Author: Dana Turnmire
Date: 18:37:39 05/04/01
Go up one level in this thread
This test was designed for master level programs and not for anything weaker. Here is the article by IM Larry Kaufman as it appeared in CCR Volume 5 Number 1 (1994). "In previous issues we have presented test sets of problems designed to rate chess programs. The last set of 25 problems compiled for CCR was primarily composed of tactical problems, and so it tended to overrate "stupid, fast" programs relative to "smart, slow" ones. In this issue I am at last unveiling my first serious attempt to rate both the positional and tactical play of chess computers in one set. The challenge I faced was to find positional problems that were sufficiently difficult to challenge strong programs but yet clear enough so that few masters would dispute the solution. I decided that such problems are best found in the opening books, since it is usually safe to say that only one move is played in a given position by the masters, it is probably the right move. This test is composed of 25 positions taken from opening theory, from which only one move is generally played or recommended. Some tactical problems are included, but the majority of the problems are primarily of a positional nature." He goes on. "One weakness of this test is that it ignores the endgame. Perhaps in a future issue we will offer an endgame test, with the rating to be averaged in with this test. One strength of the test is that the openings from which it is drawn cover a wide range, fairly representative of actual play." "So far I have been quite impressed with the correlation between this test and Swedish or British rating lists, but more programs need to be tested to confirm its accuracy." I might add that I have also been impressed with the correlation. I have Fritz 5, Genius 2, Chessmaster 5000 and 8000. Each time I upgrade my computer I usually retest a couple of programs and the increase in strength seems right. The only exception so far is CM5000 rated 2648 on my AMDk6-2/500 machine while CM8000 went down in rating to only 2600 on the same machine. The personality CM6666Deep on CM8000 however went up to 2672 I used a selective search of 9 so I just renamed it CM9Deep. I wonder if he has ever revised this test.
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.