Author: Terry Presgrove
Date: 14:34:55 04/08/99
Go up one level in this thread
On April 07, 1999 at 17:07:22, Jeremiah Penery wrote: >On April 07, 1999 at 16:32:18, James T. Walker wrote: > >>For anyone interested in Crafty engines. The Crafty 16.6 has now played 58 >>games against Fritz 5.32 and Junior 5.0. The score stands as: >> >>Junior 5.0 wins 16 to 14 >>Fritz 5.32 wins 16 to 12 >> >>This may not be "Stastically significant" but it's heading in that direction. >>These games were played at game/60. Crafty used a Pentium II 333 Mhz/128 meg >>and Fritz/Junior used a K6-2/350 Mhz with 128/64 meg hash tables. Most games >>were played with Crafty using 96 meg hash,8 meg hasp,10 meg cache. Fritz used >>128 meg hash and Fritz used 128 meg with a few games with only 64 meg hash. >> >>I have also played a large number of games at game/10. Crafty is not as >>impressive at this time control but has given Junior 5 a hard time. Scores: >>Crafty 16.6 wins Comet B00 60.5 to 39.5 >>Junior 5.0 wins Crafty 16.6 71.5 to 62.5 >>Nimzo 99 wins Crafty 16.6 52 to 30 >>Fritz 5.32 wins Crafty 16.6 93 to 39 >>This indicates that Crafty is only 25 points behind Junior but 150 points behind >>Fritz. I wish Bob were here to explain why Crafty now plays Junior 5.0 so good >>now but no improvement shown with games against Nimzo/Fritz(At Blitz). The >>game/60 results indicate Crafty may play better at the long time controls and >>should have a chance at time controls closer to tournament conditions. >>Also I am wondering if this indicates a change in Bob's approach of tuning >>Crafty to play better comp. vs comp. games vice Crafty vs humans which has >>seemingly been his goal in the past. Any comments on the above info ?? >>Jim Walker > >One reason Crafty may do better at slower games (and one thing that's always >bugged me a bit, even if for no real reason :) is that almost all (if not ALL) >of his testing is done on his super-fast Quad Xeon machine. If Crafty plays >better on his machine (at whatever time control), it is analogous to Crafty >playing better on others' machines at a much longer time control. In fast >games, I've seen almost no improvement since 15.20, and I haven't played enough >slow games to notice any difference. The version that seems to do best on my >machine [PII-300] on any time control is a slightly modified 15.20 that I have. >It almost always beats 16.6, even against things like Dual PII-350 machines. >These are just my observations, so please do not take this as definite truth. :) I've been playing crafty on ICC for about a year now and my tests show version 16.6 to be much stronger not only at slower time controls but bullet and blitz also. Using Dan Corbits MS C++ compiled version Crafty v16.6 won a series of matches against 16.5 up to and including 3 hour matches. The ratio of wins was pretty consistant v16.6 defeating v16.5 13.5 to 6.5 in 3 hour matches. Also in 3 0, 5 0, 30 0 matches the ratio was averaging 14 to 6 in favor of v16.6. Running on my k6-2 350 w/1meg cache. TP > >>P.S. I wish Bob was here. > >I suspect he's just taking a break for a while, and that he'll come back sooner >or later. :) > >Jeremiah
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.