Author: Miguel A. Ballicora
Date: 22:06:17 01/07/01
My program (it does not have a name yet) beat TSCP too! at the end of the message there is the game which looks interesting. Few weeks ago I saw a message of a new programmer celebrating that his program beat TSCP. Since I was trying to develop one myself I figured that it was a great idea to follow that lead. So, I made my experimental engine Winboard compatible and here I am ready to try it. It took me a while, to find a way to administered the time properly but it look like I got it, for now... I made it play several blitzes against TSCP and it look like it was in the same league, I was very happy! at least it won several games. The first semi-long game it defeated TSCP! I am very surprised, it plays just ugly but much better than I would have predicted. The evaluation is very thin and it is also slow (20-50 knps in the middlegame). There is tons of room for improvement and things to add. Briefly, it sucks :-) but it was a lot of fun to have reached this stage. So far it has - Aspiration search with iterative deepening, - Bitboards for move generation. - Captures generated first. MV(captured?) LVA (acronym correct?) - History table (no killers yet) - Null move, - Hash table for refutation and transposition (32 MB in this game) - Extensions in check, recaptures. Evaluation considers - Material, - A very clumsy king safety (safer in the corners) - King better is the center when few pieces are in the board. - Passed pawns bonus and also bonus for attacking the squares ahead of them. - Double pawn and pawn island penalties. - Mobility bonus for all the pieces. - Penalty if the Queen has few squares to move and is outside my own pawn chain (it worked in this game, detecting that the opponent queen was in toruble) The game it played against TSCP is interesting. It looks like my program likes closed positions. I am particularly interested in the positions after 33. Ba6 and 34. Qd2. What do your programs think? Here is the game briefly annotated. Thanks to all the people that posted here and 3 years ago in r.g.c.c. from whom I extracted all the knowledge, lurking. [Event "Computer chess game"] [Site "DEFAULT"] [Date "2001.01.07"] [Round "?"] [White "no name yet"] [Black "TSCP 1.71"] [Result "1-0"] [ECO "D01"] [PlyCount "117"] 1. d4 d5 2. Nc3 Nf6 3. Bg5 h6 4. Bxf6 exf6 5. a3 Be7 6. e3 O-O 7. g3 Be6 8. Nf3 Re8 9. h4 {This is ugly, just trying probably to gain space and increase the mobility of the bishop. TSCP does not realize that it will end up with a bad bishop, though.} 9... Bd6 10. Nd2 f5 11. f4 {Ugly, but at least makes sense. My program plays perfect anticomputer chess :-) :-) got a blockaded position and also a good bishop.} 11... Nd7 12. Bh3 c6 13. O-O Re7 14. Kh1 Qb6 15. Qc1 Rae8 16. Rf3 {Uglier, but looking at the game served the purpose to over protect e3 to free the knights to relocated them in better squares. The rook in f3 cannot be attacked for now} 16... Nf6 17. Nb3 Ne4 18. Qe1 a5 19. Na4 Qb5 20. Nc3 Qa6 21. Bf1 Qa7 22. Nc1 Qb6 23. Nd3 Bd7 24. Nd1 Qa7 25. Bh3 Qa6 26. Kh2 26... Qc4 {Terrible position for computers.} 27. Rc1 b6 28. Ne5 Bxe5 29. Bf1 Qa4 30. fxe5 c5 31. Nc3 Nxc3 32. Qxc3 Rc8 33. Ba6 {This move is very risky, but it was tactically justified under the point of view of the computer. It had a slight positive score. It seems like it is tempting for black to play later c4 but... it looked incorrect. Anyway, Ba6 looks weird but it worked against TSCP.} 33... Rc7 34. Qd2 34... c4 { TSCP falls for this, now both programs think that they are ahead!} 35. b3 Qxa3 36. bxc4 dxc4 37. Rb1 {not easy to defend b6, the position looks quite complex} 37... b5 38. Rff1 {Black queen is in trouble} 38... Be6 39. Rxb5 { with the rook??? it surprised me but it looks tactically correct. Is it?} 39... c3 40. Qd3 $5 {if Bc4/ Rb8+! the threat is d5 and Rxb6 prevented a blockade} 40... Re8 41. d5 Bd7 42. d6 Ra7 43. Rd5 {Now magically Qd3 was useful to defend the ugly bishop on a6. Not enough, the Rd5 is not very pretty but... very active} 43... Rb8 44. Bc4 44... Bc6 {Where to go with the rook now?} 45. Qd4 $1 {to c5, white managed to finally accomodate the pieces} 45... Bxd5 46. Qxa7 Ra8 47. Qc7 Bxc4 48. Qxc4 {Black is in trouble and now TSCP sees it.} 48... Qb2 49. d7 {I would have played Qd3 but...} 49... Qxc2+ 50. Kh3 Rf8 51. e6 51... g5 {TSCP gets desperate, now it is over} 52. hxg5 f4 53. Rxf4 Qd1 54. exf7+ Kh7 55. Qe4+ Kg7 56. gxh6+ Kxh6 57. Rf6+ Kg7 58. Qg6+ Kh8 59. Qh6# { White mates} 1-0
This page took 0.04 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.