Author: Shep
Date: 04:49:38 07/26/99
Go up one level in this thread
On July 23, 1999 at 22:00:48, hgkjhg wrote: >somehow, the newest versions of crafty seem to be doing much worse on this test >than crafty 9.21, but i think that the newer versions are much better. Also >Little Goliath is doing good on the louget test even positionally but it isn't >so good in real games. Comet can't even find simple positional moves like Bh3, >Nd1, and Bb5, but it seems alright positionally in real games. So what's wrong >with the test? Nothing. :) One of the problems is that some authors tend to debug their programs using some of the standard test suites. LGG2 has even been said to have been tuned for LCT 2 and other known suites. However, except for a few exceptions, the programs doing best on LCT2 are also the best in general. Me, I am only using LCT2 to get a quick first impression of the playing strength of a new program. To get some real details, I run bigger, harder and more varied test suites such as Albillo, Aemis 1&2 and such (remember most programs can solve the bigger part of LCT2 in under 10 seconds per position already). --- Shep
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.