Author: Peter Fendrich
Date: 10:37:42 01/05/99
Go up one level in this thread
On January 05, 1999 at 08:56:08, Enrique Irazoqui wrote: >On January 05, 1999 at 01:39:51, Jouni Uski wrote: > >>Why is MCP8 playing much worse in SSDF? Is it playing worse against old >>programs or is 90 games simple too few? > >I don't know what is the explanation, but the performance of Mchess 8 in my >tournament reflects quite well what I think of it in terms of strength. The >difference between my tournament and the SSDF list may be due of course to the >relatively small number of games I play, but also to the fact that I play all >programs on identical platforms and with the maximum RAM they can allocate for >hashtables. I just looked at the games posted by the SSDF with both opponents >playing on P200MMX machines. These 376 games would give the following relative >ratings: > > #Games Elo >Hiarcs 6 84 141 >Fritz 5 336 106 >MCP 7.1 90 59 >Genius 5.0 92 19 >Shredder 2.0 80 -12 >Rebel 9 30 -38 >Comet-A.90 40 -281 > >As you can see, this has little to do with the SSDF list. I don't mean to say >that this proves anything at all. There are few games played by R9 and Comet, >many of the games played by the SSDF on equal P200MMX are not posted, etc. >Still, the difference is quite striking. I would think that the 95% confidence interval for H6, MCP, G5 and Shr is about +-65 ELO-points or so. For F5 it should be about +-40. With that in mind I think that the difference is not striking at all. In fact there is no difference outside the 95% interval. Furthermore, in a 95% interval, 1 of 20 programs is expected to be wrong. That is an ELO outside the margins. //Peter
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.