Author: Andrew Williams
Date: 12:16:24 06/22/99
Go up one level in this thread
On June 22, 1999 at 14:46:16, James Swafford wrote: >On June 21, 1999 at 14:35:22, Andrew Williams wrote: > >>On June 21, 1999 at 13:53:48, James Robertson wrote: >> >>>My program does check, recapture, and pawn push extensions, but I have heard >>>that many programs do a lot more. I think Gromit said he does something like 10 >>>different extensions? >>> >>>Could someone please tell me what different kinds of extensions are sometimes >>>used? It would be cool to try them out on my program. >>> >>>James >> >>I do the above plus null move mate threat (when the null move search returns >>a -MATE score, I extend by a ply). I also have a single response to check >>extension, if there's only one way of getting out of check - but I don't >>allow too many of these in one search. I've also recently tried re-searching >>with one ply extra if I'm about to take a beta cutoff, but my null move search >>suggested I would lose a Piece if I "passed". This doesn't seem to be doing > >I'm not sure I understand this. If null_score-alpha>=some_piece_value, >&& you're about to fail high, you extend *instead* of returning beta? > That's right. I should have been clearer that this isn't my idea. Rather it comes from Chrilly Donninger's paper on null move: Null Move and Deep Search, C Donninger, ICCAJ vol 16 no 3, September 1993. But now I'm looking at the paper in detail, it looks like I've made a mistake in this because Donninger is using the *static* score for the position rather than the score returned by the search. However, I'm sure that there's a solid basis for doing it my way - I think that this extension provides a measure of protection for cases where I'm simply "putting off" some problem by interposing silly moves - but I'm going to test it by doing it "properly" to make sure. Incidentally, I think the idea was described in the way I've implemented it on CCC in the last couple of weeks. It was that which made me re-implement it. I can't remember the title of the thread or who wrote the message though. Sorry if this post is a bit confused and confusing :) Andrew Williams > > >>any harm (which makes a change) so I'm leaving it in at the moment. I'm messing >>about a lot with my extensions at the moment (not very successfully), so I'd >>also be interested in what others do. >> >>Also, there was a message just after the Rebel-Rohde match where Ed Schroeder >>gave a few hints as to what he does - might be worth looking at that for some >>very interesting ideas. >> >>Andrew Williams
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.