Author: Martin Giepmans
Date: 10:15:33 07/06/04
Go up one level in this thread
On July 06, 2004 at 12:38:14, Andrew Dados wrote: >On July 06, 2004 at 12:13:31, Stuart Cracraft wrote: > >>I'm thinking a list of features for chess >>programs to gauge effort vs. return to save >>implementation time would be useful, so that >>programmers won't get lost in a maze >>"full of twisty little passages" ... >> >>In the table below, low numbers in effort and return >>columns below correspond to lesser effort and greater >>return. The lower their product, the better >> >>If anyone wants to add to or revise this, please >>respond. This would serve as a guide to >>practical implementation since time is limited. >> >>I am also looking for additional features with >>your effort/return estimate. >> >>Thanks --Stuart >> >> FERV (Feature, Effort, Return, Value) >> >>EFFORT RETURN VALUE Feature >>1 1 1 capture ordering >>3 1 3 null move >>2 2 4 null move with verification >>4 2 6 search pv first >>5 1 5 static exchange evaluator >>4 1 4 transposition table >>3 2 6 transposition table with 2-tier replacement >>3 2 6 history heuristic, killers, other ordering >>2 3 6 aspiration >>2 2 4 iterative deepening >>2 2 4 pawn hashing w/ complex pawn evaluation >>2 3 9 capture extension >>1 1 1 check extension >>1 5 5 pawn to 6th/7th extension >>3 3 9 futility >>3 3 9 razoring >>5 3 15 mate-at-a-glance >> >>P.S. Return would imply things like reduced nodes, faster, >>stronger, etc. -- the usual. Practical results in real games... > > >capture extensions and razoring are at best dubious (eg crafty got rid of them). >Why such high values there? > >-Andrew- What I understand from Stuart is that better features have _lower_ values, not higher. This is a bit couter-intuitive, perhaps a word like 'priority' would be better than 'value'. For instance, mate-at-a glance has the lowest priority (15). Martin
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.