Mystery engine at CCC

[Chessqueen]

...

Probably next year, Torch might beat Stockfish ...

Next year, Torch might beat Stockfish, and maybe a bag of rice will fall over in China?

You forgot Bean and Fish to complete the meal

[Peter Berger]

So either you measure the ability to find critical moves in positions near the win/draw line, or you measure the ability to trick weak engines; take your pick!

This feels like a perfect summary of the situation to me.
It also has many implications, if you think about it some more.
While doing some research on the Hiarcs product, I came across a few ancient discussions. Marc Uniacke mentioned that at some point in time the German CSS journal was influential with its position tests. Professionals wanted to do good in them for commercial reasons. So they made their engines perform well in these tests.
Then there is also the notorious MChess- Rebel rivalry when book learning was new.
You mentioned how Benjamin was doing good in handicap games, better than Stockfish.
For all practical purposes, top engines are nearly equal right now in the classical setup when playing each other.
Will the engine that is best in positions near the win/draw line be able to reach these positions often enough to make this ability worthwhile in the classical setup? The answer may be yes (and I think right now it still is), but how would we know? An engine that could improve on its chances to win from not so clear positions may be more useful in the long run though.

[Chessqueen]

...

Probably next year, Torch might beat Stockfish ...

Next year, Torch might beat Stockfish, and maybe a bag of rice will fall over in China?

It might be closer than next January, here are some games that is giving us a hint that Torch can fry the Fish within a few months from NOW https://www.youtube.com/watch?v=SxynS1m_Xb0

[Peter Berger]

If you use a book with only "best" lines, the top engine ratings will depend primarily on which ones are better at avoiding draws against much weaker engines, which may basically mean which ones do "Contempt" better, since almost all the games will be drawn between the top ones

I agreed to this statement a little too fast. I have since thought about it some more.

The first obvious question for me was - so how does Dragon deal with this problem?

This was my first game:

[Event "Lang 120min+10sek"]
[Site "Berlin"]
[Date "2023.08.05"]
[Round "?"]
[White "Crafty 25.6"]
[Black "Dragon 3.1 by Komodo Chess 64-"]
[Result "0-1"]
[ECO "C67"]
[PlyCount "286"]
[TimeControl "7200+10"]

