Stockfish handicap match - PART II -

[Uri Blass]

I find arena adjudication as unfair in the following game.
position is equal does not mean it is a draw.
There was no draw by repetition and it is obvious that equal position gives better chances for stockfish so why arena adjudicate the game and does not let the engines to continue to play.


[pgn][Event "Edited game"]
[Site "DESKTOP-7QE6S12"]
[Date "2021.05.15"]
[Round "-"]
[White "-"]
[Black "-"]
[Result "1/2-1/2"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PP3PPP/RNBQKBNR w KQkq - 0 1"]
[SetUp "1"]

{--------------
r n b q k b n r
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P . . . P P P
R N B Q K B N R
white to play
--------------}
1. Bd3 {-4.80/31 1:10} d5 {+2.64/22 31} 2. Nf3 {-4.58/28 5} Nf6
{+2.67/22 12} 3. Nc3 {-4.64/29 7} e6 {+2.48/21 12} 4. O-O {-4.67/28 42} Bd6
{+2.71/22 24} 5. Re1 {-4.39/26 19} O-O {+2.88/20 10} 6. Nb5 {-5.15/30 1:00}
Nc6 {+3.11/23 11} 7. Nxd6 {-5.05/23 2.0} cxd6 {+3.11/23 18} 8. Bg5
{-4.90/23 4} e5 {+3.22/23 14} 9. Bc2 {-4.80/26 10} Be6 {+3.22/23 14} 10.
Rc1 {-4.57/24 4} h6 {+3.34/23 14} 11. Bh4 {-4.63/28 8} Rc8 {+3.40/22 14}
12. Bb1 {-4.77/30 25} Re8 {+3.24/21 14} 13. h3 {-4.70/25 6} Qb6
{+3.40/23 22} 14. Bxf6 {-4.51/23 7} gxf6 {+3.59/24 19} 15. Nh4
{-4.66/28 16} Nd4 {+3.70/23 12} 16. Qd2 {-4.52/24 5} Kf8 {+3.73/22 13} 17.
Rcd1 {-4.01/25 7} Ke7 {+3.74/21 14} 18. Kh2 {-3.89/28 13} h5 {+3.34/22 10}
19. f4 {-3.32/25 6} Rc4 {+3.09/21 11} 20. b3 {-3.24/27 9} Rb4 {+3.17/23 12}
21. Rf1 {-3.55/28 15} Bd7 {+3.36/20 11} 22. Kh1 {-2.76/30 52} a5
{+3.40/21 11} 23. Bh7 {-2.15/26 11} a4 {+3.80/23 11} 24. fxe5 {-2.25/24 5}
dxe5 {+3.88/24 17} 25. bxa4 {-2.80/27 13} Rxa4 {+3.63/20 11} 26. Rb1
{-2.07/28 27} Qa6 {+3.63/21 13} 27. Rf2 {-2.58/28 35} Ra3 {+3.12/19 11} 28.
Rbf1 {-1.57/19 2.0} Be6 {+2.30/20 12} 29. Qh6 {-1.32/26 17} Bxh3
{+2.20/21 14} 30. Qg7 {+0.00/21 2.0} Kd8 {+1.60/22 18} 31. Rxf6
{+0.00/28 10} Be6 {+1.07/23 13} 32. Qg5 {+0.00/28 4} Kc8 {+1.14/23 13} 33.
Rc1+ {+0.00/27 5} Nc6 {+1.10/24 15} 34. Nf5 {+0.00/30 4} Kb8 {+0.79/25 17}
35. Nd6 {+0.00/34 8} Re7 {+0.81/25 16} 36. Qg8+ {+0.00/33 8} Kc7
{+0.00/23 13} 37. Nxf7 {+0.00/34 6} Bxf7 {+0.00/26 16} 38. Rxf7
{+0.00/34 6} Rxf7
{0.00/26 12 Arena Adjudication} 1/2-1/2

[/pgn]

[Rebel]

I make use of the Arena setting to declare the game a draw when both engines agree on a 0.00 score for x moves. I use 5 moves.

[Rebel]

Match cancelled after 50 games, apparently a full piece behind is needed to get Stockfish on its knees. Results, cross table and download games at my forum.

[lkaufman]

Match cancelled after 50 games, apparently a full piece behind is needed to get Stockfish on its knees. Results, cross table and download games at my forum.

