repetition draw bug

[sje]

Are there any compiled linux apps that I will play a random game and save to PGN format? I could run that and collect a big library of repetitions/50 move pgn's. If left to my own devices I'm gonna fire up Perl again :p

I might be able to help with sending some random game PGN data via email. How many games would you like?

[stevemulligan]

I'm guessing around 100k each of 3 & 50 move reps would be a good test. So about 4 million random games.

[sje]

I'm guessing around 100k each of 3 & 50 move reps would be a good test. So about 4 million random games.

That's several GB of data. Even if your email can handle it, mine can't.

Give me a little time to see what I can do.

[sje]

Some files with random PGN games can be found at:

https://public.me.com/chessnotation -> RandomGames

(It may take a while to synchronize.)

Random PGN average length: about 2 KB

The result data (1-0, 0-1, 1/2-1/2) are shown; not the reason. Let me know if you want the termination reason shown as well; that would take some time.

[stevemulligan]

Thx. I'm getting the 100k game file now.

Here is the python script I'm going to use to collect all the different endings. It requires pyChess to run.

http://pwnedthegameofchess.com/pgn/

[stevemulligan]

Let me know if you want the termination reason shown as well; that would take some time.

Ok I extracted the games from the 100k file and grouped them together by ending type. These are NOT verified by anyone yet.

http://pwnedthegameofchess.com/pgn/game_endings.zip (84MB)

Here are the counts that I got.

55872 insuf
15430 checkmates
6127 stalemates
19891 fifty move
2680 repetition

[sje]

Ok I extracted the games from the 100k file and grouped them together by ending type. These are NOT verified by anyone yet.

http://pwnedthegameofchess.com/pgn/game_endings.zip (84MB)

Here are the counts that I got.

55872 insuf
15430 checkmates
6127 stalemates
19891 fifty move
2680 repetition

These look very close to the expected percentages.

When I get a chance, I'll run my own scan. But it will take a little bit of coding first.

[sje]

A conflict with the checkmate count (your number 15430):

Code: Select all

gail:tmp sje$ grep "#" rg100000.pgn | wc -l
   15438

[sje]

And upon further examination, it looks like some of the Result tag values are wrong.

Code: Select all

gail:tmp sje$ grep "1-0" rg100000.pgn | grep Result |wc -l
    7716
gail:tmp sje$ grep "0-1" rg100000.pgn | grep Result | wc -l
    7716  (Wow!  An unlikely match)
gail:tmp sje$ grep "1/2-1/2" rg100000.pgn | grep Result | wc -l
   84568

The Result tags say 15,432 checkmates but the game termination markers say 15,438 checkmates. So apparently the Result tag guy is using a different, and likely faulty, scheme compared with the game termination guy.

[stevemulligan]

Some of the termination markers end with # 1/2-1/2 (none of the stalemates..).

eg, pyChess saw this as a 50 move game but the termination marker is "# 1/2-1/2"

Code: Select all

[Event "Random Game"]
[Site "gail"]
[Date "2011.10.06"]
[Round "1"]
[White "'-'"]
[Black "'-'"]
[Result "1/2-1/2"]
[Time "19:01:00"]
[PlyCount "283"]

1. d4 f6 2. Nh3 c5 3. Bf4 Na6 4. Nd2 Nc7 5. Bxc7 f5 6. g4 h6 7. a3 e5 8. dxe5
g5 9. Qb1 Bd6 10. Qc1 Bf8 11. e3 b5 12. Rg1 d5 13. c4 Qf6 14. b4 Qxe5 15. Bxe5
Ke7 16. Rh1 Kd8 17. Qc3 Ba6 18. cxb5 Ne7 19. Be2 Kd7 20. Qxc5 Bg7 21. Bd1 Nc8
22. Nb1 Rg8 23. Nf4 gxf4 24. Bf3 Rd8 25. Bxf4 Bf8 26. Bxh6 Nb6 27. Nd2 Bb7 28.
gxf5 Nc4 29. h4 Bg7 30. e4 Nb2 31. Nb1 Bc3+ 32. Bd2 Nc4 33. h5 Ne5 34. Qd6+ Ke8
35. Qxd5 Rdc8 36. Ke2 Ng6 37. Be3 Ba6 38. Bf4 Rc4 39. fxg6 Rd4 40. Rh4 Kf8 41.
Kf1 Rb8 42. Qd8+ Kg7 43. Rh1 Rb6 44. Be5+ Rf6 45. Rh2 Bc8 46. Rh1 Bh3+ 47. Bg2
Rc4 48. Qe8 Be6 49. Rh2 Rxb4 50. Qc6 Bxe5 51. Bh3 Rf5 52. f3 Bd5 53. Kg1 Bxh2+
54. Kg2 Bg8 55. Qd6 Rf7 56. a4 Rb3 57. Qxh2 Rbxf3 58. Na3 Rg3+ 59. Kh1 Rf6 60.
Rc1 Rd6 61. Bf1 Ba2 62. Bc4 Kf8 63. Rc2 Rdxg6 64. Bd5 Rxa3 65. Qf4+ Ke8 66. Qg5
Rxg5 67. Be6 Bxe6 68. h6 Kf7 69. Rc6 Bg4 70. Rc7+ Kf6 71. e5+ Kxe5 72. b6 Be6
73. Rf7 Rg7 74. Rf4 Rg1+ 75. Kxg1 a6 76. Kf1 Bf7 77. b7 Re3 78. Rf3 a5 79.
b8=B+ Kd5 80. Rf6 Bg6 81. Ra6 Re4 82. Ba7 Re6 83. Kf2 Re1 84. Rxg6 Rf1+ 85.
Kxf1 Ke5 86. Kf2 Kf4 87. Rf6+ Kg4 88. Ke2 Kh3 89. h7 Kg2 90. Bc5 Kh3 91. Ba7
Kg3 92. h8=R Kg4 93. Rb6 Kf5 94. Kd1 Ke5 95. Ra8 Ke4 96. Rab8 Kf3 97. Rg8 Ke3
98. Rc6+ Kf3 99. Rgg6 Ke4 100. Bb8 Kf5 101. Kc1 Ke4 102. Rg8 Kf3 103. Bf4 Ke2
104. Rc7 Kf1 105. Rb8 Ke1 106. Rc6 Kf2 107. Be5 Ke2 108. Rc7 Kf3 109. Rc2 Ke4
110. Rb6 Ke3 111. Rcb2 Kd3 112. Bg7 Ke4 113. Rh2 Kf4 114. Rbb2 Kf5 115. Kd1 Kf4
116. Rhf2+ Ke3 117. Rb5 Ke4 118. Rd2 Ke3 119. Bb2 Kf4 120. Kc2 Kf3 121. Kd1 Ke4
122. Re5+ Kf4 123. Rb5 Kg3 124. Bc1 Kf4 125. Rf2+ Kg3 126. Rd2 Kh4 127. Rb1 Kh5
128. Rd3 Kg6 129. Bb2 Kf5 130. Rd8 Kg5 131. Bf6+ Kg4 132. Rd4+ Kf3 133. Rb6 Kg3
134. Ke2 Kh2 135. Rb2 Kg2 136. Kd1+ Kg3 137. Kd2 Kh3 138. Bg7 Kh2 139. Rd3 Kh1
140. Kc1 Kg1 141. Rd6 Kf1 142. Rd1# 1/2-1/2

There are 6 games like this that were counted as a 50 move ending.

The other 2 missing checkmates were counted as insufficient material. I commented out lines 51-53 in Utils/logic.py from pyChess and they were counted as checkmates.