{4096MB, LAPTOP-NCDN8BTK} 1. e4 {[%emt 0:00:00]} e5 {[%eval 39,33] [%emt 0:04:
56]} 2. Nf3 {[%emt 0:00:06]} Nc6 {[%eval 32,33] [%emt 0:03:00]} 3. Bb5 {
[%emt 0:00:06]} Nf6 {[%eval 24,33] [%emt 0:02:02]} 4. O-O {[%emt 0:00:06]} Nxe4
{[%eval 12,36] [%emt 0:02:24]} 5. d4 {[%emt 0:00:06] (Te1)} Nd6 {[%eval 24,34]
[%emt 0:02:42]} 6. Bxc6 {[%emt 0:00:06]} dxc6 {[%eval 22,36] [%emt 0:01:53]} 7.
dxe5 {[%emt 0:00:07]} Nf5 {[%eval 16,37] [%emt 0:02:49]} 8. Qxd8+ {[%emt 0:00:
07]} Kxd8 {[%eval 15,24] [%emt 0:00:01]} 9. Nc3 {[%emt 0:00:06]} Be7 {[%eval
14,36] [%emt 0:06:23]} 10. Rd1+ {[%emt 0:00:07] (h3)} Ke8 {[%eval 16,36] [%emt
0:03:32]} 11. h3 {[%emt 0:00:08] (Lf4)} h5 {[%eval 13,36] [%emt 0:02:32]} 12.
g3 {[%emt 0:00:06] (Se2)} Be6 {[%eval 0,34] [%emt 0:02:34]} 13. Ng5 {[%emt 0:
00:17] (Lg5)} Rh6 {[%eval 0,39] [%emt 0:02:17]} 14. b3 {[%emt 0:04:17] (Sce4)}
Rg6 {[%eval -10,37] [%emt 0:02:45]} 15. Nce4 {[%emt 0:08:43] (Sf3)} Bc8 {
[%eval -26,34] [%emt 0:02:48]} 16. Bf4 {[%emt 0:08:18] (Kh2)} c5 {[%eval -35,
38] [%emt 0:03:49]} 17. Nf3 {[%emt 0:03:17] (c3)} Nd4 {[%eval -38,36] [%emt 0:
02:31]} 18. Nxd4 {[%emt 0:02:57]} cxd4 {[%eval -32,39] [%emt 0:00:01]} 19. Rxd4
{[%emt 0:04:17] (Kg2)} Bxh3 {[%eval -43,35] [%emt 0:02:38]} 20. f3 {[%emt 0:00:
15] (Te1)} Bf5 {[%eval -45,36] [%emt 0:06:57]} 21. Kf2 {[%emt 0:00:07]} Rc6 {
[%eval -38,35] [%emt 0:02:14]} 22. c4 {[%emt 0:02:12]} Ra6 {[%eval -34,36]
[%emt 0:01:34]} 23. Be3 {[%emt 0:02:09] (Ke3)} Rd8 {[%eval -28,35] [%emt 0:02:
58]} 24. Rxd8+ {[%emt 0:06:08]} Kxd8 {[%eval -28,39] [%emt 0:00:01]} 25. Nc3 {
[%emt 0:04:25]} Bb4 {[%eval -20,37] [%emt 0:01:49]} 26. Nd5 {[%emt 0:01:50]
(Ld4)} Bf8 {[%eval -11,36] [%emt 0:02:00]} 27. Bd4 {[%emt 0:02:03]} c6 {
[%eval -8,42] [%emt 0:00:01]} 28. Ne3 {[%emt 0:01:38]} Be6 {[%eval -8,42]
[%emt 0:00:09]} 29. Ke2 {[%emt 0:04:38] (f4)} g6 {[%eval -11,35] [%emt 0:02:39]
} 30. f4 {[%emt 0:02:05] (a4)} Bd7 {[%eval -13,43] [%emt 0:03:35]} 31. Kd3 {
[%emt 0:01:30] (Kd2)} c5 {[%eval -11,44] [%emt 0:02:31]} 32. Bb2 {[%emt 0:02:
10] (Lc3)} b5 {[%eval -19,38] [%emt 0:01:35]} 33. cxb5 {[%emt 0:01:33] (Kc2)}
Bxb5+ {[%eval -30,34] [%emt 0:01:34]} 34. Kc3 {[%emt 0:00:08]} Bd7 {[%eval -33,
37] [%emt 0:02:06]} 35. Nd5 {[%emt 0:01:21] (Lc1)} Be6 {[%eval -42,39] [%emt 0:
01:44]} 36. Rd1 {[%emt 0:00:10] (Se3)} Kc8 {[%eval -45,44] [%emt 0:01:45]} 37.
a3 {[%emt 0:02:17]} Ra5 {[%eval -37,43] [%emt 0:00:01]} 38. a4 {[%emt 0:02:35]
(Se3)} c4 {[%eval -78,46] [%emt 0:01:47]} 39. bxc4 {[%emt 0:01:07]} Rxa4 {
[%eval -78,49] [%emt 0:00:26]} 40. Kb3 {[%emt 0:00:42]} Ra6 {[%eval -78,48]
[%emt 0:02:23]} 41. Rd3 {[%emt 0:01:07] (Te1)} Kb7 {[%eval -76,43] [%emt 0:02:
50]} 42. Kc3 {[%emt 0:01:03] (Td2)} Rc6 {[%eval -105,36] [%emt 0:01:32]} 43.
Bc1 {[%emt 0:01:46] (Kb3)} Ra6 {[%eval -75,45] [%emt 0:03:17]} 44. Bb2 {
[%emt 0:01:01]} Bc5 {[%eval -74,43] [%emt 0:02:19]} 45. Rd2 {[%emt 0:04:20]}
Bf5 {[%eval -73,48] [%emt 0:00:01]} 46. Kb3 {[%emt 0:00:55]} Bd7 {[%eval -73,
45] [%emt 0:00:31]} 47. Kc3 {[%emt 0:00:52] (Tg2)} Be6 {[%eval -72,43] [%emt 0:
01:31]} 48. Rd3 {[%emt 0:00:55] (Kb3)} Ra4 {[%eval -69,41] [%emt 0:01:10]} 49.
Kb3 {[%emt 0:00:12]} Bf5 {[%eval -67,44] [%emt 0:02:29]} 50. Rd1 {[%emt 0:00:
57]} Bg4 {[%eval -67,44] [%emt 0:00:02]} 51. Rd2 {[%emt 0:00:46]} Ra6 {[%eval
-67,40] [%emt 0:00:23]} 52. Kc3 {[%emt 0:00:27] (Th2)} Be6 {[%eval -64,45]
[%emt 0:01:36]} 53. Rd3 {[%emt 0:01:13] (Kb3)} Rc6 {[%eval -67,40] [%emt 0:01:
41]} 54. Kb3 {[%emt 0:02:37] (Td2)} Bf5 {[%eval -64,44] [%emt 0:01:19]} 55. Rd2
{[%emt 0:00:36]} Ra6 {[%eval -63,42] [%emt 0:00:38]} 56. Kc3 {[%emt 0:00:09]
(Te2)} Kc8 {[%eval -63,45] [%emt 0:02:17]} 57. Nb4 {[%emt 0:00:37] (Kb3)} Ra4 {
[%eval -57,41] [%emt 0:00:46]} 58. Nd5 {[%emt 0:00:13]} Be6 {[%eval -57,45]
[%emt 0:00:38]} 59. Kb3 {[%emt 0:00:34]} Ra6 {[%eval -56,40] [%emt 0:00:38]}
60. Rd3 {[%emt 0:00:38] (Te2)} Bd7 {[%eval -55,44] [%emt 0:01:14]} 61. Kc3 {
[%emt 0:00:35] (Td2)} Bf5 {[%eval -49,41] [%emt 0:00:45]} 62. Rf3 {[%emt 0:00:
09] (Td2)} Kb8 {[%eval -65,37] [%emt 0:00:38]} 63. Rf1 {[%emt 0:00:37]} Kb7 {
[%eval -64,36] [%emt 0:00:11]} 64. Rf3 {[%emt 0:02:58]} Bg4 {[%eval -59,40]
[%emt 0:00:01]} 65. Rf1 {[%emt 0:01:00]} Rc6 {[%eval -53,27] [%emt 0:00:01]}
66. Rb1 {[%emt 0:00:39]} Bf5 {[%eval -48,30] [%emt 0:00:02]} 67. Ra1 {[%emt 0:
00:24]} Ra6 {[%eval -47,36] [%emt 0:00:30]} 68. Rf1 {[%emt 0:00:24]} Bg4 {
[%eval -45,38] [%emt 0:00:19]} 69. Rb1 {[%emt 0:00:25]} Kc8 {[%eval -40,32]
[%emt 0:00:15]} 70. Rf1 {[%emt 0:00:23]} Be6 {[%eval -36,36] [%emt 0:02:33]}
71. Rd1 {[%emt 0:00:23] (Th1)} Bf2 {[%eval -43,35] [%emt 0:00:49]} 72. Rd3 {
[%emt 0:00:26]} Kb7 {[%eval -43,36] [%emt 0:00:05]} 73. Kb3 {[%emt 0:00:20]
(Tf3)} Bf5 {[%eval -36,35] [%emt 0:01:02]} 74. Rf3 {[%emt 0:00:11]} Bg1 {
[%eval -35,38] [%emt 0:00:30]} 75. Rf1 {[%emt 0:00:24]} Bc5 {[%eval -34,35]
[%emt 0:00:33]} 76. Kc3 {[%emt 0:00:10]} Kb8 {[%eval -33,34] [%emt 0:00:20]}
77. Nb4 {[%emt 0:00:24] (Kb3)} Ra4 {[%eval -32,37] [%emt 0:00:44]} 78. Nd5 {
[%emt 0:00:08]} Kb7 {[%eval -32,37] [%emt 0:00:19]} 79. Kb3 {[%emt 0:00:13]}
Ra5 {[%eval -31,37] [%emt 0:00:10]} 80. Rd1 {[%emt 0:00:24]} Ra6 {[%eval -31,
23] [%emt 0:00:01]} 81. Rd2 {[%emt 0:00:10]} Kc8 {[%eval -30,34] [%emt 0:00:10]
} 82. Kc3 {[%emt 0:00:18] (Td1)} Kb8 {[%eval -29,36] [%emt 0:00:28]} 83. Nf6 {
[%emt 0:00:20] (Kb3)} Kc8 {[%eval -26,36] [%emt 0:00:21]} 84. Re2 {[%emt 0:00:
19] (Sd5)} Rc6 {[%eval -44,34] [%emt 0:00:32]} 85. Nd5 {[%emt 0:00:18]} a5 {
[%eval -44,32] [%emt 0:00:06]} 86. Bc1 {[%emt 0:00:16]} Be6 {[%eval -42,37]
[%emt 0:00:25]} 87. Be3 {[%emt 0:00:08] (Tg2)} Bf8 {[%eval -79,36] [%emt 0:00:
21]} 88. Kd4 {[%emt 0:00:07]} a4 {[%eval -81,38] [%emt 0:00:17]} 89. Ra2 {
[%emt 0:00:55]} a3 {[%eval -81,42] [%emt 0:00:01]} 90. Bc1 {[%emt 0:00:09]} Ra6
{[%eval -81,43] [%emt 0:00:11]} 91. Bd2 {[%emt 0:00:07]} Ra7 {[%eval -81,44]
[%emt 0:00:33]} 92. Bb4 {[%emt 0:00:08]} Bxb4 {[%eval -81,47] [%emt 0:00:18]}
93. Nxb4 {[%emt 0:00:07]} Rb7 {[%eval -81,45] [%emt 0:00:12]} 94. Nd5 {[%emt 0:
00:07]} Rb3 {[%eval -81,46] [%emt 0:00:16]} 95. Nc3 {[%emt 0:00:19]} Kd8 {
[%eval -81,42] [%emt 0:00:01]} 96. Ra1 {[%emt 0:00:15] (Se4)} Kd7 {[%eval -81,
45] [%emt 0:00:20]} 97. Rd1 {[%emt 0:00:16] (Ta2)} Kc6 {[%eval -81,46] [%emt 0:
00:24]} 98. Ra1 {[%emt 0:00:07]} Bf5 {[%eval -80,44] [%emt 0:00:23]} 99. Ra2 {
[%emt 0:00:08]} Kb7 {[%eval -80,43] [%emt 0:00:09]} 100. Ra1 {[%emt 0:00:15]}
Kb6 {[%eval -79,33] [%emt 0:00:05]} 101. Nd5+ {[%emt 0:00:07] (Ta2)} Kc6 {
[%eval -78,42] [%emt 0:00:23]} 102. Nc3 {[%emt 0:00:15]} Be6 {[%eval -78,39]
[%emt 0:00:01]} 103. Ra2 {[%emt 0:00:30]} Kd7 {[%eval -78,41] [%emt 0:00:01]}
104. Ra1 {[%emt 0:00:12]} Bg4 {[%eval -78,42] [%emt 0:00:19]} 105. Kd3 {
[%emt 0:00:14] (Ta2)} Bf5+ {[%eval -78,45] [%emt 0:00:24]} 106. Kd4 {[%emt 0:
00:08]} Ke6 {[%eval -78,43] [%emt 0:00:17]} 107. Re1 {[%emt 0:00:14] (Ta2)} Bg4
{[%eval -78,41] [%emt 0:00:21]} 108. Ra1 {[%emt 0:00:14]} Bh3 {[%eval -78,40]
[%emt 0:00:09]} 109. Rh1 {[%emt 0:00:14] (Kd3)} Bg2 {[%eval -78,36] [%emt 0:00:
17]} 110. Rg1 {[%emt 0:00:07]} Bc6 {[%eval -78,38] [%emt 0:00:09]} 111. Ra1 {
[%emt 0:00:14] (Te1)} Kf5 {[%eval -231,32] [%emt 0:00:22]} 112. Ne2 {[%emt 0:
00:18] (c5)} Kg4 {[%eval -629,25] [%emt 0:00:22]} 113. c5 {[%emt 0:00:20] (Kc5)
} Rb4+ {[%eval -581,29] [%emt 0:00:38]} 114. Ke3 {[%emt 0:00:20] (Kd3)} Ra4 {
[%eval -628,27] [%emt 0:00:18]} 115. Kf2 {[%emt 0:00:12] (e6)} a2 {[%eval -732,
24] [%emt 0:00:12]} 116. Ke3 {[%emt 0:00:07] (Sc1)} Ra3+ {[%eval -821,23]
[%emt 0:00:11]} 117. Kd2 {[%emt 0:00:18]} Bd5 {[%eval -861,19] [%emt 0:00:01]}
118. Nc3 {[%emt 0:00:15]} Bf3 {[%eval -900,20] [%emt 0:00:01]} 119. Rxa2 {
[%emt 0:00:16] (Sb5)} Rxa2+ {[%eval -1177,23] [%emt 0:00:15]} 120. Nxa2 {
[%emt 0:00:07]} Kxg3 {[%eval -1212,23] [%emt 0:00:05]} 121. Nc3 {[%emt 0:00:19]
} h4 {[%eval -1292,20] [%emt 0:00:01]} 122. f5 {[%emt 0:00:19]} gxf5 {[%eval
-1340,18] [%emt 0:00:01]} 123. Ne2+ {[%emt 0:00:13]} Kf2 {[%eval -1379,18]
[%emt 0:00:01]} 124. Nf4 {[%emt 0:00:10]} Bb7 {[%eval -1487,24] [%emt 0:00:32]}
125. Kd3 {[%emt 0:00:14] (Kc3)} Kf3 {[%eval -1595,24] [%emt 0:00:17]} 126. e6 {
[%emt 0:00:12] (Sh5)} fxe6 {[%eval -1754,22] [%emt 0:00:14]} 127. Ng6 {[%emt 0:
00:07] (Sh3)} h3 {[%eval -2149,21] [%emt 0:00:13]} 128. Kd4 {[%emt 0:00:07]
(Kc4)} Kg3 {[%eval -4207,22] [%emt 0:00:09]} 129. Ke5 {[%emt 0:00:09] (Se7)} h2
{[%eval -6142,28] [%emt 0:01:07]} 130. Kf6 {[%emt 0:00:06]} f4 {[%eval -25000,
20] [%emt 0:00:00]} 131. c6 {[%emt 0:00:08] (Se7)} Bxc6 {[%eval -25000,22]
[%emt 0:00:14]} 132. Ne7 {[%emt 0:00:07] (Se5)} h1=Q {[%eval -25000,21] [%emt
0:00:14]} 133. Kg7 {[%emt 0:00:10] (Kf7)} Be8 {[%eval -32741,22] [%emt 0:00:03]
} 134. Kf6 {[%emt 0:00:12]} Qe4 {[%eval -32745,22] [%emt 0:00:01]} 135. Nc8 {
[%emt 0:00:06]} Qf5+ {[%eval -32749,22] [%emt 0:00:00]} 136. Ke7 {[%emt 0:00:
07]} Qf7+ {[%eval -32751,21] [%emt 0:00:01]} 137. Kd6 {[%emt 0:00:06]} Qd7+ {
[%eval -32753,20] [%emt 0:00:02]} 138. Ke5 {[%emt 0:00:07] (Kc5)} Qxc8 {
[%eval -32755,21] [%emt 0:00:04]} 139. Kd4 {[%emt 0:00:07] (Kd6)} Qc7 {[%eval
-32757,23] [%emt 0:00:02]} 140. Kd3 {[%emt 0:00:09]} Bg6+ {[%eval -32759,66]
[%emt 0:00:01]} 141. Ke2 {[%emt 0:00:07]} f3+ {[%eval -32761,67] [%emt 0:00:12]
} 142. Kd2 {[%emt 0:00:06] (Kd1)} Kf2 {[%eval -32763,99] [%emt 0:00:01]} 143.
Kd1 {[%emt 0:00:08]} Qc2# {[%eval -32765,99] [%emt 0:00:01]} 0-1

OK, technically this is somehow a solution, and it goes well with your suggestion that this is only doeable with contempt. But it is also clear, that this bully approach can't work all the time, and it is boring. (And yes, this was my one and only game in this setup with Dragon).

