Author: James Robertson
Date: 08:59:40 06/08/99
Go up one level in this thread
On June 08, 1999 at 09:30:11, Andreas Stabel wrote:
When my program finally beat me, my brother said I had released a Frankenstein,
a monster I could not control.... :)
Anyway, the way to get your program to play sensible moves very very easily is
with piece/square tables.... it will revolutionize your program's style.
James
> I'm not the worlds best chess player (ANST rated 1200-1300 on FICS), but I
>thought I knew something at least.
>
> Now I'm in serious doubt. As a professional computer programmer I of course
>couldn't keep my hands off trying to make a chess playing program. I don't
>have much time for this, so I've been on and off for many years.
>
> Now finally I've at least managed to make a program which can do the
>following: Play all legal chess moves, knows of mates and
>stalemates and has a simple evaluating function which just calculates
>material and says that a pawn is better the more advanced it is. THAT'S ALL !
>
> The search algortithm is also very simple. I think it is some sort of
>alpha-beta though I seem to only have an alpha :)
>I have no hash table, no opening book, no endgame table bases, no learning,
>a simple quiecense search and an even simpler move ordering which takes winning
>captures first (where the hitting piece is of less value than the one captured)
>and then the other moves in the sequence they are generated.
>
> After finding a lot of bugs and misunderstandings, I now have got the
>program working. I have played test games against it all the time to see
>if it did what I expected it to do, but it has been very easy to win.
>
> But now everything has changed. I have LOST the two last games I played
>against it !!! (see PGN at the end)
> This shouldn't be possible. The program plays like an idiot. It just
>pushes pawns whenever possible and messes up it's king safety and the
>positions looks terrible for the machine but still it WINS :(
>
> Of course I haven't played on its weakness. I could have played safely,
>just slowly exploiting the bad position and protecting my pieces, but I've
>never liked playing this way. I love attacking and trying crazy (and mostly
>loosing) combinations and this has of course backfired.
>
> But the big point here is that the program knows so little and gets into
>so bad positions that I thought I would win easily anyway. This has turned out
>not to be true. The program ususally only sees 3-4 plies before the
>quiecense search, but this is more than enough to find winning tactics
>against me as the two games clearly show. It seems that the ability to look
>deeper than what I do enables the program to escape the messy posisitons it
>gets into and as soon as I overlook something it jumps on me. All my chess
>knowlegde beaten by a couple of plies. I'm depressed :(
>
> My only consolation is that when I analyse the game using crafty, most of
>my moves look reasonable and the score stays at about even or + for me most
>of the time - untill the tactics turn against me.
>
> There is of course another way of looking at this. To imagine that the
>best chess players in the world beat the most powerful computers with big
>opening books, learning, endgame table bases and as much chess understanding
>as Mr. Hyatt and the rest has been able to put into them through decades.
>This is very impressive and tells us how powerful the human brain is.
>Pity my brain isn't that impressive :)
>I think I finally have gotten a practical proof of the distance between
>these masters and me.
>
> If there are somebody else who have experienced similar setbacks I hope
>this report can show that you are not alone.
>
> Where to go from here ? Tic tac toe perhaps :)
>
>P.S. Here are the games in PGN:
>
>[Event "Test of move program"]
>[Site "ANST PC"]
>[Date "1999.06.07"]
>[Round "-"]
>[White "Move"]
>[Black "ANST"]
>[Result "1-0"]
>
>1. h4 d5 2. f4 Nf6 3. d4 b6 4. b4 e6 5. c3 c5 6. Qa4+ Bd7 7. b5 a6 8. e3
>Qc8 9. h5 Qb7 10. dxc5 axb5 11. c6 Qxc6 12. Qd1 Na6 13. g4 Ne4 14. Qc2 Rc8
>15. a4 bxa4 16. Bxa6 Rc7 17. Bb2 Bb4 18. Ne2 b5 19. Qd3 Nc5 20. Qxb5 Nxa6
>21. Qxc6 Rxc6 22. cxb4 Nxb4 23. Rxa4 Nc2+ 24. Kd2 O-O 25. Ra7 Rd8 26. Bc3
>Nxe3 27. Kxe3 Bc8 28. g5 Rcd6 29. Be5 Rc6 30. Rc1 Rb6 31. Ra8 Rb3+ 32. Kf2
>Rxb1 33. Rcxc8 f6 34. Rxd8+ Kf7 35. Rd7#
>1-0
>
>[Event "Test of move program"]
>[Site "ANST PC"]
>[Date "1999.06.08"]
>[Round "-"]
>[White "ANST"]
>[Black "Move"]
>[Result "0-1"]
>
>1. e4 d5 2. e5 h5 3. d4 f5 4. b3 a5 5. c4 e6 6. Nf3 Bb4+ 7. Nbd2 Bc3 8. Rb1
>dxc4 9. bxc4 Bxd4 10. Qa4+ Bd7 11. Qb3 Bb6 12. c5 Bxc5 13. Qxb7 Nc6 14. Bb5
>Ra7 15. Bxc6 Rxb7 16. Bxb7 a4 17. O-O g5 18. Nc4 g4 19. Bg5 Qb8 20. Nh4 a3
>21. Rfd1 Qa7 22. Rd2 Qa4 23. Ne3 f4 24. Nc2 Qc4 25. Rbd1 Ba4 26. Rd8+ Kf7
>27. Rc8 Qxc2 28. Rxc7+ Ne7 29. Rf1 Bxf2+ 30. Rxf2 Qxc7 31. Ba6 Qxe5 32.
>Bxe7 Qa5 33. Bxa3 Qxa6 34. Rxf4+ Ke8 35. Ng6 Qb6+ 36. Kf1 Qb1+ 37. Kf2 Qxg6
>38. Rxa4 Qc2+ 39. Kg3 h4+ 40. Kxg4 Qf5#
>{Black mates} 0-1
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.