Author: Terry Presgrove
Date: 17:30:57 04/30/99
Go up one level in this thread
On April 30, 1999 at 11:52:54, James T. Walker wrote: >On April 29, 1999 at 21:38:35, Terry Presgrove wrote: > ><snip> > >>#games 50 20 50 50 22 22 20 >> >>v16.6 27.5 11.5 24.5 21* 11 11 13.5 >> >>v15.20 22.5** 8.5 25.5 29 11 11 6.5 >> >>I haven't had a chance to examine all the data as am currently >>running 40/2 and have been using notepad to look at the scores. >>Apologies in advance for any inadvertant errors......testing is >>very time consuming and difficult to be sure that the data is >>solid and not corrupt. >> >>* lost one game on time in which it appears to have locked up after only 22 >>moves >> >>** lost 11 games on time many of which should have been drawn >> >>TP > >Hello TP, >Your numbers are not significant if taken individually by time but I tried to >add them together in my head and came out with about 119-114. This tells me not >much difference between the two. I played 300 games of Crafty 16.6 vs (Crafty >16.6 w/tablebases)(3&5& some 5man). In the three 100 game matches(all blitz) >Crafty with tablebases won the first 50 games while Crafty w/o tablebases won >the second 50 games. The final score was 152-148 for the Crafty with >tablebases. The point being that even 50 games at a certain time control is not >enough to week out the best program especially if they are fairly evenly >matched. >Jim Walker Jim, While your point is well taken I'm not sure there is that much of an elo difference between 16.6 with or without tablebases. That is at least my opinion ....maybe 20 points tops. It seems your data bears that out. I think your mixing apples and oranges when it comes to comparing 5 0 blitz games to 3 hour matches. I think some programs play much better at blitz then standard . While crafty may not play as well as voyager at blitz I doubt that voyager comes close to crafty at say 40/2. I feel you have to seperate chess for programs into time frame categories just like some humans play better at faster time controls so too programs. As to 50 games not being enough to determine which program is the strongest I'm not sure I agree. But not being that good at probabilty and statistics I will defer to others. TP
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.