Author: Robert Hyatt
Date: 08:13:36 01/25/01
Go up one level in this thread
On January 25, 2001 at 05:34:35, Gian-Carlo Pascutto wrote: >On January 24, 2001 at 23:39:12, Robert Hyatt wrote: > >>pushing a passed pawn had hardly any effect, but that might be a >>result of the test suite rather than a useless extension. The default >>value for this is also .75 >> >>recapture is the most interesting case. It actually appears that removing >>this _completely_ gave the best score, which is surprising since >>win at chess is a tactical suite. Lowering this extension obviously >>speeds up the overall search. I currently use .75 here as well. >> >>The most interesting thing is the recapture extension. I have used it forever, >>and took the idea from Cray Blitz since I used it there as well. I am not yet >>sure why it actually hurts since it seems (to me) to be a logical idea. I am >>going to test this more, just for my own interest. > >I have hesitated for a long time to add this extension for the following >reason: you should use extensions to reduce the horizon effect, or on >_really_ forced moves. Neither pushpp nor recapture do that when used >like you do. > >Always extending checks, threats and onereply doesn't hurt much because >they are forced anyway. For recaptures and pushpp that just isn't true. > >Recapturing happens often on different squares. Transpositions of moves >fool the extension. Recapturing is a normal move. If you extend them you >are blowing up your search with a _very_ small chance to find anything >interesting. > >IMHO recapture extensions should only be done when close to the leaves, >in order to prevent capture/recapture sequences from pushing a threat >away. But when there is enough depth left, I see no reason to extend. > >Try your recapture extensions with varying trigger depth also. I think >the results will be interesting at least. > >-- >GCP I disagree to an extent. IE if I do a 12 ply search and see that I lose a pawn then I can insert BxN NxB and eat two plies of search and "hide" that pawn loss by using BxN as a "horizon-pusher" move. The recapture partially eliminates the ability of a capture to hide something deeper in the tree by pushing it over the horizon and out of the universe. Junior does something similar with the way it counts plies differently depending on whether the move is a capture or non-capture. So it _does_ work reasonably for some. Perhaps the aggressive null-move usage in Crafty is the big difference here...
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.