Computer Chess Club Archives


Search

Terms

Messages

Subject: PASSED_PAWN_PUSH extension scheme (and SmarThink)

Author: Sergei S. Markoff

Date: 10:39:07 09/18/02


Hello all!

It's surprise for me that PASSED_PAWN_PUSH extension is off in SmarThink since
0.00a!!! Anyway SmarThink is playing OK!? Is it means that PASSED_PAWN_PUSH
extension is not such imortant or it means than SmarThink is really much
stronger than now?

It is also interesting term to discuss: how PPP-extension must work. I think
that Crafty's (and also SmarThink's and others scheme) is not quite well. I
think that it is possible to enlarge search depth not only for pushing pawn to
7th (2th) rank but also for other pushing when result of SEE for this move is
not pawn losing... But I'm don't know the optimal extension value. May be it's
about 1/2 ply for 6th rank and 1/4 ply for 5th rank?

PPP-extension must resolve two type of problems:

1) we can't see the losing of pushed pawn
2) we can't see the promotion

But the probability of existing (1) and (2) are the different because of some
properties of eval. function. "Evaluate" _want_ to push pawn. And that's why (1)
is more important than (2).

Losing of our passed pawn for engine is much terrible than no promotion. Because
of losing game is worse than not winning.

I think that it's ideas can be used in creating good non-Crafty scheme of
PPP-extension.

Sincerelly yours,
Sergei



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.