Author: Dann Corbit
Date: 16:12:54 09/14/99
Go up one level in this thread
On September 14, 1999 at 17:21:43, alfred palang wrote: >this might be a matter of choice.but not all programs play decently..not all. a >lot of people sight the outcomes of ssdf testing but i think they are using >obsolete equipment. The equipment does not matter. Their results would be just as valid on Pentium 90's as on 32 CPU EV-7's. The SSDF is (quite frankly) the best and most statistically sound chess program testing organization in the world. There have been some attempts at alternatives. As far as I can see, all of them have failed miserably. It should be obvious that the SSDF at least does their math correctly. I understand that there is some controversy about autoplayer learning by some programs and the like. In any case, the results are valid for the contests and configurations involved. I think (also) that those crying about missing results have a point but that it is ludicrous to suggest that the games for which the PGN scores are missing are faked. Since we are after the fact you can either ignore those results or accept them at face value. In summary, SSDF is the closest thing to true science for the testing of chess programs that we have right now for machine verses machine contests. There is also some interesting work for Winboard programs that can be found at: http://f11.parsimony.net/forum16635/ and related sites. However, the conditions between contesting programs and the hardware in use is not nearly so well spelled out and controlled, nor are the mathematics as complete or accurate.
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.