Topic Rybka Support & Discussion / Rybka Discussion / repeating same moves - repeating same question
when Rybka finds that some moves share the first place in a given position (or have very similar scores) , she can choose the one to play "randomly".
I post this idea, because I couldn't get a satisfiying answer to my "repeated games in engine-engine test tournaments" question.
is the opening book only factor which prevents Rybka from always playing the same moves under same conditions ?
- - -
my mind needs a rest
I post this idea, because I couldn't get a satisfiying answer to my "repeated games in engine-engine test tournaments" question.
is the opening book only factor which prevents Rybka from always playing the same moves under same conditions ?
- - -
my mind needs a rest
I agree to randomize the choice into a certain interval (smaller than the overall %error of evaluation/Elo rating) to generate homogeneous test tournaments. We can also ponder the randomization by a normal curve to get more probabilities to have the choise more close to the better choise (I think it's difficult to share exactly the first place).
Actually, I already have a private capability to do this. Rybka can play against herself a few thousand games from the same position, and cover the tree of reasonable possibilities quite well.
If there is time, I will add this to the Rybka 2.3 release.
Vas
If there is time, I will add this to the Rybka 2.3 release.
Vas
Topic Rybka Support & Discussion / Rybka Discussion / repeating same moves - repeating same question
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill
