Author: Enrique Irazoqui
Date: 13:27:50 09/27/99
Go up one level in this thread
On September 27, 1999 at 15:10:06, Ratko V Tomic wrote: >> Overwhelming evidence of fraud and corruption. The probabilities that >> you or anyone had to be conceived and born are almost infinitesimal. >> Therefore, and according to your statistics , you probably don't exist. > >Not exactly the same. Nothing is exactly the same as something else, this is obvious, but what your theory and my joke have in common is that neither makes sense. If you look at the 6 best rated programs on a P200MMX, you will notice that 5 of them (F532, Hiarcs, N99, F5 and Junior) are CB programs. The exception is CM6K, that must be played manually. It is only natural that the highest rated programs are tested first on new platforms, like the SSDF did. This, I think, explains things clearly enough, at least to me, without resorting to your statistics and other theories. I don't think it is fair to blame the SSDF for reflecting the situation of the market. Nimzo 7.32 (also from CB) and Tiger will end up high in their list. Meanwhile, there is no commercial program that in comp-comp can touch these 6 I mentioned before. Enrique > If you give some general properties, such as >"only programs from CB on top hardware" than it is perfectly legitimate >to compute in how many ways such properties are satisfied compared to all >possible combinations. It is conceptually no different than estimating >the odds of "4 dice giving 6" etc. While generally it would not have >occured to me to even look at the odds here, there was an ongoing >controversy with SSDF and their pro-CB bias, so much so that they have >a response to the accusations on their web page. > >The fact that, just in time for the Christmas shopping season, they hand-picked >4 CB products to be at the top of their chart (by virtue of putting them, >and only them, on top hardware) only adds another bit to the existent picture. >If they had only limited number of fast machines, there are so many other >ways to arrange tests to be even-handed and fair to all other programs, >that no amount of the rationalizations on their web page can explain it. > >In addition to this latest pick and the earlier controversy about the >proprietary CB autoplayer, one can also question their allowing their >testers to mostly overlook the problems with killer books, opponent >recognition, and other similar tricks. Since such gimmicks are well known, >there should be in place a set of effective and strict measures to detect >such and disqualify offending programs, or at least disqualify their >results achieved by such means (unrelated to program's chess playing >quality/strength, as would be perceived by an end user).
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.