Then I thought: maybe this MCTS approach can do better (this time, this is the second game I tried). For my personal taste, this is a much nicer game, but well, it was a draw

[Event "Lang 120min+10sek"]
[Site "Berlin"]
[Date "2023.08.06"]
[Round "?"]
[White "Crafty 25.6"]
[Black "Dragon 3.1 by Komodo Chess 64-, MCTS."]
[Result "1/2-1/2"]
[ECO "C13"]
[PlyCount "86"]
[TimeControl "7200+10"]

{4096MB, LAPTOP-NCDN8BTK} 1. e4 {[%emt 0:00:00]} e6 {[%eval 26,18] [%emt 0:02:
51]} 2. d4 {[%emt 0:00:06]} d5 {[%eval 27,18] [%emt 0:02:26]} 3. Nc3 {[%emt 0:
00:06] (e5)} Nf6 {[%eval 17,18] [%emt 0:03:22]} 4. Bg5 {[%emt 0:00:06]} dxe4 {
[%eval 17,18] [%emt 0:02:42]} 5. Nxe4 {[%emt 0:00:06]} Be7 {[%eval 17,19]
[%emt 0:02:17]} 6. Bxf6 {[%emt 0:00:07]} Bxf6 {[%eval 12,20] [%emt 0:05:51]} 7.
Nf3 {[%emt 0:00:06]} O-O {[%eval 13,19] [%emt 0:02:18]} 8. Qd2 {[%emt 0:00:06]}
Be7 {[%eval 16,19] [%emt 0:02:22]} 9. Bd3 {[%emt 0:00:08]} Nd7 {[%eval 27,19]
[%emt 0:02:26]} 10. O-O-O {[%emt 0:00:07]} c5 {[%eval 19,21] [%emt 0:06:03]}
11. Nxc5 {[%emt 0:00:07] (dxc5)} Nxc5 {[%eval 5,18] [%emt 0:04:09]} 12. dxc5 {
[%emt 0:00:06]} Qc7 {[%eval 7,19] [%emt 0:02:16]} 13. Qe2 {[%emt 0:09:55] (g3)}
Bd7 {[%eval 14,19] [%emt 0:05:08]} 14. Bxh7+ {[%emt 0:02:51] (Se5)} Kxh7 {
[%eval 1,19] [%emt 0:03:52]} 15. Qd3+ {[%emt 0:00:06]} Kg8 {[%eval 0,20] [%emt
0:02:07]} 16. Qxd7 {[%emt 0:00:31]} Qxc5 {[%eval 0,20] [%emt 0:02:36]} 17. Qd4
{[%emt 0:00:06]} Qa5 {[%eval 0,21] [%emt 0:02:05]} 18. Kb1 {[%emt 0:01:09]} Bf6
{[%eval -1,20] [%emt 0:01:02]} 19. Qd2 {[%emt 0:02:43]} Qb5 {[%eval -2,19]
[%emt 0:00:01]} 20. Nd4 {[%emt 0:02:37]} Qd5 {[%eval -1,20] [%emt 0:01:11]} 21.
f4 {[%emt 0:05:03] (Sf3)} Rfd8 {[%eval -2,19] [%emt 0:04:15]} 22. Nf3 {[%emt 0:
00:07] (Sb3)} Qe4 {[%eval -1,20] [%emt 0:07:15]} 23. Qc1 {[%emt 0:02:24]} Qe2 {
[%eval 0,19] [%emt 0:00:10]} 24. Rhg1 {[%emt 0:02:03] (Txd8+)} Rxd1 {[%eval
-12,19] [%emt 0:05:33]} 25. Qxd1 {[%emt 0:02:17]} Qb5 {[%eval -12,19] [%emt 0:
00:01]} 26. Qc1 {[%emt 0:02:04] (Se5)} Rc8 {[%eval -9,20] [%emt 0:06:08]} 27.
Re1 {[%emt 0:02:15]} Rd8 {[%eval -4,18] [%emt 0:00:01]} 28. a3 {[%emt 0:01:51]
(Se5)} Bc3 {[%eval -9,18] [%emt 0:02:52]} 29. Rg1 {[%emt 0:01:32]} g6 {[%eval
-12,19] [%emt 0:01:03]} 30. h4 {[%emt 0:03:17]} Rd6 {[%eval -8,19] [%emt 0:00:
01]} 31. Ka1 {[%emt 0:02:09] (Ka2)} Bf6 {[%eval -9,19] [%emt 0:03:44]} 32. c3 {
[%emt 0:00:07]} Qc5 {[%eval -7,20] [%emt 0:02:28]} 33. Ng5 {[%emt 0:03:51]
(Te1)} Qc4 {[%eval -2,19] [%emt 0:03:46]} 34. g4 {[%emt 0:00:15]} Rd3 {[%eval
-3,20] [%emt 0:03:10]} 35. Kb1 {[%emt 0:00:07]} Bxc3 {[%eval -3,19] [%emt 0:01:
14]} 36. f5 {[%emt 0:01:50] (h5)} exf5 {[%eval -4,18] [%emt 0:02:04]} 37. gxf5
{[%emt 0:00:07]} Qb5 {[%eval -4,19] [%emt 0:00:58]} 38. fxg6 {[%emt 0:01:41]}
fxg6 {[%eval -2,19] [%emt 0:00:01]} 39. Qc2 {[%emt 0:02:02]} Bxb2 {[%eval 0,20]
[%emt 0:00:01]} 40. a4 {[%emt 0:01:52]} Qb3 {[%eval 0,20] [%emt 0:00:01]} 41.
Qxb2 {[%emt 0:01:24]} Rd1+ {[%eval 0,21] [%emt 0:01:41]} 42. Rxd1 {[%emt 0:00:
07]} Qxd1+ {[%eval 0,21] [%emt 0:00:55]} 43. Qc1 {[%emt 0:00:27] (Ka2)} Qb3+ {
[%eval 0,20] [%emt 0:01:45]} 1/2-1/2

So - Dragon seems to have the exact same problem as Stockfish. It can't beat Crafty as often as it should.

With every recent Stockfish version (and now with Dragon) I get the same general result - they give Crafty way too many draws according to their rating.

As there are other programs that can beat Crafty like always ( and without obvious contempt trolling), I have begun to wonder if part of the current "draw problem" doesn't just show a conceptional weakness in current chess programs.

[lkaufman]

If you use a book with only "best" lines, the top engine ratings will depend primarily on which ones are better at avoiding draws against much weaker engines, which may basically mean which ones do "Contempt" better, since almost all the games will be drawn between the top ones

I agreed to this statement a little too fast. I have since thought about it some more.

The first obvious question for me was - so how does Dragon deal with this problem?

This was my first game:

[Event "Lang 120min+10sek"]
[Site "Berlin"]
[Date "2023.08.05"]
[Round "?"]
[White "Crafty 25.6"]
[Black "Dragon 3.1 by Komodo Chess 64-"]
[Result "0-1"]
[ECO "C67"]
[PlyCount "286"]
[TimeControl "7200+10"]

