Author: Vincent Diepeveen
Date: 09:21:45 12/11/02
Go up one level in this thread
On December 11, 2002 at 10:46:08, Sune Fischer wrote: >On December 11, 2002 at 10:00:38, Johan Hutting wrote: > >>On December 11, 2002 at 07:44:43, Sune Fischer wrote: >> >>>Chess is primarily tactics, >> >>Vincent does have a point, to get into a position where you can use those >>tactics you require positional knowledge. Any 1600+ rated player can tell you >>that :) > >Vincent was arguing a different point. I agree with him that it would be hard to >write a top level program with only 2000 lines of code, but I don't believe it >is impossible. > >I'm pretty sure there is still room for improvement in the area of search. >And probably evaluation is secondary to search in 95% of the cases. First of all you write 'code' i talk about code for evaluation function here. Not about 'code'. Secondly my evaluation with simple alfabeta and simple nullmove and a simple move generator (my move generator is pretty simple and stripped off actually like 40 lines including en passant and castlign) and a simple move ordering, so that the search is perhaps a 200 lines or so in itself and a simple recapturing qsearch. Yes even a static qsearch of 20 lines. That very inefficient search with diep's evaluation will of course kick the hell out of the best search in the world with movei evaluation. Especially if both are without book. Best regards, Vincent >>My guess is Uri uses a very simple patzer eval with big bonusses for good pieces >>along with a fast search. So far it makes movei play good, but if you wish to >>improve you'll run into a wall. > >I think you can write a lot of important eval in 2000 lines. >The basic pawn structure and king safety issues can be done in 2000 lines. > >>>if you have near _optimal_ extensions and near >>>_optimal_ pruning rules, then you could probably get those extra plies on your >>>opponent and could win WCCC using tactics. >> >>but, with a bigger eval and good pruning you will still search the promising >>lines deeper, right? > >Of course, but you also see eval hurting at times, no rules without excepton as >they say. For instance the trapped bishop codes are often wrong, Bxa2 b3 a5! and >maybe we have time to free the bishop here or can simply protect it with Ra8 or >Qb2... > >Not a whole lot of things are easy to staticly evaluate, only simple patterns >like isolated double pawns can be relatively safely scored, IMO. > >>>My guess is that Fritz would still be mighty strong even if they had to limit >>>their eval to 2000 lines. >> >>It would be outplayed by more positional programs and perhaps score some 'lucky' >>wins. But that's in 'slow' tempo games, blitz is another matter. > >With basic eval I wouldn't fear any program that I outsearched two plies in the >midgame, endgame is another matter as the tactics is often few and far between. > >IIRC the author of Ruffian said the secret was in the extensions and pruning >rules. :) > >-S.
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.