Author: David Fotland
Date: 13:34:25 01/15/98
Go up one level in this thread
No actually, someone suggested I write a chess program the Tuesday before last Thanksgiving and I started coding the next day. I probably spent about 3 days on the win-95 user interface, and I've worked on it every evening for an hour or two before bed, and maybe 5-10 hours each Saturday and Sunday. I was discouraged by a 5th place finish in the Computer Go World championship, with 3 very close games against the programs that finished 2ns, 3rd, and 4th, so I was ready to work on a different project. The rating I gave was Blitz, since all test games against people so far are at 30 moves in 15 minute time limits. I'm sure it would be much weaker at tournament time controls. But I'm an experienced strategy game developer, and I've coded alpha-beta and transposition tables several times before for other games. I wrote The Many Faces of Go, one of the strongest computer go programs, and a very strong Domineering program. I didn't use any code from my other programs, since I wanted this to be all in C++, and others are in ANSI C. rec.games.computer.chess and this forum gave me lots of good ideas so I didn't waste a lot of time trying things that don't work well. I don't know what you mean by operational, but it played its first game of chess with all rules except 50 move and repetition draw before the end of thanksgiving weekend. It didn't have a transposition table, and it was only about 5K nps on my Pentium-166, so it only looked about 5 ply deep, but it could already beat me. By Christmas I added the transposition table (Thompson style), and some basic move sorting. Between christmas and new years I profiled it a few times and got the speed up to about 35k nps on the Pentium-166. Last week I added killers. Next week I will implement a SEE to cut the Q search size way down, and put in pawn structure caching for the evaluation function, and I hope to get over 50k nps on the Pentium, and search 7 ply in less than 500k nodes most of the time. I'm still finding bugs of course. The most recent, amusing one, is that I forgot to say castling was illegal if a rook was captured. I only unset the castle-legal flag if the king or rook moved. So in a game, a rook was captured, then the program castled to that side anyway, getting its rook back :) I can't spend much more time on this program, since I have to get back to working on Computer Go. The next big competition is in August... But it's been a lot of fun working on something where I can see the program improve from day to day, rather than the very slow improvements you get with a more mature game program. David Fotland David On January 14, 1998 at 20:13:53, Willie Wood wrote: > >On January 14, 1998 at 14:02:22, David Fotland wrote: > >>I looked, and didn't see "Blitzen", so that's the name for my chess >>program. >>Please let me know if someone else is already using it. It's fast, but >>stupid, hence the name. 8 weeks into the project, it's now rated about >>2000. >> >>David >> >> > > >Uh, 8 weeks? I can barely tune my eval effectively in 8 weeks. Oh, you >must mean that's it's been operational for 8 weeks, and it took at least >a year to write. > >I wouldn't use the name Blitzen, that's the name of the ICC interface. >If you want a name that's "fast and stupid," how about Clinton? > >WW
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.