Author: Uri Blass
Date: 01:06:51 07/13/04
Go up one level in this thread
On July 13, 2004 at 02:54:02, John Hamlen wrote: >Hi Uri, > >Many thanks for the kind words for Woodpusher. > >I was quite shocked when Mark Winands contacted me just before the tournament >and asked if I could bring Woodpusher in order to make an even number of >competitors. (I was originally entered in just the Go tournament). "But I >haven't touched the program since 1997", I protested. "Don't worry", he said, >"you might not come last". > >Surprisingly, he was right, and it turned out to be a very interesting >experiment. I'm planning to jot down a few of my impressions about how things >have moved on in the intervening 7 years in a note for the ICGA Journal and/or >Eric Hallsworth's excellent Selective Search Newsletter. > >You are right, the only modifications were: 1) a large negative contempt factor >of 1 pawn as I went to Tel-Aviv with the target of snatching a draw in one of >the games, and 2) A larger transposition table, because of the much faster >hardware (2.8GHz vs. 166MHz). If my memory is correct then PII300Mhz was the hardware of WCCC1997 so maybe another point is that woodpusher had slower hardware in that tournament. Another possibility is that woodpusher earns more from time relative to crazybishop because of things like better order of moves. It may be interesting if you can test woodpusher in some test suites and the interesting question is not number of solution but progress. If based on test suites 5 seconds of woodpusher is equivalent to 3 seconds of crazybishop(same number of solutions) when 50 seconds of woodpusher are equivalent to 70 seconds of crazybishop then it suggests that the search of woodpusher is better. I think that it can be interesting if someone compare engines not based on speed but based on progress when for every engine 1 unit of time is the time that it needs to solve half of the positions in the test and the table give results after 10 units of time for every engine. Uri
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.