{4096MB, LAPTOP-NCDN8BTK} 1. e4 {[%emt 0:00:00]} e5 {[%eval 39,33] [%emt 0:04:
56]} 2. Nf3 {[%emt 0:00:06]} Nc6 {[%eval 32,33] [%emt 0:03:00]} 3. Bb5 {
[%emt 0:00:06]} Nf6 {[%eval 24,33] [%emt 0:02:02]} 4. O-O {[%emt 0:00:06]} Nxe4
{[%eval 12,36] [%emt 0:02:24]} 5. d4 {[%emt 0:00:06] (Te1)} Nd6 {[%eval 24,34]
[%emt 0:02:42]} 6. Bxc6 {[%emt 0:00:06]} dxc6 {[%eval 22,36] [%emt 0:01:53]} 7.
dxe5 {[%emt 0:00:07]} Nf5 {[%eval 16,37] [%emt 0:02:49]} 8. Qxd8+ {[%emt 0:00:
07]} Kxd8 {[%eval 15,24] [%emt 0:00:01]} 9. Nc3 {[%emt 0:00:06]} Be7 {[%eval
14,36] [%emt 0:06:23]} 10. Rd1+ {[%emt 0:00:07] (h3)} Ke8 {[%eval 16,36] [%emt
0:03:32]} 11. h3 {[%emt 0:00:08] (Lf4)} h5 {[%eval 13,36] [%emt 0:02:32]} 12.
g3 {[%emt 0:00:06] (Se2)} Be6 {[%eval 0,34] [%emt 0:02:34]} 13. Ng5 {[%emt 0:
00:17] (Lg5)} Rh6 {[%eval 0,39] [%emt 0:02:17]} 14. b3 {[%emt 0:04:17] (Sce4)}
Rg6 {[%eval -10,37] [%emt 0:02:45]} 15. Nce4 {[%emt 0:08:43] (Sf3)} Bc8 {
[%eval -26,34] [%emt 0:02:48]} 16. Bf4 {[%emt 0:08:18] (Kh2)} c5 {[%eval -35,
38] [%emt 0:03:49]} 17. Nf3 {[%emt 0:03:17] (c3)} Nd4 {[%eval -38,36] [%emt 0:
02:31]} 18. Nxd4 {[%emt 0:02:57]} cxd4 {[%eval -32,39] [%emt 0:00:01]} 19. Rxd4
{[%emt 0:04:17] (Kg2)} Bxh3 {[%eval -43,35] [%emt 0:02:38]} 20. f3 {[%emt 0:00:
15] (Te1)} Bf5 {[%eval -45,36] [%emt 0:06:57]} 21. Kf2 {[%emt 0:00:07]} Rc6 {
[%eval -38,35] [%emt 0:02:14]} 22. c4 {[%emt 0:02:12]} Ra6 {[%eval -34,36]
[%emt 0:01:34]} 23. Be3 {[%emt 0:02:09] (Ke3)} Rd8 {[%eval -28,35] [%emt 0:02:
58]} 24. Rxd8+ {[%emt 0:06:08]} Kxd8 {[%eval -28,39] [%emt 0:00:01]} 25. Nc3 {
[%emt 0:04:25]} Bb4 {[%eval -20,37] [%emt 0:01:49]} 26. Nd5 {[%emt 0:01:50]
(Ld4)} Bf8 {[%eval -11,36] [%emt 0:02:00]} 27. Bd4 {[%emt 0:02:03]} c6 {
[%eval -8,42] [%emt 0:00:01]} 28. Ne3 {[%emt 0:01:38]} Be6 {[%eval -8,42]
[%emt 0:00:09]} 29. Ke2 {[%emt 0:04:38] (f4)} g6 {[%eval -11,35] [%emt 0:02:39]
} 30. f4 {[%emt 0:02:05] (a4)} Bd7 {[%eval -13,43] [%emt 0:03:35]} 31. Kd3 {
[%emt 0:01:30] (Kd2)} c5 {[%eval -11,44] [%emt 0:02:31]} 32. Bb2 {[%emt 0:02:
10] (Lc3)} b5 {[%eval -19,38] [%emt 0:01:35]} 33. cxb5 {[%emt 0:01:33] (Kc2)}
Bxb5+ {[%eval -30,34] [%emt 0:01:34]} 34. Kc3 {[%emt 0:00:08]} Bd7 {[%eval -33,
37] [%emt 0:02:06]} 35. Nd5 {[%emt 0:01:21] (Lc1)} Be6 {[%eval -42,39] [%emt 0:
01:44]} 36. Rd1 {[%emt 0:00:10] (Se3)} Kc8 {[%eval -45,44] [%emt 0:01:45]} 37.
a3 {[%emt 0:02:17]} Ra5 {[%eval -37,43] [%emt 0:00:01]} 38. a4 {[%emt 0:02:35]
(Se3)} c4 {[%eval -78,46] [%emt 0:01:47]} 39. bxc4 {[%emt 0:01:07]} Rxa4 {
[%eval -78,49] [%emt 0:00:26]} 40. Kb3 {[%emt 0:00:42]} Ra6 {[%eval -78,48]
[%emt 0:02:23]} 41. Rd3 {[%emt 0:01:07] (Te1)} Kb7 {[%eval -76,43] [%emt 0:02:
50]} 42. Kc3 {[%emt 0:01:03] (Td2)} Rc6 {[%eval -105,36] [%emt 0:01:32]} 43.
Bc1 {[%emt 0:01:46] (Kb3)} Ra6 {[%eval -75,45] [%emt 0:03:17]} 44. Bb2 {
[%emt 0:01:01]} Bc5 {[%eval -74,43] [%emt 0:02:19]} 45. Rd2 {[%emt 0:04:20]}
Bf5 {[%eval -73,48] [%emt 0:00:01]} 46. Kb3 {[%emt 0:00:55]} Bd7 {[%eval -73,
45] [%emt 0:00:31]} 47. Kc3 {[%emt 0:00:52] (Tg2)} Be6 {[%eval -72,43] [%emt 0:
01:31]} 48. Rd3 {[%emt 0:00:55] (Kb3)} Ra4 {[%eval -69,41] [%emt 0:01:10]} 49.
Kb3 {[%emt 0:00:12]} Bf5 {[%eval -67,44] [%emt 0:02:29]} 50. Rd1 {[%emt 0:00:
57]} Bg4 {[%eval -67,44] [%emt 0:00:02]} 51. Rd2 {[%emt 0:00:46]} Ra6 {[%eval
-67,40] [%emt 0:00:23]} 52. Kc3 {[%emt 0:00:27] (Th2)} Be6 {[%eval -64,45]
[%emt 0:01:36]} 53. Rd3 {[%emt 0:01:13] (Kb3)} Rc6 {[%eval -67,40] [%emt 0:01:
41]} 54. Kb3 {[%emt 0:02:37] (Td2)} Bf5 {[%eval -64,44] [%emt 0:01:19]} 55. Rd2
{[%emt 0:00:36]} Ra6 {[%eval -63,42] [%emt 0:00:38]} 56. Kc3 {[%emt 0:00:09]
(Te2)} Kc8 {[%eval -63,45] [%emt 0:02:17]} 57. Nb4 {[%emt 0:00:37] (Kb3)} Ra4 {
[%eval -57,41] [%emt 0:00:46]} 58. Nd5 {[%emt 0:00:13]} Be6 {[%eval -57,45]
[%emt 0:00:38]} 59. Kb3 {[%emt 0:00:34]} Ra6 {[%eval -56,40] [%emt 0:00:38]}
60. Rd3 {[%emt 0:00:38] (Te2)} Bd7 {[%eval -55,44] [%emt 0:01:14]} 61. Kc3 {
[%emt 0:00:35] (Td2)} Bf5 {[%eval -49,41] [%emt 0:00:45]} 62. Rf3 {[%emt 0:00:
09] (Td2)} Kb8 {[%eval -65,37] [%emt 0:00:38]} 63. Rf1 {[%emt 0:00:37]} Kb7 {
[%eval -64,36] [%emt 0:00:11]} 64. Rf3 {[%emt 0:02:58]} Bg4 {[%eval -59,40]
[%emt 0:00:01]} 65. Rf1 {[%emt 0:01:00]} Rc6 {[%eval -53,27] [%emt 0:00:01]}
66. Rb1 {[%emt 0:00:39]} Bf5 {[%eval -48,30] [%emt 0:00:02]} 67. Ra1 {[%emt 0:
00:24]} Ra6 {[%eval -47,36] [%emt 0:00:30]} 68. Rf1 {[%emt 0:00:24]} Bg4 {
[%eval -45,38] [%emt 0:00:19]} 69. Rb1 {[%emt 0:00:25]} Kc8 {[%eval -40,32]
[%emt 0:00:15]} 70. Rf1 {[%emt 0:00:23]} Be6 {[%eval -36,36] [%emt 0:02:33]}
71. Rd1 {[%emt 0:00:23] (Th1)} Bf2 {[%eval -43,35] [%emt 0:00:49]} 72. Rd3 {
[%emt 0:00:26]} Kb7 {[%eval -43,36] [%emt 0:00:05]} 73. Kb3 {[%emt 0:00:20]
(Tf3)} Bf5 {[%eval -36,35] [%emt 0:01:02]} 74. Rf3 {[%emt 0:00:11]} Bg1 {
[%eval -35,38] [%emt 0:00:30]} 75. Rf1 {[%emt 0:00:24]} Bc5 {[%eval -34,35]
[%emt 0:00:33]} 76. Kc3 {[%emt 0:00:10]} Kb8 {[%eval -33,34] [%emt 0:00:20]}
77. Nb4 {[%emt 0:00:24] (Kb3)} Ra4 {[%eval -32,37] [%emt 0:00:44]} 78. Nd5 {
[%emt 0:00:08]} Kb7 {[%eval -32,37] [%emt 0:00:19]} 79. Kb3 {[%emt 0:00:13]}
Ra5 {[%eval -31,37] [%emt 0:00:10]} 80. Rd1 {[%emt 0:00:24]} Ra6 {[%eval -31,
23] [%emt 0:00:01]} 81. Rd2 {[%emt 0:00:10]} Kc8 {[%eval -30,34] [%emt 0:00:10]
} 82. Kc3 {[%emt 0:00:18] (Td1)} Kb8 {[%eval -29,36] [%emt 0:00:28]} 83. Nf6 {
[%emt 0:00:20] (Kb3)} Kc8 {[%eval -26,36] [%emt 0:00:21]} 84. Re2 {[%emt 0:00:
19] (Sd5)} Rc6 {[%eval -44,34] [%emt 0:00:32]} 85. Nd5 {[%emt 0:00:18]} a5 {
[%eval -44,32] [%emt 0:00:06]} 86. Bc1 {[%emt 0:00:16]} Be6 {[%eval -42,37]
[%emt 0:00:25]} 87. Be3 {[%emt 0:00:08] (Tg2)} Bf8 {[%eval -79,36] [%emt 0:00:
21]} 88. Kd4 {[%emt 0:00:07]} a4 {[%eval -81,38] [%emt 0:00:17]} 89. Ra2 {
[%emt 0:00:55]} a3 {[%eval -81,42] [%emt 0:00:01]} 90. Bc1 {[%emt 0:00:09]} Ra6
{[%eval -81,43] [%emt 0:00:11]} 91. Bd2 {[%emt 0:00:07]} Ra7 {[%eval -81,44]
[%emt 0:00:33]} 92. Bb4 {[%emt 0:00:08]} Bxb4 {[%eval -81,47] [%emt 0:00:18]}
93. Nxb4 {[%emt 0:00:07]} Rb7 {[%eval -81,45] [%emt 0:00:12]} 94. Nd5 {[%emt 0:
00:07]} Rb3 {[%eval -81,46] [%emt 0:00:16]} 95. Nc3 {[%emt 0:00:19]} Kd8 {
[%eval -81,42] [%emt 0:00:01]} 96. Ra1 {[%emt 0:00:15] (Se4)} Kd7 {[%eval -81,
45] [%emt 0:00:20]} 97. Rd1 {[%emt 0:00:16] (Ta2)} Kc6 {[%eval -81,46] [%emt 0:
00:24]} 98. Ra1 {[%emt 0:00:07]} Bf5 {[%eval -80,44] [%emt 0:00:23]} 99. Ra2 {
[%emt 0:00:08]} Kb7 {[%eval -80,43] [%emt 0:00:09]} 100. Ra1 {[%emt 0:00:15]}
Kb6 {[%eval -79,33] [%emt 0:00:05]} 101. Nd5+ {[%emt 0:00:07] (Ta2)} Kc6 {
[%eval -78,42] [%emt 0:00:23]} 102. Nc3 {[%emt 0:00:15]} Be6 {[%eval -78,39]
[%emt 0:00:01]} 103. Ra2 {[%emt 0:00:30]} Kd7 {[%eval -78,41] [%emt 0:00:01]}
104. Ra1 {[%emt 0:00:12]} Bg4 {[%eval -78,42] [%emt 0:00:19]} 105. Kd3 {
[%emt 0:00:14] (Ta2)} Bf5+ {[%eval -78,45] [%emt 0:00:24]} 106. Kd4 {[%emt 0:
00:08]} Ke6 {[%eval -78,43] [%emt 0:00:17]} 107. Re1 {[%emt 0:00:14] (Ta2)} Bg4
{[%eval -78,41] [%emt 0:00:21]} 108. Ra1 {[%emt 0:00:14]} Bh3 {[%eval -78,40]
[%emt 0:00:09]} 109. Rh1 {[%emt 0:00:14] (Kd3)} Bg2 {[%eval -78,36] [%emt 0:00:
17]} 110. Rg1 {[%emt 0:00:07]} Bc6 {[%eval -78,38] [%emt 0:00:09]} 111. Ra1 {
[%emt 0:00:14] (Te1)} Kf5 {[%eval -231,32] [%emt 0:00:22]} 112. Ne2 {[%emt 0:
00:18] (c5)} Kg4 {[%eval -629,25] [%emt 0:00:22]} 113. c5 {[%emt 0:00:20] (Kc5)
} Rb4+ {[%eval -581,29] [%emt 0:00:38]} 114. Ke3 {[%emt 0:00:20] (Kd3)} Ra4 {
[%eval -628,27] [%emt 0:00:18]} 115. Kf2 {[%emt 0:00:12] (e6)} a2 {[%eval -732,
24] [%emt 0:00:12]} 116. Ke3 {[%emt 0:00:07] (Sc1)} Ra3+ {[%eval -821,23]
[%emt 0:00:11]} 117. Kd2 {[%emt 0:00:18]} Bd5 {[%eval -861,19] [%emt 0:00:01]}
118. Nc3 {[%emt 0:00:15]} Bf3 {[%eval -900,20] [%emt 0:00:01]} 119. Rxa2 {
[%emt 0:00:16] (Sb5)} Rxa2+ {[%eval -1177,23] [%emt 0:00:15]} 120. Nxa2 {
[%emt 0:00:07]} Kxg3 {[%eval -1212,23] [%emt 0:00:05]} 121. Nc3 {[%emt 0:00:19]
} h4 {[%eval -1292,20] [%emt 0:00:01]} 122. f5 {[%emt 0:00:19]} gxf5 {[%eval
-1340,18] [%emt 0:00:01]} 123. Ne2+ {[%emt 0:00:13]} Kf2 {[%eval -1379,18]
[%emt 0:00:01]} 124. Nf4 {[%emt 0:00:10]} Bb7 {[%eval -1487,24] [%emt 0:00:32]}
125. Kd3 {[%emt 0:00:14] (Kc3)} Kf3 {[%eval -1595,24] [%emt 0:00:17]} 126. e6 {
[%emt 0:00:12] (Sh5)} fxe6 {[%eval -1754,22] [%emt 0:00:14]} 127. Ng6 {[%emt 0:
00:07] (Sh3)} h3 {[%eval -2149,21] [%emt 0:00:13]} 128. Kd4 {[%emt 0:00:07]
(Kc4)} Kg3 {[%eval -4207,22] [%emt 0:00:09]} 129. Ke5 {[%emt 0:00:09] (Se7)} h2
{[%eval -6142,28] [%emt 0:01:07]} 130. Kf6 {[%emt 0:00:06]} f4 {[%eval -25000,
20] [%emt 0:00:00]} 131. c6 {[%emt 0:00:08] (Se7)} Bxc6 {[%eval -25000,22]
[%emt 0:00:14]} 132. Ne7 {[%emt 0:00:07] (Se5)} h1=Q {[%eval -25000,21] [%emt
0:00:14]} 133. Kg7 {[%emt 0:00:10] (Kf7)} Be8 {[%eval -32741,22] [%emt 0:00:03]
} 134. Kf6 {[%emt 0:00:12]} Qe4 {[%eval -32745,22] [%emt 0:00:01]} 135. Nc8 {
[%emt 0:00:06]} Qf5+ {[%eval -32749,22] [%emt 0:00:00]} 136. Ke7 {[%emt 0:00:
07]} Qf7+ {[%eval -32751,21] [%emt 0:00:01]} 137. Kd6 {[%emt 0:00:06]} Qd7+ {
[%eval -32753,20] [%emt 0:00:02]} 138. Ke5 {[%emt 0:00:07] (Kc5)} Qxc8 {
[%eval -32755,21] [%emt 0:00:04]} 139. Kd4 {[%emt 0:00:07] (Kd6)} Qc7 {[%eval
-32757,23] [%emt 0:00:02]} 140. Kd3 {[%emt 0:00:09]} Bg6+ {[%eval -32759,66]
[%emt 0:00:01]} 141. Ke2 {[%emt 0:00:07]} f3+ {[%eval -32761,67] [%emt 0:00:12]
} 142. Kd2 {[%emt 0:00:06] (Kd1)} Kf2 {[%eval -32763,99] [%emt 0:00:01]} 143.
Kd1 {[%emt 0:00:08]} Qc2# {[%eval -32765,99] [%emt 0:00:01]} 0-1

