Stockfish 1.7 vs Rybka 3 tests

[WuShock]

Small sample , maybe not significant , but maybe an interesting data point.

4'2" / i7 @ 3.75 / F11 GUI / Vista x64 / R3 cont=0 / klo openings / no egtb / 512 MB /

With 1 core each , and Ponder -on- , Stockfish wins :
53-47 , +30 , =46 , -24

With 3 cores each , and ponder -off- , Rybka wins :
54.5 - 45.5 , +31 , =47 , -22


Tom

[S.Taylor]

This means to me, that Rybka is stronger, as I'm not very interested in how many nano seconds it takes for a computer to make a move, but rather, what moves it can find......... as long as it is within an amount of time which i am prepared to wait for it (not over the board, of course, because any modern computer is strong enough to challenge ME OTB).

If i knew that a computer played perfect chess at 48 hours per move, i would get that, and nothing else! even if all other programs could beat it at a minute per game. Not only that, but i would gladly pay much more than a normal price, for it.
Even if any program could beat it at 10 hours, or less, per game, if it ranked as perfect, solved chess, at 48 hours per move, then I'll be coming for it on all fours!

[Uri Blass]

This means to me, that Rybka is stronger, as I'm not very interested in how many nano seconds it takes for a computer to make a move, but rather, what moves it can find......... as long as it is within an amount of time which i am prepared to wait for it (not over the board, of course, because any modern computer is strong enough to challenge ME OTB).

If i knew that a computer played perfect chess at 48 hours per move, i would get that, and nothing else! even if all other programs could beat it at a minute per game. Not only that, but i would gladly pay much more than a normal price, for it.
Even if any program could beat it at 10 hours, or less, per game, if it ranked as perfect, solved chess, at 48 hours per move, then I'll be coming for it on all fours!

I do not see how you get the consequence that rybka is stronger.

Not enough games and we also do not know about long time control based on the results.

[S.Taylor]

This means to me, that Rybka is stronger, as I'm not very interested in how many nano seconds it takes for a computer to make a move, but rather, what moves it can find......... as long as it is within an amount of time which i am prepared to wait for it (not over the board, of course, because any modern computer is strong enough to challenge ME OTB).

If i knew that a computer played perfect chess at 48 hours per move, i would get that, and nothing else! even if all other programs could beat it at a minute per game. Not only that, but i would gladly pay much more than a normal price, for it.
Even if any program could beat it at 10 hours, or less, per game, if it ranked as perfect, solved chess, at 48 hours per move, then I'll be coming for it on all fours!

I do not see how you get the consequence that rybka is stronger.

Not enough games and we also do not know about long time control based on the results.

Isn't 3 cores, instead of 1 core, like a longer time control?

[Uri Blass]

No 3 cores vs 1 core is not equivalent to longer time control.

It is possible that rybka is better with more processors when stockfish is better with the same number of processors with longer time control.

It is possible that for rybka 3 processors is equivalent to time control that is 2.4 times faster when for stockfish 3 prossesors are equivalent to time control that is 2 times faster.

ponder on is also not the same as ponder off and it is possible that stockfish is relatively better with ponder on and not with ponder off.

finally number of games is not enough to get conclusions and if you get result of
+30 , =46 , -24 or result of +31 , =47 , -22 then you need more games to know which program is better.



Uri