for the interested fast compiles are online

[mainsworthy]

[Event "Computer Chess Game"]
[Site "MARKS-PC"]
[Date "2010.04.16"]
[Round "-"]
[White "Chad's Chess V015"]
[Black "SuperMainsworthy96f"]
[Result "0-1"]
[TimeControl "40/300"]

1. b4 e6 2. Ba3 Nc6 3. Bb2 b5 4. e3 Ba6 5. a3 e5 6. Ra2 Bd6 7. Bd3 f5 8.
Bxf5 Be7 9. d4 exd4 10. Qg4 Kf8 11. Bxd7 Nf6 12. Qh3 Qxd7 13. Qxd7 Nxd7 14.
exd4 Nf6 15. Ba1 Kf7 16. Nc3 Bd6 17. Nge2 Kg6 18. d5 Ne7 19. f4 Nf5 20. h3
Ne3 21. g3 Rhe8 22. g4 h6 23. f5+ Kh7 24. Nb1 Nfxd5 25. Kf2 Bb7 26. Re1 Bc8
27. Ng3 a6 28. c3 Nxg4+ 29. hxg4 Bxg3+ 30. Kxg3 Rxe1 31. Rb2 Rg1+ 32. Kf3
Nf6 33. g5 Rxg5 34. c4 Rxf5+ 35. Kg2 Rg5+ 36. Kf2 Rf5+ 37. Ke2 Re5+ 38. Kf3
Rf5+ 39. Ke2 Re5+ 40. Kd3 bxc4+ 41. Kxc4 Be6+ 42. Kd4 Re1 43. Kd3 Rd8+ 44.
Kc2 Kg6 45. a4 Kg5 46. b5 Bf5+ 47. Kc3 Rxb1 48. Rg2+ Kh4 49. Rh2+ Kg3 50.
Ra2 axb5 51. Bb2 Be6 52. Ra3 Kf4 53. axb5 Bf5 54. Ra2 g5 55. Ra5 Nd5+ 56.
Kb3 Kg4 57. Ra4+ Kh3 58. Ra2 Bd3 59. Ra5 Kg4 60. Ka3 Bc2 61. Bg7 h5 62. Ka2
Re1 63. Ra7 Rb8 64. Ra5 Rb6 65. Bf8 Rf1 66. Bc5 Rb8 67. Bd4 Bb1+ 68. Kb3
Bd3 69. b6 c5 70. Bxc5 Rb1+ 71. Ka2 Nc3+ 72. Ka3 Ra1+ 73. Kb3 Rxa5 74. Bd6
Rxb6+ 75. Kxc3 Rxd6 76. Kd2 Bf5+ 77. Kc3 Ra7 78. Kb4 Re6 79. Kc4 Rd7 80.
Kb5 Rc7 81. Ka4 Bd3 82. Kb4 Ra6 83. Kb3 Be2 84. Kb4 Kf4 85. Kb3 Rb6+ 86.
Ka4 Ra7#
{Xboard adjudication: Checkmate} 0-1




WOW!! 96f just beat chads chess in winboard, thankyou so so much HG! I owe you a thankyou from my engine, which has stopped blundering

[beachknight]

Hi Mark,

I have not tried 96f that seems much stronger yet.
But a prev. version:

Code: Select all

[Event "CET2436"]
[Site "CET24_Antalya_TUR"]
[Date "2010.04.06"]
[Round "1"]
[White "Slibo 0.51 JA w"]
[Black "Mainsworthy S14 w"]
[Result "1-0"]
[ECO "D43"]
[WhiteElo "2391"]
[BlackElo "2048"]
[PlyCount "67"]
[EventDate "2010.03.08"]
[EventType "tourn"]

1. d4 d5 2. c4 dxc4 3. e3 Nc6 4. Bxc4 Bf5 5. Qb3 Be6 6. d5 Ne5 7. dxe6 Nxc4 8.
exf7+ Kxf7 9. Qxc4+ Ke8 10. Qb5+ Qd7 11. Qxb7 Rd8 12. Nc3 c6 13. Qxd7+ Kxd7 14.
Nf3 Nf6 15. Ne5+ Ke6 16. Nxc6 Rd6 17. Nxa7 Ra6 18. Nab5 Kd7 19. e4 h5 20. Bf4
e6 21. O-O-O+ Nd5 22. exd5 Bc5 23. dxe6+ Kxe6 24. Nc7+ Kf5 25. Nxa6 Bxf2 26. g3
Rc8 27. Rhf1 Ba7 28. Bb8+ Bf2 29. Rxf2+ Kg4 30. Be5 g5 31. Rd4+ Kh3 32. Nc7 Rb8
33. Ne4 Rxb2 34. Nxg5# 1-0

Please look at the fifth move: 5. ... Be6 ??

Best,

PS: ECO Code generated by Arena [D43] should read [D20].

[Sven]

HG , Ive updated Super to 96f, this update is because of your pointers, I think Its working ok now! thanks HG! you seem to know whats going-on before anyone else me included, it was a simple random factor that was being mangled, i think its ok now time will tell, soon be on Leos website i hope.

Mark

I tried 96f, it plays the same 4.Nxf7 blunder.

If you send me your code I can try if it is possible for me to have a brief look at it. Contact me by PM if you like.