OK, technically this is somehow a solution, and it goes well with your suggestion that this is only doeable with contempt. But it is also clear, that this bully approach can't work all the time, and it is boring. (And yes, this was my one and only game in this setup with Dragon).

Then I thought: maybe this MCTS approach can do better (this time, this is the second game I tried). For my personal taste, this is a much nicer game, but well, it was a draw

[Event "Lang 120min+10sek"]
[Site "Berlin"]
[Date "2023.08.06"]
[Round "?"]
[White "Crafty 25.6"]
[Black "Dragon 3.1 by Komodo Chess 64-, MCTS."]
[Result "1/2-1/2"]
[ECO "C13"]
[PlyCount "86"]
[TimeControl "7200+10"]

{4096MB, LAPTOP-NCDN8BTK} 1. e4 {[%emt 0:00:00]} e6 {[%eval 26,18] [%emt 0:02:
51]} 2. d4 {[%emt 0:00:06]} d5 {[%eval 27,18] [%emt 0:02:26]} 3. Nc3 {[%emt 0:
00:06] (e5)} Nf6 {[%eval 17,18] [%emt 0:03:22]} 4. Bg5 {[%emt 0:00:06]} dxe4 {
[%eval 17,18] [%emt 0:02:42]} 5. Nxe4 {[%emt 0:00:06]} Be7 {[%eval 17,19]
[%emt 0:02:17]} 6. Bxf6 {[%emt 0:00:07]} Bxf6 {[%eval 12,20] [%emt 0:05:51]} 7.
Nf3 {[%emt 0:00:06]} O-O {[%eval 13,19] [%emt 0:02:18]} 8. Qd2 {[%emt 0:00:06]}
Be7 {[%eval 16,19] [%emt 0:02:22]} 9. Bd3 {[%emt 0:00:08]} Nd7 {[%eval 27,19]
[%emt 0:02:26]} 10. O-O-O {[%emt 0:00:07]} c5 {[%eval 19,21] [%emt 0:06:03]}
11. Nxc5 {[%emt 0:00:07] (dxc5)} Nxc5 {[%eval 5,18] [%emt 0:04:09]} 12. dxc5 {
[%emt 0:00:06]} Qc7 {[%eval 7,19] [%emt 0:02:16]} 13. Qe2 {[%emt 0:09:55] (g3)}
Bd7 {[%eval 14,19] [%emt 0:05:08]} 14. Bxh7+ {[%emt 0:02:51] (Se5)} Kxh7 {
[%eval 1,19] [%emt 0:03:52]} 15. Qd3+ {[%emt 0:00:06]} Kg8 {[%eval 0,20] [%emt
0:02:07]} 16. Qxd7 {[%emt 0:00:31]} Qxc5 {[%eval 0,20] [%emt 0:02:36]} 17. Qd4
{[%emt 0:00:06]} Qa5 {[%eval 0,21] [%emt 0:02:05]} 18. Kb1 {[%emt 0:01:09]} Bf6
{[%eval -1,20] [%emt 0:01:02]} 19. Qd2 {[%emt 0:02:43]} Qb5 {[%eval -2,19]
[%emt 0:00:01]} 20. Nd4 {[%emt 0:02:37]} Qd5 {[%eval -1,20] [%emt 0:01:11]} 21.
f4 {[%emt 0:05:03] (Sf3)} Rfd8 {[%eval -2,19] [%emt 0:04:15]} 22. Nf3 {[%emt 0:
00:07] (Sb3)} Qe4 {[%eval -1,20] [%emt 0:07:15]} 23. Qc1 {[%emt 0:02:24]} Qe2 {
[%eval 0,19] [%emt 0:00:10]} 24. Rhg1 {[%emt 0:02:03] (Txd8+)} Rxd1 {[%eval
-12,19] [%emt 0:05:33]} 25. Qxd1 {[%emt 0:02:17]} Qb5 {[%eval -12,19] [%emt 0:
00:01]} 26. Qc1 {[%emt 0:02:04] (Se5)} Rc8 {[%eval -9,20] [%emt 0:06:08]} 27.
Re1 {[%emt 0:02:15]} Rd8 {[%eval -4,18] [%emt 0:00:01]} 28. a3 {[%emt 0:01:51]
(Se5)} Bc3 {[%eval -9,18] [%emt 0:02:52]} 29. Rg1 {[%emt 0:01:32]} g6 {[%eval
-12,19] [%emt 0:01:03]} 30. h4 {[%emt 0:03:17]} Rd6 {[%eval -8,19] [%emt 0:00:
01]} 31. Ka1 {[%emt 0:02:09] (Ka2)} Bf6 {[%eval -9,19] [%emt 0:03:44]} 32. c3 {
[%emt 0:00:07]} Qc5 {[%eval -7,20] [%emt 0:02:28]} 33. Ng5 {[%emt 0:03:51]
(Te1)} Qc4 {[%eval -2,19] [%emt 0:03:46]} 34. g4 {[%emt 0:00:15]} Rd3 {[%eval
-3,20] [%emt 0:03:10]} 35. Kb1 {[%emt 0:00:07]} Bxc3 {[%eval -3,19] [%emt 0:01:
14]} 36. f5 {[%emt 0:01:50] (h5)} exf5 {[%eval -4,18] [%emt 0:02:04]} 37. gxf5
{[%emt 0:00:07]} Qb5 {[%eval -4,19] [%emt 0:00:58]} 38. fxg6 {[%emt 0:01:41]}
fxg6 {[%eval -2,19] [%emt 0:00:01]} 39. Qc2 {[%emt 0:02:02]} Bxb2 {[%eval 0,20]
[%emt 0:00:01]} 40. a4 {[%emt 0:01:52]} Qb3 {[%eval 0,20] [%emt 0:00:01]} 41.
Qxb2 {[%emt 0:01:24]} Rd1+ {[%eval 0,21] [%emt 0:01:41]} 42. Rxd1 {[%emt 0:00:
07]} Qxd1+ {[%eval 0,21] [%emt 0:00:55]} 43. Qc1 {[%emt 0:00:27] (Ka2)} Qb3+ {
[%eval 0,20] [%emt 0:01:45]} 1/2-1/2

So - Dragon seems to have the exact same problem as Stockfish. It can't beat Crafty as often as it should.

