Author: Robert Hyatt
Date: 11:57:07 06/02/04
Go up one level in this thread
On June 02, 2004 at 14:09:40, Sune Fischer wrote: >On June 02, 2004 at 13:48:50, Robert Hyatt wrote: > >>On June 02, 2004 at 12:30:50, Sune Fischer wrote: >> >>>On June 02, 2004 at 11:49:48, Robert Hyatt wrote: >>> >>>>>Don't complain about it saying that's not how it meant to be played. >>>>>If you won't acknowledge that the user is boss at his end then don't release the >>>>>engine to him. >>>> >>>>That's a completely nonsensical approach to anything. I designed Crafty with >>>>one goal in mind, playing chess. Crafty is a "package". Its book. Its >>>>learning. Its search. Its evaluation. Its pondering. Breaking any of those >>>>makes little sense since it becomes "not crafty" at that instant. >>> >>>Ok that is your privilege, but take Mr. Smith here he is a very strong player >>>and in the need for good engine to help him analyse his games. >>> >>>Mr. Smith has never heard of the computer chess club and couldn't care less >>>about computer games. All Mr. Smith needs is an engine that will assist him in >>>his analysis. >>> >>>As it happens Mr. Smith asks me for advice, now, should I tell him that Crafty >>>can't do analysis because Crafty is a "chess playing system" and he _must_ do >>>analysis with ponder ON (whatever in the world that means) and he must enable >>>learning before beginning analysis (whatever effect that would have?)? >> >>Learning has no affect in analysis mode. ponder=on is fine becausee it also has >>no effect. IE crafty is designed sensibly and doesn't need the operator to say >>"you are analyzing a game so you can't ponder anything." It already knows that. >> It also knows that learning makes no sense when analyzing and it doesn't do >>that either... >> >>As I said, it is designed to be run "as is" and it does just fine that way, the >>operator does _not_ have to twiddle with anything. Annotating games? No >>problem. No learning or pondering required. No change to their settings >>required either as the program is written to do this stuff correctly. > >Mr. Smith has just asked me another question. >Now he want's to know how good Crafty is for use in analysis compared to other >engines. How are you going to answer that? Turn off learning and play matches? It gets killed due to bad book lines and you draw conclusions from that? Leave learning on and it doesn't get killed with bad book lines. Which _really_ shows the analysis capabilities better? > >Due to his special needs it seems that Mr. Smith is not interested in the way >learning influences Crafty's performance. > >He is troubled that the only way Crafty is allowed to be tested is with learning >on, he feels this would be distorting the picture too much for his intended use. > >What advice should I give to my dear friend Mr. Smith? > To study the problem a bit more. Then he'll realize it is a non-issue. No matter whether learn=on or learn=off, the book _will_ influence the match outcome and the wrong conclusion can easily be made. So learning is simply not something that should be considered as bad or good. Just "as a part of the complete engine". If "Mr. Smith" can't understand that he doesn't need to worry about the skill of the analysis engine. Any engine will be so much better than him that it simply won't matter. >-S.
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.