According to my tests, Komodo Dragon 2 performs around CCRL 2600 level in Rapid when giving knight odds to other engines (except for your Benjamin, which is much better than other similarly rated engines with the knight odds). Although MCTS does better against humans, the standard version does better in giving knight odds to other engines (by roughly the elo gap between the two), so you should use standard. Since only one of your engines is below 2600 (a bit under 2400), and other a bit under 2700, with the rest much higher, I would expect Dragon to score about 1.5 against those ten opponents at knight odds. If you don't already have Dragon 2, we can send it to you. What your tests do show is that a piece is a much bigger handicap than three pawns. It looks like my claim that a knight is worth 4 pawns in the opening is validated. I don't know whether Stockfish 13 or Dragon 2 will do better giving knight odds to other engines, but I know that setting Contempt to 100 will help Dragon when giving knight odds, whereas I don't know about Stockfish. If you really want to compare them at knight odds, I think you need to pick engines rated in the 2500 to 2700 range on the CCRL Rapid list. It's just a waste of time to try to give knight odds in Rapid to engines rated over 3000; knight odds is roughly a thousand elo handicap.

[Uri Blass]

Some comments:

1)I read rating of the opponents 2400-3200 so I expected average rating near 2800
Practically the opponents were stronger and the median is 3073(I guess Texel1.7 is Texel1.07 from the CCRL because I do not see 1.7 version).

Practically the 9 non buggy opponents are:
Ethereal 12.75 64-bit 3320 +12 −12 46.9% +22.9 60.7% 2004
Arasan 22.2 64-bit 3159 +22 −22 54.0% −23.1 51.9% 592
Vajolet2 2.8 64-bit 3098 +15 −15 48.9% +5.8 53.4% 1355
Komodo 3 64-bit 3089 +14 −14 58.5% −56.6 44.3% 1737
Texel 1.07 64-bit 3073 +11 −11 49.8% −2.0 47.8% 2662
Cheng 4.41 64-bit 2949 +94 −98 35.0% +90.3 56.7% 30 at 40/15(not enough games at 40/15 so I looked also at blitz rating)
Cheng 4.41 64-bit 3058 +73 −71 57.8% −51.5 37.5% 64 at blitz

Senpai 1.0 64-bit 2960 +13 −13 47.7% +14.5 47.3% 2072
ProDeo 3.1 2795 +31 −31 50.6% −3.8 40.1% 344
Fruit 2.1 2695 +25 −25 51.5% −13.9 37.2% 522

2)I believe that latest stockfish is stronger.

Here are games from my own knight handicap tournament of later version of Stockfish against fruit and first Rybka.
both default stockfish with 3 cores and stockfish with 3 cores and contempt 100 made result of 1-1 at 40 moves/13 minutes

[pgn][Event "My Tournament"]
[Site "?"]
[Date "2021.05.14"]
[Round "1"]
[White "stockfish_21051119_x64contempt100"]
[Black "Rybka v1.0 Beta.x64"]
[Result "1-0"]
[TimeControl "40/780:300"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "77"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {-4.94/36 58s} Nf6 {+1.42/14 20s} 2. g3 {-4.71/28 4.8s} d5 {+1.58/15
21s} 3. Bg2 {-4.85/30 7.0s} Nc6 {+1.57/14 4.1s} 4. c4 {-4.58/32 22s} dxc4
{+1.83/15 14s} 5. d5 {-4.43/32 49s} Ne5 {+2.40/15 14s} 6. Qd4 {-4.35/32
4.5s} Ned7 {+3.95/15 13s} 7. Nc3 {-4.55/35 85s} Nb6 {+4.12/13 5.1s} 8. Be3
{-4.63/31 14s} Bd7 {+4.28/15 19s} 9. O-O {-4.50/29 21s} a5 {+4.29/13 15s}
10. Rad1 {-3.55/28 12s} Qc8 {+4.35/14 16s} 11. Qf4 {-3.21/30 18s} h6
{+4.30/14 19s} 12. Bd4 {-2.54/28 4.5s} Ra6 {+4.09/14 20s} 13. e4 {-2.62/29
11s} Nh7 {+4.28/14 23s} 14. Qe3 {-2.06/35 35s} Ng5 {+4.33/14 17s} 15. f4
{-0.91/30 12s} Nh7 {+3.48/15 45s} 16. e5 {-0.47/30 19s} Bg4 {+2.02/14 26s}
17. d6 {+1.50/30 4.9s} Na8 {+2.02/14 16s} 18. Ne4 {+1.37/30 9.4s} cxd6
{+1.17/13 13s} 19. exd6 {+1.59/37 7.4s} Rc6 {-0.20/15 171s} 20. Bc5
{+3.06/29 7.7s} e5 {-0.35/14 6.7s} 21. Rd5 {+3.31/30 8.1s} Kd8 {+0.06/13
12s} 22. fxe5 {+3.63/31 12s} Be6 {-0.62/13 11s} 23. Rdd1 {+3.95/31 11s} Qd7
{-0.67/13 12s} 24. Ba3 {+5.80/29 12s} b6 {-0.43/14 10s} 25. Nc3 {+6.73/27
10s} Ke8 {-0.70/13 7.5s} 26. Nb5 {+7.22/28 9.8s} Rc8 {-1.27/14 9.3s} 27.
Nd4 {+7.51/30 11s} Kd8 {-1.34/14 12s} 28. Rd2 {+9.01/26 11s} c3 {-3.18/12
11s} 29. bxc3 {+9.94/28 14s} f6 {-3.65/12 9.1s} 30. exf6 {+12.78/27 18s}
Ng5 {-5.47/12 19s} 31. h4 {+14.37/27 14s} gxf6 {-7.56/14 79s} 32. hxg5
{+18.05/26 19s} Bc4 {-9.10/14 39s} 33. Qf4 {+21.01/25 16s} Qf7 {-9.06/10
5.8s} 34. g6 {+M19/54 16s} Qxg6 {-17.10/11 17s} 35. d7 {+M17/76 18s} Qxg3
{-M384/3 0.031s} 36. Qxf6+ {+M9/245 15s} Kxd7 37. Ne6+ {+M7/245 0.37s} Bd5
38. Qf7+ {+M5/245 0.009s} Kc6 39. Nd4# {+M1/245 0.004s, White mates} 1-0[/pgn]