With every recent Stockfish version (and now with Dragon) I get the same general result - they give Crafty way too many draws according to their rating.

As there are other programs that can beat Crafty like always ( and without obvious contempt trolling), I have begun to wonder if part of the current "draw problem" doesn't just show a conceptional weakness in current chess programs.

If you can name some engines that would win almost always as Black from the Berlin endgame (starting after Black's eighth move) while showing that SF (plus Dragon as shown) cannot, that would give us a clue. Perhaps NNUE engines are too "honest", they prioritize not losing over winning in slightly inferior positions. Regarding Contempt, the default value in Dragon is too small to have major consequences other than avoiding early repetitions as Black; it should help to avoid major trades but it's too late to avoid the queen trade once you have committed to the Berlin.

[carldaman]

If you use a book with only "best" lines, the top engine ratings will depend primarily on which ones are better at avoiding draws against much weaker engines, which may basically mean which ones do "Contempt" better, since almost all the games will be drawn between the top ones

I agreed to this statement a little too fast. I have since thought about it some more.

The first obvious question for me was - so how does Dragon deal with this problem?

This was my first game:

[Event "Lang 120min+10sek"]
[Site "Berlin"]
[Date "2023.08.05"]
[Round "?"]
[White "Crafty 25.6"]
[Black "Dragon 3.1 by Komodo Chess 64-"]
[Result "0-1"]
[ECO "C67"]
[PlyCount "286"]
[TimeControl "7200+10"]

