Author: James Robertson
Date: 09:59:10 04/12/00
Go up one level in this thread
On April 12, 2000 at 12:19:09, Graham Laight wrote: >Thanks for a good, helpful post. > >Tony's calibration page is at http://home.interact.se/~w100107/level.htm > >I welsome the SSDF's openness in publishing this data. Unfortunately, I don't >think that it provides strong enough evidence for the year 2000. The following >claims can reasonably be made: > >* Humans have improved against computers in the last 10 years > >* Even allowing for the above, it is possible that computer "rating inflation" >has occured since 1990, in which case the current SSDF ratings could possibly be >too high now > >This is why we need to look at the evidence of the Israeli league. > >However, thanks again to Chris and Tony for being open with the "older" >evidence. I think that the older machines are accurately rated by the SSDF (they were used to calibrate the SSDF, after all). A computer gains 60-100 points for every hardware doubling when it plays other computers. There seems to be a lot of evidence this is not true when playing humans, though. As a result, it is impossible to predict with any certainty how much better a program will fare if you give it twice the hardware. The 232 point difference between Fritz 6's actual TPR and the SSDF rating is very strong evidence of this. James > >-g > >On April 12, 2000 at 11:28:41, Chris Carson wrote: > >>Based on the data, SSDF is the most valid and most reliable source >>of Computer ratings. The SSDF ratings have held up well against >>humans. See the SSDF list and Tony's page for a comparison of the >>SSDF with the Human Calibration games and the Micro vs FIDE human >>ratings. All are 40/2. I will post a consolidated data analysis >>later, when I have time. :) >> >>You should expect performance ratings above and below SSDF. This is >>what I see when I look at the performance ratings. :) >> >>Best Regards, >>Chris Carson
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.