[pgn][Event "My Tournament"]
[Site "?"]
[Date "2021.05.14"]
[Round "2"]
[White "stockfish_21051119_x64_avx2"]
[Black "Rybka v1.0 Beta.x64"]
[Result "0-1"]
[TimeControl "40/780:300"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "136"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {-5.40/36 60s} Nf6 {+1.42/14 20s} 2. g3 {-5.44/31 5.7s} d5 {+1.58/15
21s} 3. Bg2 {-5.42/32 12s} Nc6 {+1.57/14 4.1s} 4. O-O {-5.13/31 13s} h5
{+1.89/15 15s} 5. c4 {-4.82/26 5.7s} dxc4 {+2.01/14 15s} 6. d5 {-4.88/33
15s} Ne5 {+2.26/15 12s} 7. Qd4 {-4.76/32 7.6s} Qd6 {+2.12/15 17s} 8. Bf4
{-3.66/29 8.1s} Nfd7 {+2.19/16 19s} 9. Nc3 {-3.82/33 40s} c5 {+3.87/16 17s}
10. Qe3 {-3.47/26 6.2s} f6 {+3.52/16 35s} 11. a4 {-3.09/30 17s} g5
{+3.98/14 14s} 12. Nb5 {-3.12/32 8.2s} Qb6 {+3.45/14 4.0s} 13. Bxe5
{-3.66/37 25s} Nxe5 {+3.00/16 37s} 14. d6 {-3.77/36 7.9s} Rb8 {+3.41/15
39s} 15. Rad1 {-2.50/31 9.2s} Bd7 {+3.86/14 19s} 16. f4 {-2.79/35 27s} gxf4
{+4.19/14 9.7s} 17. Nc7+ {-3.00/36 18s} Kd8 {+4.40/15 17s} 18. gxf4
{-2.62/34 11s} exd6 {+2.75/15 41s} 19. Nd5 {-3.78/36 28s} Ng4 {+3.18/15
21s} 20. Qc3 {-4.12/37 34s} Qb3 {+3.36/16 10s} 21. Qa5+ {-4.12/39 14s} b6
{+3.57/17 31s} 22. Qxa7 {-4.16/37 15s} Rc8 {+3.16/16 25s} 23. Nxb6
{-4.25/39 16s} Rc7 {+3.95/15 21s} 24. Qa8+ {-4.35/38 11s} Bc8 {+3.95/15
12s} 25. Nxc8 {-4.55/42 25s} Rxc8 {+3.95/15 23s} 26. Qa6 {-4.28/37 12s} Rh7
{+3.95/14 11s} 27. h3 {-4.04/36 13s} Qe3+ {+3.95/13 3.6s} 28. Kh1 {-5.59/40
40s} Nf2+ {+3.95/13 3.8s} 29. Rxf2 {-5.48/37 12s} Qxf2 {+3.95/14 7.9s} 30.
Bf3 {-5.91/40 57s} Rd7 {+4.18/15 57s} 31. Qxc4 {-6.01/35 11s} Qg3 {+4.24/15
28s} 32. Qe6 {-5.57/34 11s} Qxf4 {+4.27/14 9.2s} 33. a5 {-6.02/39 71s} Rb8
{+4.65/13 12s} 34. a6 {-5.98/36 46s} Qa4 {+4.69/14 15s} 35. Qd5 {-6.25/33
31s} h4 {+4.97/15 61s} 36. Qd3 {-6.23/31 19s} Rb6 {+5.40/13 6.9s} 37. Rg1
{-6.45/29 10s} Qxa6 {+4.71/14 10.0s} 38. Qxa6 {-6.54/27 3.0s} Rxa6
{+5.51/18 13s} 39. Rg4 {-6.65/23 1.8s} Rh7 {+5.57/17 12s} 40. Rf4 {-6.06/18
0.35s} Be7 {+6.14/18 14s} 41. Bd5 {-7.40/37 102s} Ra1+ {+6.15/19 22s} 42.
Kg2 {-7.59/31 14s} Rd1 {+6.33/19 14s} 43. Be6 {-8.18/36 144s} Rd2 {+6.33/18
21s} 44. Kf3 {-7.78/31 22s} Rxb2 {+6.34/17 21s} 45. Bc4 {-7.86/35 40s} Rb4
{+7.05/18 13s} 46. Ba2 {-9.12/35 85s} Rg7 {+8.80/20 24s} 47. Re4 {-12.01/38
87s} Rg3+ {+10.27/19 23s} 48. Kf4 {-23.31/27 13s} Rxe4+ {+10.27/19 12s} 49.
Kxe4 {-39.18/30 22s} Rxh3 {+11.13/20 24s} 50. e3 {-62.49/34 62s} f5+
{+18.77/19 21s} 51. Kxf5 {-M34/60 8.3s} Rxe3 {+M384/19 3.1s} 52. Kf4
{-M32/66 5.5s} Rd3 {+M384/16 0.034s} 53. Ke4 {-M24/71 4.1s} c4 {+M384/11
0.015s} 54. Bxc4 {-M22/72 4.0s} d5+ {+M384/9 0.002s} 55. Bxd5 {-M20/85
4.2s} Rxd5 {+M384/9 0.002s} 56. Kxd5 {-M18/102 4.4s} h3 {+M384/7 0.001s}
57. Kc4 {-M16/77 4.3s} h2 {+M384/7 0.012s} 58. Kd4 {-M14/81 4.3s} h1=Q
{+M384/4 0s} 59. Ke3 {-M12/156 4.4s} Qd5 {+M384/5 0.004s} 60. Kf4 {-M10/245
2.8s} Bc5 {+M384/3 0s} 61. Kg4 {-M10/245 3.6s} Qe4+ {+M384/3 0.003s} 62.
Kg5 {-M8/245 0.22s} Be7+ {0s} 63. Kh6 {-M6/245 0.006s} Bf6 64. Kh5 {-M6/245
0.024s} Qf5+ 65. Kh6 {-M4/245 0.005s} Qg5+ 66. Kh7 {-M2/245 0.003s} Qh5+
67. Kg8 {-M4/245 0.005s} Qg6+ 68. Kf8 {-M2/245 0.003s} Be7# {Black mates}
0-1
[/pgn]