{4096MB, LAPTOP-NCDN8BTK} 1. e4 {[%emt 0:00:00]} e5 {[%eval 39,33] [%emt 0:04:
56]} 2. Nf3 {[%emt 0:00:06]} Nc6 {[%eval 32,33] [%emt 0:03:00]} 3. Bb5 {
[%emt 0:00:06]} Nf6 {[%eval 24,33] [%emt 0:02:02]} 4. O-O {[%emt 0:00:06]} Nxe4
{[%eval 12,36] [%emt 0:02:24]} 5. d4 {[%emt 0:00:06] (Te1)} Nd6 {[%eval 24,34]
[%emt 0:02:42]} 6. Bxc6 {[%emt 0:00:06]} dxc6 {[%eval 22,36] [%emt 0:01:53]} 7.
dxe5 {[%emt 0:00:07]} Nf5 {[%eval 16,37] [%emt 0:02:49]} 8. Qxd8+ {[%emt 0:00:
07]} Kxd8 {[%eval 15,24] [%emt 0:00:01]} 9. Nc3 {[%emt 0:00:06]} Be7 {[%eval
14,36] [%emt 0:06:23]} 10. Rd1+ {[%emt 0:00:07] (h3)} Ke8 {[%eval 16,36] [%emt
0:03:32]} 11. h3 {[%emt 0:00:08] (Lf4)} h5 {[%eval 13,36] [%emt 0:02:32]} 12.
g3 {[%emt 0:00:06] (Se2)} Be6 {[%eval 0,34] [%emt 0:02:34]} 13. Ng5 {[%emt 0:
00:17] (Lg5)} Rh6 {[%eval 0,39] [%emt 0:02:17]} 14. b3 {[%emt 0:04:17] (Sce4)}
Rg6 {[%eval -10,37] [%emt 0:02:45]} 15. Nce4 {[%emt 0:08:43] (Sf3)} Bc8 {
[%eval -26,34] [%emt 0:02:48]} 16. Bf4 {[%emt 0:08:18] (Kh2)} c5 {[%eval -35,
38] [%emt 0:03:49]} 17. Nf3 {[%emt 0:03:17] (c3)} Nd4 {[%eval -38,36] [%emt 0:
02:31]} 18. Nxd4 {[%emt 0:02:57]} cxd4 {[%eval -32,39] [%emt 0:00:01]} 19. Rxd4
{[%emt 0:04:17] (Kg2)} Bxh3 {[%eval -43,35] [%emt 0:02:38]} 20. f3 {[%emt 0:00:
15] (Te1)} Bf5 {[%eval -45,36] [%emt 0:06:57]} 21. Kf2 {[%emt 0:00:07]} Rc6 {
[%eval -38,35] [%emt 0:02:14]} 22. c4 {[%emt 0:02:12]} Ra6 {[%eval -34,36]
[%emt 0:01:34]} 23. Be3 {[%emt 0:02:09] (Ke3)} Rd8 {[%eval -28,35] [%emt 0:02:
58]} 24. Rxd8+ {[%emt 0:06:08]} Kxd8 {[%eval -28,39] [%emt 0:00:01]} 25. Nc3 {
[%emt 0:04:25]} Bb4 {[%eval -20,37] [%emt 0:01:49]} 26. Nd5 {[%emt 0:01:50]
(Ld4)} Bf8 {[%eval -11,36] [%emt 0:02:00]} 27. Bd4 {[%emt 0:02:03]} c6 {
[%eval -8,42] [%emt 0:00:01]} 28. Ne3 {[%emt 0:01:38]} Be6 {[%eval -8,42]
[%emt 0:00:09]} 29. Ke2 {[%emt 0:04:38] (f4)} g6 {[%eval -11,35] [%emt 0:02:39]
} 30. f4 {[%emt 0:02:05] (a4)} Bd7 {[%eval -13,43] [%emt 0:03:35]} 31. Kd3 {
[%emt 0:01:30] (Kd2)} c5 {[%eval -11,44] [%emt 0:02:31]} 32. Bb2 {[%emt 0:02:
10] (Lc3)} b5 {[%eval -19,38] [%emt 0:01:35]} 33. cxb5 {[%emt 0:01:33] (Kc2)}
Bxb5+ {[%eval -30,34] [%emt 0:01:34]} 34. Kc3 {[%emt 0:00:08]} Bd7 {[%eval -33,
37] [%emt 0:02:06]} 35. Nd5 {[%emt 0:01:21] (Lc1)} Be6 {[%eval -42,39] [%emt 0:
01:44]} 36. Rd1 {[%emt 0:00:10] (Se3)} Kc8 {[%eval -45,44] [%emt 0:01:45]} 37.
a3 {[%emt 0:02:17]} Ra5 {[%eval -37,43] [%emt 0:00:01]} 38. a4 {[%emt 0:02:35]
(Se3)} c4 {[%eval -78,46] [%emt 0:01:47]} 39. bxc4 {[%emt 0:01:07]} Rxa4 {
[%eval -78,49] [%emt 0:00:26]} 40. Kb3 {[%emt 0:00:42]} Ra6 {[%eval -78,48]
[%emt 0:02:23]} 41. Rd3 {[%emt 0:01:07] (Te1)} Kb7 {[%eval -76,43] [%emt 0:02:
50]} 42. Kc3 {[%emt 0:01:03] (Td2)} Rc6 {[%eval -105,36] [%emt 0:01:32]} 43.
Bc1 {[%emt 0:01:46] (Kb3)} Ra6 {[%eval -75,45] [%emt 0:03:17]} 44. Bb2 {
[%emt 0:01:01]} Bc5 {[%eval -74,43] [%emt 0:02:19]} 45. Rd2 {[%emt 0:04:20]}
Bf5 {[%eval -73,48] [%emt 0:00:01]} 46. Kb3 {[%emt 0:00:55]} Bd7 {[%eval -73,
45] [%emt 0:00:31]} 47. Kc3 {[%emt 0:00:52] (Tg2)} Be6 {[%eval -72,43] [%emt 0:
01:31]} 48. Rd3 {[%emt 0:00:55] (Kb3)} Ra4 {[%eval -69,41] [%emt 0:01:10]} 49.
Kb3 {[%emt 0:00:12]} Bf5 {[%eval -67,44] [%emt 0:02:29]} 50. Rd1 {[%emt 0:00:
57]} Bg4 {[%eval -67,44] [%emt 0:00:02]} 51. Rd2 {[%emt 0:00:46]} Ra6 {[%eval
-67,40] [%emt 0:00:23]} 52. Kc3 {[%emt 0:00:27] (Th2)} Be6 {[%eval -64,45]
[%emt 0:01:36]} 53. Rd3 {[%emt 0:01:13] (Kb3)} Rc6 {[%eval -67,40] [%emt 0:01:
41]} 54. Kb3 {[%emt 0:02:37] (Td2)} Bf5 {[%eval -64,44] [%emt 0:01:19]} 55. Rd2
{[%emt 0:00:36]} Ra6 {[%eval -63,42] [%emt 0:00:38]} 56. Kc3 {[%emt 0:00:09]
(Te2)} Kc8 {[%eval -63,45] [%emt 0:02:17]} 57. Nb4 {[%emt 0:00:37] (Kb3)} Ra4 {
[%eval -57,41] [%emt 0:00:46]} 58. Nd5 {[%emt 0:00:13]} Be6 {[%eval -57,45]
[%emt 0:00:38]} 59. Kb3 {[%emt 0:00:34]} Ra6 {[%eval -56,40] [%emt 0:00:38]}
60. Rd3 {[%emt 0:00:38] (Te2)} Bd7 {[%eval -55,44] [%emt 0:01:14]} 61. Kc3 {
[%emt 0:00:35] (Td2)} Bf5 {[%eval -49,41] [%emt 0:00:45]} 62. Rf3 {[%emt 0:00:
09] (Td2)} Kb8 {[%eval -65,37] [%emt 0:00:38]} 63. Rf1 {[%emt 0:00:37]} Kb7 {
[%eval -64,36] [%emt 0:00:11]} 64. Rf3 {[%emt 0:02:58]} Bg4 {[%eval -59,40]
[%emt 0:00:01]} 65. Rf1 {[%emt 0:01:00]} Rc6 {[%eval -53,27] [%emt 0:00:01]}
66. Rb1 {[%emt 0:00:39]} Bf5 {[%eval -48,30] [%emt 0:00:02]} 67. Ra1 {[%emt 0:
00:24]} Ra6 {[%eval -47,36] [%emt 0:00:30]} 68. Rf1 {[%emt 0:00:24]} Bg4 {
[%eval -45,38] [%emt 0:00:19]} 69. Rb1 {[%emt 0:00:25]} Kc8 {[%eval -40,32]
[%emt 0:00:15]} 70. Rf1 {[%emt 0:00:23]} Be6 {[%eval -36,36] [%emt 0:02:33]}
71. Rd1 {[%emt 0:00:23] (Th1)} Bf2 {[%eval -43,35] [%emt 0:00:49]} 72. Rd3 {
[%emt 0:00:26]} Kb7 {[%eval -43,36] [%emt 0:00:05]} 73. Kb3 {[%emt 0:00:20]
(Tf3)} Bf5 {[%eval -36,35] [%emt 0:01:02]} 74. Rf3 {[%emt 0:00:11]} Bg1 {
[%eval -35,38] [%emt 0:00:30]} 75. Rf1 {[%emt 0:00:24]} Bc5 {[%eval -34,35]
[%emt 0:00:33]} 76. Kc3 {[%emt 0:00:10]} Kb8 {[%eval -33,34] [%emt 0:00:20]}
77. Nb4 {[%emt 0:00:24] (Kb3)} Ra4 {[%eval -32,37] [%emt 0:00:44]} 78. Nd5 {
[%emt 0:00:08]} Kb7 {[%eval -32,37] [%emt 0:00:19]} 79. Kb3 {[%emt 0:00:13]}
Ra5 {[%eval -31,37] [%emt 0:00:10]} 80. Rd1 {[%emt 0:00:24]} Ra6 {[%eval -31,
23] [%emt 0:00:01]} 81. Rd2 {[%emt 0:00:10]} Kc8 {[%eval -30,34] [%emt 0:00:10]
} 82. Kc3 {[%emt 0:00:18] (Td1)} Kb8 {[%eval -29,36] [%emt 0:00:28]} 83. Nf6 {
[%emt 0:00:20] (Kb3)} Kc8 {[%eval -26,36] [%emt 0:00:21]} 84. Re2 {[%emt 0:00:
19] (Sd5)} Rc6 {[%eval -44,34] [%emt 0:00:32]} 85. Nd5 {[%emt 0:00:18]} a5 {
[%eval -44,32] [%emt 0:00:06]} 86. Bc1 {[%emt 0:00:16]} Be6 {[%eval -42,37]
[%emt 0:00:25]} 87. Be3 {[%emt 0:00:08] (Tg2)} Bf8 {[%eval -79,36] [%emt 0:00:
21]} 88. Kd4 {[%emt 0:00:07]} a4 {[%eval -81,38] [%emt 0:00:17]} 89. Ra2 {
[%emt 0:00:55]} a3 {[%eval -81,42] [%emt 0:00:01]} 90. Bc1 {[%emt 0:00:09]} Ra6
{[%eval -81,43] [%emt 0:00:11]} 91. Bd2 {[%emt 0:00:07]} Ra7 {[%eval -81,44]
[%emt 0:00:33]} 92. Bb4 {[%emt 0:00:08]} Bxb4 {[%eval -81,47] [%emt 0:00:18]}
93. Nxb4 {[%emt 0:00:07]} Rb7 {[%eval -81,45] [%emt 0:00:12]} 94. Nd5 {[%emt 0:
00:07]} Rb3 {[%eval -81,46] [%emt 0:00:16]} 95. Nc3 {[%emt 0:00:19]} Kd8 {
[%eval -81,42] [%emt 0:00:01]} 96. Ra1 {[%emt 0:00:15] (Se4)} Kd7 {[%eval -81,
45] [%emt 0:00:20]} 97. Rd1 {[%emt 0:00:16] (Ta2)} Kc6 {[%eval -81,46] [%emt 0:
00:24]} 98. Ra1 {[%emt 0:00:07]} Bf5 {[%eval -80,44] [%emt 0:00:23]} 99. Ra2 {
[%emt 0:00:08]} Kb7 {[%eval -80,43] [%emt 0:00:09]} 100. Ra1 {[%emt 0:00:15]}
Kb6 {[%eval -79,33] [%emt 0:00:05]} 101. Nd5+ {[%emt 0:00:07] (Ta2)} Kc6 {
[%eval -78,42] [%emt 0:00:23]} 102. Nc3 {[%emt 0:00:15]} Be6 {[%eval -78,39]
[%emt 0:00:01]} 103. Ra2 {[%emt 0:00:30]} Kd7 {[%eval -78,41] [%emt 0:00:01]}
104. Ra1 {[%emt 0:00:12]} Bg4 {[%eval -78,42] [%emt 0:00:19]} 105. Kd3 {
[%emt 0:00:14] (Ta2)} Bf5+ {[%eval -78,45] [%emt 0:00:24]} 106. Kd4 {[%emt 0:
00:08]} Ke6 {[%eval -78,43] [%emt 0:00:17]} 107. Re1 {[%emt 0:00:14] (Ta2)} Bg4
{[%eval -78,41] [%emt 0:00:21]} 108. Ra1 {[%emt 0:00:14]} Bh3 {[%eval -78,40]
[%emt 0:00:09]} 109. Rh1 {[%emt 0:00:14] (Kd3)} Bg2 {[%eval -78,36] [%emt 0:00:
17]} 110. Rg1 {[%emt 0:00:07]} Bc6 {[%eval -78,38] [%emt 0:00:09]} 111. Ra1 {
[%emt 0:00:14] (Te1)} Kf5 {[%eval -231,32] [%emt 0:00:22]} 112. Ne2 {[%emt 0:
00:18] (c5)} Kg4 {[%eval -629,25] [%emt 0:00:22]} 113. c5 {[%emt 0:00:20] (Kc5)
} Rb4+ {[%eval -581,29] [%emt 0:00:38]} 114. Ke3 {[%emt 0:00:20] (Kd3)} Ra4 {
[%eval -628,27] [%emt 0:00:18]} 115. Kf2 {[%emt 0:00:12] (e6)} a2 {[%eval -732,
24] [%emt 0:00:12]} 116. Ke3 {[%emt 0:00:07] (Sc1)} Ra3+ {[%eval -821,23]
[%emt 0:00:11]} 117. Kd2 {[%emt 0:00:18]} Bd5 {[%eval -861,19] [%emt 0:00:01]}
118. Nc3 {[%emt 0:00:15]} Bf3 {[%eval -900,20] [%emt 0:00:01]} 119. Rxa2 {
[%emt 0:00:16] (Sb5)} Rxa2+ {[%eval -1177,23] [%emt 0:00:15]} 120. Nxa2 {
[%emt 0:00:07]} Kxg3 {[%eval -1212,23] [%emt 0:00:05]} 121. Nc3 {[%emt 0:00:19]
} h4 {[%eval -1292,20] [%emt 0:00:01]} 122. f5 {[%emt 0:00:19]} gxf5 {[%eval
-1340,18] [%emt 0:00:01]} 123. Ne2+ {[%emt 0:00:13]} Kf2 {[%eval -1379,18]
[%emt 0:00:01]} 124. Nf4 {[%emt 0:00:10]} Bb7 {[%eval -1487,24] [%emt 0:00:32]}
125. Kd3 {[%emt 0:00:14] (Kc3)} Kf3 {[%eval -1595,24] [%emt 0:00:17]} 126. e6 {
[%emt 0:00:12] (Sh5)} fxe6 {[%eval -1754,22] [%emt 0:00:14]} 127. Ng6 {[%emt 0:
00:07] (Sh3)} h3 {[%eval -2149,21] [%emt 0:00:13]} 128. Kd4 {[%emt 0:00:07]
(Kc4)} Kg3 {[%eval -4207,22] [%emt 0:00:09]} 129. Ke5 {[%emt 0:00:09] (Se7)} h2
{[%eval -6142,28] [%emt 0:01:07]} 130. Kf6 {[%emt 0:00:06]} f4 {[%eval -25000,
20] [%emt 0:00:00]} 131. c6 {[%emt 0:00:08] (Se7)} Bxc6 {[%eval -25000,22]
[%emt 0:00:14]} 132. Ne7 {[%emt 0:00:07] (Se5)} h1=Q {[%eval -25000,21] [%emt
0:00:14]} 133. Kg7 {[%emt 0:00:10] (Kf7)} Be8 {[%eval -32741,22] [%emt 0:00:03]
} 134. Kf6 {[%emt 0:00:12]} Qe4 {[%eval -32745,22] [%emt 0:00:01]} 135. Nc8 {
[%emt 0:00:06]} Qf5+ {[%eval -32749,22] [%emt 0:00:00]} 136. Ke7 {[%emt 0:00:
07]} Qf7+ {[%eval -32751,21] [%emt 0:00:01]} 137. Kd6 {[%emt 0:00:06]} Qd7+ {
[%eval -32753,20] [%emt 0:00:02]} 138. Ke5 {[%emt 0:00:07] (Kc5)} Qxc8 {
[%eval -32755,21] [%emt 0:00:04]} 139. Kd4 {[%emt 0:00:07] (Kd6)} Qc7 {[%eval
-32757,23] [%emt 0:00:02]} 140. Kd3 {[%emt 0:00:09]} Bg6+ {[%eval -32759,66]
[%emt 0:00:01]} 141. Ke2 {[%emt 0:00:07]} f3+ {[%eval -32761,67] [%emt 0:00:12]
} 142. Kd2 {[%emt 0:00:06] (Kd1)} Kf2 {[%eval -32763,99] [%emt 0:00:01]} 143.
Kd1 {[%emt 0:00:08]} Qc2# {[%eval -32765,99] [%emt 0:00:01]} 0-1

OK, technically this is somehow a solution, and it goes well with your suggestion that this is only doeable with contempt. But it is also clear, that this bully approach can't work all the time, and it is boring. (And yes, this was my one and only game in this setup with Dragon).

