Author: Dave Gomboc
Date: 21:17:05 04/07/99
Go up one level in this thread
On April 07, 1999 at 17:20:07, Roberto Waldteufel wrote: >On April 07, 1999 at 16:27:10, Dave Gomboc wrote: > >>On April 07, 1999 at 12:10:20, Roberto Waldteufel wrote: >> >>>Hi all, >>> >>>I wonder how many here have programmed other games besides chess. Before I wrote >>>my chess program I cut my teeth on several games of gradually increasing >>>complexity until finally I felt ready to tackle chess. One of the most >>>interesting of these was checkers. In fact, I used bitboards for my first ever >>>attempt at checkers before knowing about their widespread use in chess, so when >>>I came to program chess I was naturally inclined to lean heavily in the >>>direction of bitboard representations of information. Recently I returned to my >>>old checkers program and rewrote it from scratch making use of many new things I >>>have learnt from programming chess, resulting in a strong program based on Aske >>>Plaat's MTDF algorithm. On 24 April there is to be a match between my program >>>and Nexus99, one of the top commercial checkers programs. I may even release my >>>own checkers program commercially in due course. >> >>>I think the same techniques that have proved themselves in computer chess are >>>applicable to several other games, such as Shogi, Go and of course checkers too. >>>I would be interested to hear if anyone else here has found the same to be true. >>>In particular, if anyone else has programmed checkers, it would be interesting >>>to "compare notes". >>> >>>Best wishes, >>>Roberto >> >>I haven't programmed checkers, but I talk virtually daily with someone who has. >>;-) Are the top commercial programs in the same league as Chinook yet? >> >>Dave Gomboc > >Hi Dave, > >The main things that separate Chinook from top PC programs are the hardware and >the endgame databases. Chinook has tablebases for all endings with 8 or fewer >pieces, whereas none of the commercials have more than 6-piece tablebases. >Furthermore, Chinook has gigabytes of RAM so as to hold all these tablebases in >memory rather than on disk. However, the gap is narrower than you might think. >After some "friendly" games between Chinook and Colossus, Jonathan Schaefer said >that he thought that the 8-piece endgame databases were all that really >separated them. Of course, in a really big match, like the famous encounters >with Marion Tinsley, Chinook runs on much more powerful (massively parallel) >hardware than normal, making about 40 times faster. If, for example, Chinook >were to play a match against Ron King (currant World Champion), then this >faster hardware would definitely be pressed into service. Remember that Jonathan hasn't really done TOO much with it since 1994. He does have an experimental version with some new stuff, but mainly he has been working on other things. Ron King would just get crushed by Chinook anyway, IMO that's not a match worth playing. Colossus is written by Martin Bryant, right? He helped Jonathan with opening book stuff for Chinook (also Norm Treloar did this earlier.) The program doesn't have gigs of RAM, but the supercomputer at the U of A does. And yes, I can confirm that the 8-piece databases fit into the 10 gigs. :-) If I remember correctly, Mark Brockington's Ph.D. work sped up Chinook by a factor of four again on parallel machines (compared to what it was already getting for speedup). Chinook would kick the crap out of anything that wasn't parallelized. The experimental version hasn't really been stress tested, but I'm sure it would be better than the version that played in the aborted match vs. Tinsley. The question really is, would it make a difference... at the level Chinook plays, the game is very drawish. I am confident that sooner or later, he will bother to strongly solve the game. It is already possible to prove that a few selected positions 5 ply deep are drawn. >Best wishes, >Roberto 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.