Author: Pete Galati
Date: 09:21:39 04/25/00
Go up one level in this thread
On April 25, 2000 at 11:52:52, Peter Skinner wrote: >What is Crafty's strength with tablebases, and without? >Or any program for that matter? >Are tablebases really that needed for these engines that use them to perform >well? > >Only reason I ask is, I watch several games in a series between Little Goliath, >and Crafty 17.10, and Little Goliath played the endgame better than Crafty did >with tablebases? > >Why would this be? > >How much weaker would Crafty be let's say without them? > >Many engines play endgames perfectly without tablebases, and usually do a bit >better than ones that use them. They may not move instantly, but they certainly >get the job done.... This isn't really an answer actually, but in the dreaded single computer games between Crafty & another program, if I go and look through Crafty's logs, I'll almost always have several tablebase probes. And that's even thought I'm only useing 3 & 4 piece tablebases. But would Crafty do better without them? I can't back up what I say, but I think I've noticed less draws since starting to use the 3 & 4 pc TBs. Maybe a posible way to test it would be to play a game between Crafty and a program that's close enough to Crafty's strength to have a chance to have a chance, and _also_ able to start a game from an epd position (can LGol?) and run the same position, maybe from ten moves into the 1st game, over and over between the 2 programs, both with and without the use of the tablebases, to see which way resulted in the most wins. Does that make sense? Does it seem like a good way to test that? I'm not sure, I hadn't given it that much thought. The reason I suggest starting the games from a position is to keep the games following the same basic game plan, so maybe instead of saving the position from move 10, how about more towards the end of the game, say move 35, 40, 45??? Pete
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.