Author: Vincent Diepeveen
Date: 09:17:00 12/11/02
Go up one level in this thread
On December 11, 2002 at 04:42:41, Nicolas GUIBERT wrote: > >>it means your positional score for a single position >>cannot be big. in DIEP in a normal position it is pawns big. > >>crucial positions the king safety gets like 10 pawns penalty >>which takes care i don't need to see the mate but evaluate it >>already correctly as -10 (which gives such a fail low that >>no further search is lucky needed there). Even if you sacrafice >>then a piece it doesn't matter. > >>Seems you do not do this in buggy to me. >> >>How can you have more than 4000 lines of eval code without >>having this problem? > > >Well obviously if you have a lazy evaluation function that is far (10 pawns) >from the full evaluation function in many cases, then you cannot use it. also in draughts a good evaluation should take care that you search more efficient, so adding lazy evaluation in those few positions where sometimes it helps to lazy eval and in like 1% of the situations it doesn't, it should mean that you get a smaller cutoff, which means in short that you search less efficient, so you should waste more nodes searching to the same depth with lazy evaluation where the % overhead is bigger than the efficiency win for evaluating without lazy eval. If not then either your search is very buggy or your evaluation is. My guess is you should start fixing hashtable :) Best regards, Vincent >To use lazy evaluation, you need to be almost sure that the lazy evaluation >(when used) will not change the result of the alpha-beta and is therefore a >pretty good estimate of the full evaluation. In draughts, there is nothing like >King safety, and I very rarely have a position where positional stuffs are worth >more than a piece. The main exception to this is what you call passed pans in >Chess. Of course. You have to take this into account. > >In some positions, when the alpha-beta root score was close to one piece more, I >had some problems with lazy evaluation, the alpha-beta suddenly being worth 1 >exact piece. But that was easy to fix. I don't have this problem any more. > >>I can use lazy eval already hardly in napoleon. i still use it, but only >>a small % of the positions i use it. > > >>i partly agree and disagree. a lot of the knowledge you can get >>from others. but i agree in this sense that fixing knowledge brought >>by others is so time consuming that you simply either cannot do it >>or you are busy too long with a few patterns. > >Of course, you can add knowledge that other people teach you. But well, it is >always easier to communicate inside one single brain, than between two brains... >:o)))) There is always less loss of information in the process. And both parties >are better satisfied :o))))
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.