Author: Uri Blass
Date: 10:08:11 06/01/03
Go up one level in this thread
On June 01, 2003 at 12:28:27, Georg v. Zimmermann wrote: >Hi, > >The first bughouse chess engine I know of was a modified crafty, but it was >buggy and did hang often. Then came "Skyshark" and "DeepBug", which play at >around 1800 level, and can take advise from its partners. >GCP's older Sjeng versions played bughouse, and did very well. Actually the name >of the program "Sjeng" is the reverse of the long time #1 human bughouse player, >"Gnejs". > >Some 3-4 years ago, I dont remember exactly, I began improving the DeepBug >program. I was very lucky to have picked a program to improve upon whos author >was extremely helpfull and which has very clear data structures. Even after >years of programming, and several complete redesigns of search and evaluation, >big changes to the move generation and innumerable bug fixes the underlying >structure is obviously the same. I called that "clone" Sunsetter. > >With the most recent Sunsetter versions I have abandoned bughouse support and >first concentrated on crazyhouse, and then for a few weeks on normal chess. >The main reason are this: >a) many strong bughouse players understandably dont enjoy playing engines , >because playing them has a very artificial feeling, therefore it is hard to find >good opponents , which is frustrating for an author >b) a really strong bughouse program would have to know what is going on on the >other board. Winbaord doesnt suport that. And I was too lazy to write >interfacing software to the servers just for that task of observing the partners >game and playing at the same time. I never thought that programs play that game without knowing what happens in the other board. I think that the biggest problem is how to use knowledge about what happen in the other board except times in a productive way. >c) and this is probably the most important point: Bughouse largely depends on >speed (I can explain if it isnt clear why so). It is pretty easy to make a >program fast , doubling the speed only costs you 50 raw "elo". By making a >program that takes only 100ms per move, I could create something very strong, >which would be even more artificial. I know that in a lot of games one player has a mate threat that cannot be stopped and the opponent tries to save the game by not playing and hoping that his partner is going to win. It is clear that this strategy can work only if the opponent of the parner has not enough time to decide to stop playing and lose on time. Inspite of that it seems to me surprising if you can do the program stronger by using 0.1 second per move because intuition tells me that the demage from getting bad position by playing very fast is bigger because in this case you will probably not get a position with unstoppable mate. Did you find that in comp-comp games the strategy of using 0.1 second per move is a winner? 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.