141112 climbed again over 67% score in my Endless RoundRobin...So it doesnt seem, that the result gets "more normal" with more games. Next week, I will update the Endless RoundRobin on my website.
Stay tuned!
Stefan
Looking over various rating lists, it seems that the SF vs K matchup may become easier for SF as the length of increment is decreased (measured as the ratio to the base TC); i.e., you can validate this claim by re-running your SF-K games, keeping the same base TC, but increasing the increment, and seeing if SF's performance degrades (and thus becomes closer to what is expected).
(For examples of this, see ccrl 40/40 vs 40/4 (K tops the first, SF the second), and cegt 40/40 vs 40/4 (once again, K tops the first, SF the second.)
PS., I tried to email you about this via your websites tool, but it seems you didn't receive the message?
beram wrote:Could be because of your used SALC openingsbook. Perhaps latest SFdevs could be improved especially in the field of opposite castling situations.
Just my wild guess.
I doubt that. If the SALC-set is the reason, then why only against Houdini and Komodo a very big gain and against Gull and Equinox not?!? And in my singlecore Stockfish-testruns I use SALC-positions, too. So why is there no big Elo-gain against Komodo and Houdini?
Really strange...
And in the meantime, Stockfish 141112 climbed again over 67% score in my Endless RoundRobin...So it doesnt seem, that the result gets "more normal" with more games. Next week, I will update the Endless RoundRobin on my website.
Stay tuned!
Stefan
Looking over various rating lists, it seems that the SF vs K matchup may become easier for SF as the length of increment is decreased (measured as the ratio to the base TC); i.e., you can validate this claim by re-running your SF-K games, keeping the same base TC, but increasing the increment, and seeing if SF's performance degrades (and thus becomes closer to what is expected).
(For examples of this, see ccrl 40/40 vs 40/4 (K tops the first, SF the second), and cegt 40/40 vs 40/4 (once again, K tops the first, SF the second.)
PS., I tried to email you about this via your websites tool, but it seems you didn't receive the message?
oops, major typo/malfunction in that post. For examples of the increment relationship, http://fastgm.de/ is good. Also, I believe Larry Kaufman mentioned it in one of his posts here.
beram wrote:Could be because of your used SALC openingsbook. Perhaps latest SFdevs could be improved especially in the field of opposite castling situations.
Just my wild guess.
I doubt that. If the SALC-set is the reason, then why only against Houdini and Komodo a very big gain and against Gull and Equinox not?!? And in my singlecore Stockfish-testruns I use SALC-positions, too. So why is there no big Elo-gain against Komodo and Houdini?
Really strange...
And in the meantime, Stockfish 141112 climbed again over 67% score in my Endless RoundRobin...So it doesnt seem, that the result gets "more normal" with more games. Next week, I will update the Endless RoundRobin on my website.
Stay tuned!
Stefan
Looking over various rating lists, it seems that the SF vs K matchup may become easier for SF as the length of increment is decreased (measured as the ratio to the base TC); i.e., you can validate this claim by re-running your SF-K games, keeping the same base TC, but increasing the increment, and seeing if SF's performance degrades (and thus becomes closer to what is expected).
(For examples of this, see ccrl 40/40 vs 40/4 (K tops the first, SF the second), and cegt 40/40 vs 40/4 (once again, K tops the first, SF the second.)
PS., I tried to email you about this via your websites tool, but it seems you didn't receive the message?
The score of Komodo against Stockfish 140928 was very good (50.7% !), so I doubt, that the small Fischerbonus is really a problem for Komodo 8 in my Endless RoundRobin-tournament. Against Stockfish 141112 Komodo scores not good, but this Stockfish is incredible strong and it scores very good against all opponents.