Sven

[Sven]

Code: Select all

[Event "Computer Chess Game"]
[Site "Sven"]
[Date "2010.04.16"]
[Round "-"]
[White "SuperMainsworthy96f"]
[Black "Sven"]
[Result "*"]
[TimeControl "40/300"]

1. e4 c5 2. Nf3 Nc6 3. Bb5 a6 4. Bxc6 bxc6 5. Nc3 d6 6. h3 Nf6 7. O-O g6 8.
e5 Nd5 9. Nxd5 cxd5 10. d3 Bg7 11. Be3 d4 12. Bd2 dxe5 13. g4 O-O 14. c4 f5
15. g5 e4 16. Ne1 e5 17. Qb3 e3 18. fxe3 Qxg5+ 19. Kh1 dxe3 20. Bc3 f4 21.
Qb6 Bxh3 22. Rg1 Qh4 23. Nf3 Qf2 24. Rg5 Qxf3+ 25. Kh2 Rf5 26. Qe6+ Kh8 27.
Rxg6 hxg6 28. Qxg6 Rh5 29. Kg1 Qf2+ 30. Kh1 Bg2#
*

After the last move which mates by giving double check, 96f sends the previous move Kh1 again which is illegal.

Extract from winboard.debug:

Code: Select all

232026 >first : f3f2
232479 <first : move g1h1
machine move 58, castling = -1 0 -1 -1 0 -1
move to parse: g1h1
TC string = '+40/5'
mps=40 tc=300000 inc=0
move: g1h1
, parse: Kh1 (
)
silence
moveType 4 = 33, promochar = 0
TC string = '+40/5'
mps=40 tc=300000 inc=0
move: h3g2
, parse: Bg2# (
)
time odds: 1 1
Interrupting first
237541 >first : time 27156
237541 >first : otim 10548
book hit = (NULL)
237541 >first : h3g2
237588 <first : move g1h1
machine move 60, castling = -1 0 -1 -1 0 -1
move to parse: g1h1
GameEnds(40, xboard exit, 2)
Interrupting first
296258 >first : result * {xboard exit}
296258 >first : force
296258 >first : quit

Sven

[mainsworthy]

Hi Sven

thanks for your time, I havnt replied as I was working over the weekend, and didnt check internet, so sorry about that!

I know 94f was a bad version Super96f seems to be better , and in some games will blunder a piece, Ive almost totaly rewrote the evaluator, now the current version is 130f the 'f' just means a fast compile, the 130 means save eval()


I have thought about giveing you the code to look at , but I think Im going to goit alone (sink or swim because I want it to be my work only, but thank you so much.) sorry glad hgm didnt look

Im constantly trying to increase the eval beyond its natural ply, wish success on some engines but fails to make progress on others, but then the opposite 130f seems to be copeing as an allrounder

Im just about to look at the games you posted, I think I already agree with your comments, but new version I think dosnt do to bad.

I would be happy to email you current version, but I think it may be online soon? Ive just downloaded 'suprise' thanks for the free engine to test on mine

Mark

[mainsworthy]

http://wbec-ridderkerk.nl/html/details1 ... orthy.html

147f is online,, it does quite well , the sacrifices, are a minimum, no knight bluders(sometimes! it thinks it will do a sacrifice, but it has almost dissappeared)

HG Ive put a variable in my syntax named hg after you

and thankyou Sven and Harun, it was your reports that put me on a better path.

what problems today I ponder

[hgm]

A good indication if you have bugs or not would be if you are able to beat micro-Max 1.6. This is a extremely basic engine, without hash table, without any pruning (no null move), using classical piece values (1,4,4,5,9) which are known to be not particularly good for computer Chess, not using a particularly fast algorithm (no piece list, looping over the board to search for its own pieces), weak move ordering (just the best move of the previous iteration first, the rest in move-generation order). When you systematically lose from it, it can only be due to bugs. A good way to hunt for them is play a game against it, look where errors are made (where uMax' score jumps up compared to the previous move), and then figure out why it did that move, rather than the one mucro-Max was expecting.

[mainsworthy]

HG thats an amazing engine, realy small code! thats realy realy good.. I am installing it now, but I never imagined a program like chess could be that small.

Ive FINALY I think found the Mainsworthy bug, everything seems to be makeing sence, the bug is fixed in version 153f ,I hope it will be on Leos site soon http://wbec-ridderkerk.nl/html/details1 ... orthy.html

Thanks to all of you, this has been a productive few days! Im now off to test some more Mark

[mainsworthy]

micro beat me by grabing little pawns untill it got me in trouble then finished me. I gave my eval() some positional knowledge that can over-ride the the basic algo 1,3,3,5,9,1000,but it can play badly sometimes(it is going to be a pleasure, setting up mesurses now the engine seems ok) but plays better on other game engines.

I normaly stop the engine to see what went wrong, I never actualy watched the meters before thanks for pointing that out,.

I think my main problems are due to ply depth, it beats engines like ACE simply because its less ply, but a deeper ply will always beat mainsworthy, I tried a 2 different algos in the same engine, that works to gain a little strength, deeper ply foe 30% of moves.

I realy thought Id never find the bug, but it may still show up maybe I think I find it but then it comes back. so we will see.

Mark

[Ovyron]

using classical piece values (1,4,4,5,9)

Interesting, did that prove better than 1,3,3,5,9?