TalkChess.com

I think i read Ed saying not to search loosing catures in QS.
Does this work just as well ?
You would still sort by mvv/lva but stop when that is negative.
Any comments ?

lauriet wrote:I think i read Ed saying not to search loosing catures in QS.
Does this work just as well ?
You would still sort by mvv/lva but stop when that is negative.
Any comments ?

No, you may not want to search captures with negative SEE, but more valuable pieces capturing less valuable pieces should be explored, if the victim is undefended.

I tried QS with all captures till actual score is worst for engine than the score before start normal search.

I mean:

Without QS if maximum depth is n and in ply n a queen capture a pawn and there was no other capture in the pevious n-1 plies, the engine decides is a good move.
With my QS if at ply n a queen capture a pawn and there was no other capture in the pevious n-1 plies, it continues one more ply but if an opponent piece capturing the queen, it decides is not a good move to capture the pawn previously.

But I hope my conclusion that the performance was worst than without QS could be wrong!!!
Actual score with vs without is 18-7-0 in 300 40/4 games I started after a little adjust!!

Well QS means computing additional stuff. Each QS is a little search tree of its own, with the eval being called at the leaves. If you compute more stuff, of course that takes more time.

But in the end, the raw plies search depth don't matter; what matters is finding good moves.