search guidance

[jarkkop]

Is there a way the make the search keep the advantage (a passed pawn ) and then try to add advantage in the other side of the board. Sometimes it seems that eval lets the promising passer to be sacrificed of other meaningless so-called advantage. Can this conversion from feature to another be somehow be restricted/controlled until it is enough to be accepted ?

[Ron Murawski]

Is there a way the make the search keep the advantage (a passed pawn ) and then try to add advantage in the other side of the board. Sometimes it seems that eval lets the promising passer to be sacrificed of other meaningless so-called advantage. Can this conversion from feature to another be somehow be restricted/controlled until it is enough to be accepted ?

I don't know how to do this in the search, but in the eval you can try reducing bonuses for "other meaningless things" when a side has a passed pawn.

Ron

[hgm]

It seems to me this is caused because the passed Pawn is not valued high enough in the first place.

[jarkkop]

I think that not all passed pawns are so important. I would think that one pawn you can promote is enough for a win, but e.g. Fruit and Crafty in the most extreme case if you have 8 passers in 4th row.They are all as valuable. One would think that last 8th is not as important anymore. I would be concentrating more into making a queen from the most advanced.

The reason you can change a parameter to another and ELO doesn't change at all is that search can find many ways to combine the eval terms to make the same score? Another proof that chess is a draw if played without major errors.

My guess why Rybka is so good is that it somehow succeeds to guide the game in sharp imbalance which is not a clear draw anymore and find a consistent way keep the game the near the draw cliff from which many opponent fall because they don't see that is not a draw anymore.
That seemed to be case with up till R232 , but R3 must have some kind of singular extension because it homes so quickly very tough tactical solutions. All speculations thou just by watching the evals swing for Rybka most of the time in the end.

[Ryan Benitez]

I think that not all passed pawns are so important. I would think that one pawn you can promote is enough for a win, but e.g. Fruit and Crafty in the most extreme case if you have 8 passers in 4th row.They are all as valuable. One would think that last 8th is not as important anymore. I would be concentrating more into making a queen from the most advanced.

Connected passers are a very good thing and should get a hefty bonus. Also I always pick out the passer I consider most likely to become a queen and focus on that one with more factors than the rest. Many simple things like don't get your rook stuck in front of your best passer, don't let the opp get his rook behind your passer, and also material quantity over material quality when pushing the pawn with a few exceptions.