Then I thought: maybe this MCTS approach can do better (this time, this is the second game I tried). For my personal taste, this is a much nicer game, but well, it was a draw

[Event "Lang 120min+10sek"]
[Site "Berlin"]
[Date "2023.08.06"]
[Round "?"]
[White "Crafty 25.6"]
[Black "Dragon 3.1 by Komodo Chess 64-, MCTS."]
[Result "1/2-1/2"]
[ECO "C13"]
[PlyCount "86"]
[TimeControl "7200+10"]

{4096MB, LAPTOP-NCDN8BTK} 1. e4 {[%emt 0:00:00]} e6 {[%eval 26,18] [%emt 0:02:
51]} 2. d4 {[%emt 0:00:06]} d5 {[%eval 27,18] [%emt 0:02:26]} 3. Nc3 {[%emt 0:
00:06] (e5)} Nf6 {[%eval 17,18] [%emt 0:03:22]} 4. Bg5 {[%emt 0:00:06]} dxe4 {
[%eval 17,18] [%emt 0:02:42]} 5. Nxe4 {[%emt 0:00:06]} Be7 {[%eval 17,19]
[%emt 0:02:17]} 6. Bxf6 {[%emt 0:00:07]} Bxf6 {[%eval 12,20] [%emt 0:05:51]} 7.
Nf3 {[%emt 0:00:06]} O-O {[%eval 13,19] [%emt 0:02:18]} 8. Qd2 {[%emt 0:00:06]}
Be7 {[%eval 16,19] [%emt 0:02:22]} 9. Bd3 {[%emt 0:00:08]} Nd7 {[%eval 27,19]
[%emt 0:02:26]} 10. O-O-O {[%emt 0:00:07]} c5 {[%eval 19,21] [%emt 0:06:03]}
11. Nxc5 {[%emt 0:00:07] (dxc5)} Nxc5 {[%eval 5,18] [%emt 0:04:09]} 12. dxc5 {
[%emt 0:00:06]} Qc7 {[%eval 7,19] [%emt 0:02:16]} 13. Qe2 {[%emt 0:09:55] (g3)}
Bd7 {[%eval 14,19] [%emt 0:05:08]} 14. Bxh7+ {[%emt 0:02:51] (Se5)} Kxh7 {
[%eval 1,19] [%emt 0:03:52]} 15. Qd3+ {[%emt 0:00:06]} Kg8 {[%eval 0,20] [%emt
0:02:07]} 16. Qxd7 {[%emt 0:00:31]} Qxc5 {[%eval 0,20] [%emt 0:02:36]} 17. Qd4
{[%emt 0:00:06]} Qa5 {[%eval 0,21] [%emt 0:02:05]} 18. Kb1 {[%emt 0:01:09]} Bf6
{[%eval -1,20] [%emt 0:01:02]} 19. Qd2 {[%emt 0:02:43]} Qb5 {[%eval -2,19]
[%emt 0:00:01]} 20. Nd4 {[%emt 0:02:37]} Qd5 {[%eval -1,20] [%emt 0:01:11]} 21.
f4 {[%emt 0:05:03] (Sf3)} Rfd8 {[%eval -2,19] [%emt 0:04:15]} 22. Nf3 {[%emt 0:
00:07] (Sb3)} Qe4 {[%eval -1,20] [%emt 0:07:15]} 23. Qc1 {[%emt 0:02:24]} Qe2 {
[%eval 0,19] [%emt 0:00:10]} 24. Rhg1 {[%emt 0:02:03] (Txd8+)} Rxd1 {[%eval
-12,19] [%emt 0:05:33]} 25. Qxd1 {[%emt 0:02:17]} Qb5 {[%eval -12,19] [%emt 0:
00:01]} 26. Qc1 {[%emt 0:02:04] (Se5)} Rc8 {[%eval -9,20] [%emt 0:06:08]} 27.
Re1 {[%emt 0:02:15]} Rd8 {[%eval -4,18] [%emt 0:00:01]} 28. a3 {[%emt 0:01:51]
(Se5)} Bc3 {[%eval -9,18] [%emt 0:02:52]} 29. Rg1 {[%emt 0:01:32]} g6 {[%eval
-12,19] [%emt 0:01:03]} 30. h4 {[%emt 0:03:17]} Rd6 {[%eval -8,19] [%emt 0:00:
01]} 31. Ka1 {[%emt 0:02:09] (Ka2)} Bf6 {[%eval -9,19] [%emt 0:03:44]} 32. c3 {
[%emt 0:00:07]} Qc5 {[%eval -7,20] [%emt 0:02:28]} 33. Ng5 {[%emt 0:03:51]
(Te1)} Qc4 {[%eval -2,19] [%emt 0:03:46]} 34. g4 {[%emt 0:00:15]} Rd3 {[%eval
-3,20] [%emt 0:03:10]} 35. Kb1 {[%emt 0:00:07]} Bxc3 {[%eval -3,19] [%emt 0:01:
14]} 36. f5 {[%emt 0:01:50] (h5)} exf5 {[%eval -4,18] [%emt 0:02:04]} 37. gxf5
{[%emt 0:00:07]} Qb5 {[%eval -4,19] [%emt 0:00:58]} 38. fxg6 {[%emt 0:01:41]}
fxg6 {[%eval -2,19] [%emt 0:00:01]} 39. Qc2 {[%emt 0:02:02]} Bxb2 {[%eval 0,20]
[%emt 0:00:01]} 40. a4 {[%emt 0:01:52]} Qb3 {[%eval 0,20] [%emt 0:00:01]} 41.
Qxb2 {[%emt 0:01:24]} Rd1+ {[%eval 0,21] [%emt 0:01:41]} 42. Rxd1 {[%emt 0:00:
07]} Qxd1+ {[%eval 0,21] [%emt 0:00:55]} 43. Qc1 {[%emt 0:00:27] (Ka2)} Qb3+ {
[%eval 0,20] [%emt 0:01:45]} 1/2-1/2

So - Dragon seems to have the exact same problem as Stockfish. It can't beat Crafty as often as it should.

With every recent Stockfish version (and now with Dragon) I get the same general result - they give Crafty way too many draws according to their rating.

As there are other programs that can beat Crafty like always ( and without obvious contempt trolling), I have begun to wonder if part of the current "draw problem" doesn't just show a conceptional weakness in current chess programs.

NNUE engines are too risk averse. It's a serious flaw that detracts from the great strength.

It won't be fixed until testing begins to include more 'lopsided' matchups such as this where the much weaker engine can draw the stronger one an inordinate amount of times.

[smatovic]

Unbalanced openings are perfectly fine if each engine gets to play it with both sides against the same opponents. A better engine might convert the advantage into a win or hold the disadvantage to a draw, compared to drawing with advantage and losing with disadvantage. It just shouldn't be so unbalanced that the results are obvious, such as starting with a piece handicap.

For Dragon 3.2 4 cpu on CCRL Rapid I count about 96% draws against the seven opposing engines over 3500 (all SF or SF derivatives of course). Most likely the few decisive games were from openings that would now be considered somewhat dubious, even if they were thought to be reasonably balanced when the books were made. If you use books that never exit with Black more than say 5 centipawns worse than "par" (the initial eval) I think it would be over 99% draws. There are no other valid pairings over 3500 that are not between SF derivates or other Dragon versions.

AFAIK SF uses UHO books for testing, do they/the rating lists/tournaments do what Ras suggests? As far as I got it, you need to play the same UHO opening as white and black, and if the score is 1:1: then it is a draw?

--
Srdja

[CornfedForever]

NNUE engines are too risk averse. It's a serious flaw that detracts from the great strength.

Some do say that...but isn't part of the beauty of the nets to come up with potentially risky lines which calculation alone will not? I mean, haven't we all seen those moves/lines an a/b engine would never consider simply because they can't 'take that leap of faith'?

h4 and people shout 'Alpha Zero'!....of course, they should really be shouting Bent Larsen!

[carldaman]

NNUE engines are too risk averse. It's a serious flaw that detracts from the great strength.

Some do say that...but isn't part of the beauty of the nets to come up with potentially risky lines which calculation alone will not? I mean, haven't we all seen those moves/lines an a/b engine would never consider simply because they can't 'take that leap of faith'?

h4 and people shout 'Alpha Zero'!....of course, they should really be shouting Bent Larsen!

Right, I'd been a long time fan of kingside pawnstorms, pushing the h-pawn early and often

What I wrote is not always true, though - nn-based engines like the new CSTal, Slowchess and Revenge take more risks and can make some daring sacs. Some of dkappe's nets also achieve this.
So NNUE does NOT necessarily have to mean risk averse...

[Chessqueen]

Announcing Torch: New #2 Chess Engine
CHESScom
CHESScom
Jul 13, 2023, 6:17 PM

Chess.com News
Torch is a brand-new chess engine built from the ground up by top chess engine developers. The team is composed of: Andrew Grant (Ethereal), Finn Eggers & Kim Kåhre (Koivisto), Jay Honnold (Berserk), and Michael Whiteley & Dietrich Kappe (Dragon). Mark Lefler and Larry Kaufman have been advisors to the project.

Many on the team have long discussed joining forces to produce an engine capable of contending with the best out there. Through the support and funding of Chess.com, we have done exactly that. A single unified effort by our developers has allowed us to rapidly produce the new number-two engine according to Torch’s performance in the Chess.com Computer Chess Championship. Our goal is to be number one.

Either Torch is getting stronger or Stockfish play some positions horrible ==>https://www.youtube.com/watch?v=3RJFt5fQZY8