Author: Enrique Irazoqui
Date: 21:35:24 12/10/99
Go up one level in this thread
On December 10, 1999 at 23:29:28, Chuck wrote: >I would like to make a statement which supports Bob Hyatt's position that most >ratings given for computer chess programs are inflated, including recent >releases on the SSDF list. > >I have noticed that programs go through a cycle, which is obvious and easy to >see. Now, an unmodified program and hardware should play at a very consistent >level over long periods of time (really, for as long as it works). > >But when a new program is released, it is better and faster than most of its' >competition, thus, it's rating gets inflated. As it gets older, it is surpassed >by more and more of its' competitors and it starts getting pounded, whereby it's >rating takes a pounding, too. > >This DOES happen on SSDF. The best example is easily the Mach IV. Bertil Eklund >claims this to not be true, but, the Mach IV had a SSDF rating of 2282 in >January 1993 but now has a SSDF rating of 2074(?)!! Thats a 208-point drop!! > >Obviously, the Mach IV was not modified, so this is clear evidence of the cycle >I mentioned. > >Now, I am grateful for the work SSDF does, and Bertil, and I respect them for >doing a thankless job. The SSDF team is, I believe, fair, and they are a >valuable resource. But this rating cycle does exist and I don't understand why >Bertil would want to cover up the truth. And what if this is not the truth? I can't imagine Bertil covering up anything. Maybe what happens with this alleged inflation in the SSDF list is that you are as prone as I am to reach conclusions too quickly. This pattern you mention may or may not apply to Mach IV, depending on the number of games played when listed with the 2280 rating, but it doesn't show in the cases of Rebel 8 and Mchess 6 I mentioned in another post. I also checked the rating of F5 since it first appeared in February 98: 2589 (198 games), 2573 (already 794 games), 2574, 2574, 2570, 2566, 2567, 2564, 2565, 2566, 2561. It seems pretty consistent to me. Now, why would this pattern you mention apply to Mach IV and not to Rebel 8, Mchess 6 and Fritz 5? It seems to make sense to think that there is no such pattern. Enrique >Regards, > >Chuck
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.