Author: Paul Richards
Date: 13:25:58 03/27/99
Go up one level in this thread
[snip] >The question: "which program is better" is somewhat ambiguous. I'm going to throw my 2 cents in...with all respect to Dr. Hyatt, my experience with Fritz vs. Crafty also concludes that Fritz is better in terms of evaluating a position. This is easy to test 'fairly', independent of hardware considerations or compiled speed. I also think it is the basis for Kasparov's comments regarding the relative strength of the programs, though I know Dr. Hyatt's general feeling toward Kasparov. ;) Personality aside he does know chess though. ;) The test is very simple. Choose a position of interest and analyze it without the benefit of opening books. Analyze the position to a good fixed depth, say 15 ply. Look at the recommendations given by each program. This is easy to do by following main line opening variations where the evaluation of a position is generally well established. Left to fend without an opening book, my experience is that Fritz' and Junior's recommendations fall more in line with accepted theory than does Crafty. The conclusion is that regardless of the speed of reaching a given depth on whatever hardware, the evaluation of a position at a fixed depth is an intrinsic factor of the code and not its speed or efficiency. King safety, space, tactical considerations, pawn structure, etc., all taken together. Different programs will evaluate a position differently at the same depth, and will give different numbers and recommend different moves. The questions are why, and which move is the better one? Programs such as Fritz and Junior, based on what they will recommend for a given position, seem to have an overall grasp of these factors which is slightly more in line with what human masters and GMs recommend, thus the comments from Kasparov and others regarding their strength. In this way Junior is somewhat 'better' than Crafty. The fact that the programs are fast on typical PC hardware adds to the strength and the perception of strength, but the real strength appears to be the overall evaluation and knowledge. Have Crafty and Fritz analyze your pet lines without a book and see the difference in evaluations and recommendations for yourself, and make your own conclusion. This is the easiest way to see how each program really 'plays'. A Kasparov can certainly see whether a program recommends strategically sound moves or not, and with what frequency, and so can lesser players. This is independent of whether Crafty can defeat one of them running on a 1200 processor machine, the issue is how the program implements chess knowledge. Crafty already has book learning, tablebase access, SMP ability and everything else it needs from a perspective of 'technical' hardware-dependent strength. IMVHO Dr. Hyatt would do best to concentrate on its fundamental evaluations, then it will really become a monster. No offence intended whatsoever of course, only trying to give fair and objective commentary.
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.