Author: Robert Hyatt
Date: 15:34:36 09/11/00
Go up one level in this thread
On September 11, 2000 at 09:25:36, Mogens Larsen wrote: >On September 11, 2000 at 09:04:01, Robert Hyatt wrote: > >>I'm not quite sure who you are talking about here, but I really don't see a >>lot of "excuse company" activity. I _always_ see a lot of "explanation" >>discussion. Because I am always interested in what is happening. > >I'm not talking about you as the author behind Crafty. Your interest is obvious. >But persons who can't wait to invent excuses or refer to ancient versions that >once upon a time performed quite well according to their own tests, or claiming >that the native version used is significantly weaker without a shred of proof. > There is _way_ too much superstition used nowadays. I see lots of this "version x.y-5 is better than version x.y". And the conclusion is based on the fact that x.y-5 won 85 and lost 15 on a chess server, while version x.y (on the same hardware) won 70 and lost 30. No real investigation into the opponents to see if they were equal... no investigation into the book lines played.. etc... Better "statistics" would prevent much of that... >>IE in the SOS/Crafty disaster with the SSDF guys, I was interested in what >>was going wrong. It later turned out to be just bad opening line choices for >>the most part, as halfway through the 'match' results started evening out after >>Crafty lost the first N games badly. > >That is a reasonable explanation, but not the one chosen by the majority as far >as I can recall. It was what I saw. The first dozen or two games were horrible results. Then things changed drastically. > >>You can go a long way only if you play a very long match. Because the learning >>has to find ways to circumvent the many hand-prepared lines that some programs >>have in their books. This isn't quick, but does make progress. slowly... > >As almost everything else... > >Mogens.
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.