There was an interesting thread about material values etc, initiated by Larry Kaufman. Rether than prolonging it, I'd like to ask two related questions here
- what is an estimated playing strenght increase gained by moving from single bad trade penalty and a couple of conditional clauses to the current imbalance table?
- has it been tried to use different imbalance tables for endgames with a small number of pawns (i.e. at post 2 per side)?
I ask this because during a massive retuning of Glass evaluation function, trying to use material imbalance table (slightly modified for minor pieces worth 325 centipawns) I saw a couple of funny scores in the endgame, like a plus for KRN vs KRPP (the old version calling it a draw, the new - an advantage). Or are Crafty's passed pawn values alone capable to offset the imbalance bonus in such a case?
regards,
Pawel Koziol
Crafty material imbalance question
Moderator: Ras
-
- Posts: 908
- Joined: Mon Jan 15, 2007 11:23 am
- Location: Warsza
-
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: Crafty material imbalance question
In Crafty? Practically nothing. But it is a simpler approach, and you can fine-tune it on a case-by-case basis rather than having a complex set of conditional tests that we used before changing. It is a tiny bit faster, but significantly easier to understand.PK wrote:There was an interesting thread about material values etc, initiated by Larry Kaufman. Rether than prolonging it, I'd like to ask two related questions here
- what is an estimated playing strenght increase gained by moving from single bad trade penalty and a couple of conditional clauses to the current imbalance table?
Not by me yet, no.
- has it been tried to use different imbalance tables for endgames with a small number of pawns (i.e. at post 2 per side)?
Crafty handles that in a separate way. It knows that the KRN can never be worth more than the KR, since that is a dead draw. So it would prefer the two pawns instead of the knight, knowing that preserves winning chances. Material imbalance would give a penalty for being a knight down. Most likely, since the two pawns are passed, passed pawn scoring would call them more valuable than the knight.I ask this because during a massive retuning of Glass evaluation function, trying to use material imbalance table (slightly modified for minor pieces worth 325 centipawns) I saw a couple of funny scores in the endgame, like a plus for KRN vs KRPP (the old version calling it a draw, the new - an advantage). Or are Crafty's passed pawn values alone capable to offset the imbalance bonus in such a case?
regards,
Pawel Koziol