Author: Robert Hyatt
Date: 13:04:51 12/24/97
Go up one level in this thread
On December 24, 1997 at 11:56:01, Don Dailey wrote: >On December 24, 1997 at 04:28:16, Amir Ban wrote: > >>On December 23, 1997 at 18:12:25, Robert Hyatt wrote: >> >>> >>>and in the endgame, extend on passed pawn pushes to other ranks than >>>just the 7th, since they are a threat to run at any time... >> >> >>Small tip for those who didn't think of it themselves: Don't extend on >>advancing to 3rd rank. A pawn is as far from queening on the 3rd rank as >>on the 2nd. >> >>Amir > >Hi Amir, > >In my experience with passed pawn issues, it is dangerous >to extend most passed pawn pushes except for the extreme >cases, where a pawn is on the 7th and maybe other advanced ranks if you >constrain it correctly. Like any other extension that may seem to >make great sense, they can blow up the search in certain positions. >Keep in mind that MOST pushes of passed pawns (in the context of a >search) are blunders. Your program will look great in a few tactical >positions where your extension really helps you nail it but in most >cases all you will be left with is a general slowdown. > I would amend the above to say "most pushes of passed pawns, *until you reach the endgame* are blunders." If you don't extend once material gets low, you can fall for lots of passed pawn tactics... and if you do it carefully, it doesn't have to be expensive. IE only extend safe passed pawn pushes (as verified by the SEE routine)... only extend when the opponent has few pieces left... and so forth... >But that is not to say there are not good passed pawn extension >algorithms. They just have to make some sense and be limited in >highly useful ways. For instance, you wouldn't extend a pawn to >5th rank on a node that is close to the leaf nodes of your search, >the chances of seeing something useful are close to nil. this depends. IE in some endings, advancing a pawn one rank is enough to win. But it can take *many* plies to reach a position where the pawn can move at all. But your eval has to realize that advancing one rank is a big deal in the right circumstances. > >The real point with extensions is not to try to identify good moves >and extend them but to identify lines of play that will return >meaningful information. Usually these turn out to be moves that are >likely to be the best move but this is not the real point of them. > >A case in point is check extensions, most programs extend all >checking moves, and it's certainly not because they are usually >good moves because they are only occasionally good. But checks >are usually just annoying delaying moves. They are important >to extend even if they are completely irrelevant because they >steal 2 ply from your tactical awareness. So even bad checks, >or unimportant checks help return meaningful information. > >But passed pawn pushes are very important with regard to selectivity. >Null move selectivity with serious depth reduction can be blind >to moves of passed pawns. A very useful improvement to your >chess program might be to handle some of these cases statically. >You should probably confine yourself to the advanced cases and >don't try to cover the cases that null move pruning is comfortable >with. You can calculate how many ply it will take to promote a >pawn and if it's less than or equal to the remaining depth you can >consider it a threat. If there are pieces in the path of the pawn >you should increment your expected queening depth appropriately. >If you do this well your program can remain quite selective without >being stupid with regard to passed pawns. > >Now you can try to cover this stuff with passed pawn extensions >too but in my opinion these are 2 separate issues that shouldn't >be confused. I don't personally believe it is correct to cover >a deficiency of the selectivity with extensions, the deficiency >is still there and the blowup of the extensions are still there. > >- Don > > > > > > > >- Don
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.