Author: Robert Hyatt
Date: 12:35:42 01/28/99
Go up one level in this thread
On January 28, 1999 at 09:35:18, ramon b. de robles wrote: >On January 28, 1999 at 07:21:31, Djordje Vidanovic wrote: > >>On January 28, 1999 at 06:09:14, Ulrich Tuerke wrote: >> >>>On January 28, 1999 at 05:26:41, ramon b. de robles wrote: >>> >>>>On January 27, 1999 at 21:22:47, Djordje Vidanovic wrote: >>>> >> >>Uli, hi. I don't think that it was Crafty that went down 20-0. I believe that >>it was Bionic that lost :) That's why the question dealt with the possibility of >>Bionic playing better under Winboard... >> >>Perhaps I am wrong -- have no idea until the question is rephrased. >> >>Anyway, I have no idea what went wrong. These two programs (or versions of one >>and the same program) are very close in strength, with Bionic being definitely >>better positionally and Crafty tactically. And yes, I don't think that any >>program can beat Crafty 20-0 at blitz, yet... At lightning (bullet) I know of >>one that is close to it now :)) >> >>Regards, >>Djordje > >You are right Djordje, it is Bionic who lost by 20 - 0. Anyway, the hash table >is 16 MB. I repeat the test and use 32 MB and the score is 8-2 in favor of >Crafty 16.3 (Bionic with two(2) draws and one(1) win. I matched it with Little >Goliath and the score is 5 - in favor of Little Goliath. Of course, I used the >Fritz 5.32 interface. > >Regards, > >ramon This sounds really wrong. IE I can't imagine beating any program with any program 20-0, unless the loser is really a new program with lots of bugs, a fairly basic search, and a very rudimentary evaluation. And even then a program like that ought to win a game here and there. 20-0 is a huge margin of winning. I'd check the 'bionic' log file (assuming they left that in) to see if it is searching some ugly number of nodes per second (like 100 or something) and if so, then try to find out what is going on... I'm not sure the 'winboard interface' you are using is really well-thought-out yet, and programs can be sensitive to a lot of nonsense like forcing it to read the entire move list in before each move. As a program doing a lot of stuff at the root might be tricked into doing that for every move, and end up spending its allowed time pre-processing and not searching. needs examination...
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.