Author: Dave Gomboc
Date: 13:29:16 04/05/99
Go up one level in this thread
On April 01, 1999 at 13:11:29, Bruce Moreland wrote: > >On April 01, 1999 at 05:51:55, Inmann Werner wrote: > >>On March 31, 1999 at 16:11:27, Bruce Moreland wrote: >> >>> >>>On March 31, 1999 at 15:36:02, Will Singleton wrote: >>> >>>>I've seen similar situations. My opponent makes a mistake, then I move, he then >>>>moves the piece back where it was (unusual), and I now repeat instead of taking >>>>advantage of the better situation. This can occur only when I want the draw, of >>>>course, and usually it's not harmful. But it could be. >>>> >>>>Is this worth fooling with? >>> >>>I think it is good to fix, yes. I forgot to mention I've seen other cases as >>>well, the one that I used is just easiest to describe. >>> >>>bruce >> >>One new thought for this discussion. >>I changed from rep in 3 to rep in 2, and my program got a lot faster!(also the >>cache should recognize the repetition?) If I find repetition, I stop the search >>and return 0. 0 scores, I never put into Hashtables. >>On the other hand, you are true with your irregular examples. >>It is difficult to say, which thing is better. > >What I do is make a little table of positions that I don't want to repeat, and >incrementally update this throughout the search. > >And position found during the search is in this table (of course it's removed >from the table after the move is unmade). > >If I put every position that has actually occurred in the *game* in this table >as well, I would have something that returned on 2x repetition, and which would >exhibit the behavior I mentioned. > >But what I do is only include *game* position that have occurred *twice*. A >position that occurs *once* is not included in this table. And you can clear out this table after every non-reversible move, yes? >So, my program will repeat moves without going crazy expecting a draw, but will >understand all the appropriate 3x repetition cases, and will find draws in the >search more quickly, because if you repeat twice in the *search* you probably >really should score it as a draw. > >Of course, there are other problems now. My program will repeat positions a >lot, and if it's in an ending where progress is slow, it runs up against the >50-move rule. > >It also bugs the hell out of its opponents when it does this, but I'm not sure >if that's a good thing or a bad thing. It's a good thing. :) >bruce Dave Gomboc
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.