Author: Sune Fischer
Date: 04:44:13 05/27/02
Go up one level in this thread
On May 27, 2002 at 06:58:03, José Carlos wrote: >On May 27, 2002 at 06:14:38, Sune Fischer wrote: > >>On May 27, 2002 at 05:52:21, José Carlos wrote: >> >>> If I understood you right, you can simple put this on top of AlphaBeta() and >>>QSearch(): >>> >>> if (ply >= MAX_PLY) return Eval(); >>> >>> With 'ply' being the current depth, that you increment on every MakeMove() and >>>decrement on UnMakeMove(). >>> >>> José C. >> >>Yes, I guess that would work, technically. >>But having a check like that in the alpha-beta and the qsearch is all too often >>just a waste of time. > > Don't worry about that. That check is free compared to Eval(), MakeMove() or >any other thing you do during search. You can test it yourself, just measure >time with and without that check, and you'll see no difference. > >>I would like to solve the problem at the root if possible. >> >>Also I don't know what happens to the scores if you just start returning the >>evaluation with no regards to horizon effects. > > You gotta put the limit somewhere. At some point, you can't search further and >just rely on eval. Put that at 60 or 100 or 200 plies, it's a design choice. But >anyway, I wouldn't care much about horizon effect at 60 plies :) > >>I'd expect the result returned from a 45 ply search without these cuts to be >>more exact than a 46 ply search that does the cutting. >> >>-S. > > I don't understand this. At some point, you must return Eval() (or Alpha or >Beta). I don't know what you mean by "without these cuts". If you enter search() and return eval() there, then you might have a queen en prise. It seems better both with respect to speed and accuracy to deal with this problem at the root, if possible. After all, scores probably don't change much from a 45 ply search to a 60 ply search (theoreticly), but the strange scores you get around ply 60 by cutting may change the scores all the way up the tree, I don't know. Perhaps the best would be to completely remove the depth limit and just let it go as deep as it can in the given timeframe? -S. > José C.
This page took 0.01 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.