IGarcia wrote:
I have been reading in forums that critter 0.9 is probably the best tactical engine now. I think if this is true it will get a nice STS score.
...
Ignacio.
What a paradox ! An engine scoring well in a strategic testsuite because of tactics.
Matthias.
Well, It's certainly possible for an engine to improve the strength in both the strategy and tactics.
Most EPD's in STS have immediate short term advantage in strategy. It could be called tactical Strategy (or threatening moves).
Strategy has numerous elements. A lot of them are ordinary plans with no significant plus in evaluation -- they are ignored. Only those plans with definite plus in evaluation are recorded in STS.
I have tested Critter 0.90 in STS test. In 2 CPU / 64b / 10s it solved 1138/1300 - good result. For comparison Houdini 1174, Rybka 4 1141 and Stockfish 1107.
IGarcia wrote:
I have been reading in forums that critter 0.9 is probably the best tactical engine now. I think if this is true it will get a nice STS score.
...
Ignacio.
What a paradox ! An engine scoring well in a strategic testsuite because of tactics.
Matthias.
Yes, Critter its good in tactical and im expecting a good strategic result.
I am try to express is if the engine is getting better in one thing, probably its geting better in overal, because better optimal code, etc...
Maybe i do too much simplification on my post.
I really dislike to express my self in a language like "english". And mistakes on interpretation and expressions happens. Like the bugs in programas (not speaking on your GUI).
We have to live with this issues and try to get better every day.
Jouni wrote:Swaminathan, what's the reason to test with single cpu in Q6600
Jouni
It's because I have tested many engines in single CPU. It's for comparison purposes, besides I just wanted to analyze the data later - to see which specific EPD correlates with rating on single CPU, and how much it contributes as per measure of elo.
Jouni wrote:Swaminathan, what's the reason to test with single cpu in Q6600
Because time to fixed depth is practically a random number in SMP and you would need to run 20 times each test to average the values. Otherwise results are just meaningless.