[pgn][Event "My Tournament"]
[Site "?"]
[Date "2021.05.14"]
[Round "3"]
[White "stockfish_21051119_x64_avx2"]
[Black "Fruit2.2.1"]
[Result "1-0"]
[TimeControl "40/780:300"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "95"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {-5.42/35 54s} Nf6 {+2.77/15 17s} 2. g3 {-5.32/33 16s} d5 {+3.06/16
39s} 3. Bg2 {-5.44/30 6.8s} Nc6 {+3.10/15 19s} 4. O-O {-5.26/31 18s} e6
{+3.05/16 45s} 5. b3 {-5.05/31 7.1s} Be7 {+3.16/14 23s} 6. e3 {-5.21/36
69s} h5 {+3.27/14 20s} 7. h3 {-4.86/28 8.4s} h4 {+3.63/14 13s} 8. g4
{-4.85/29 9.1s} g5 {+3.38/14 23s} 9. c4 {-3.99/30 16s} b6 {+3.00/14 27s}
10. Bb2 {-3.90/32 28s} Ba6 {+3.03/14 23s} 11. Nd2 {-3.61/32 17s} Qd7
{+3.28/13 15s} 12. a3 {-3.56/32 13s} O-O-O {+3.35/13 12s} 13. Qc2 {-3.79/33
32s} Kb7 {+3.30/13 12s} 14. Rfc1 {-3.17/34 27s} Rh6 {+2.98/14 28s} 15. Rab1
{-2.65/31 10s} Rc8 {+3.03/14 17s} 16. b4 {-1.75/30 5.1s} dxc4 {+2.74/15
14s} 17. Bc3 {-2.36/36 21s} b5 {+2.60/15 22s} 18. a4 {-2.63/41 71s} Ka8
{+3.05/15 12s} 19. Qa2 {-2.22/33 13s} Nd5 {+2.96/15 14s} 20. axb5 {-2.06/34
6.3s} Bxb5 {+2.91/15 21s} 21. Nxc4 {-1.96/33 9.3s} Rb8 {+2.79/13 17s} 22.
Bd2 {-2.21/39 39s} f6 {+2.57/13 21s} 23. Na3 {+0.39/32 5.5s} Bd3 {+1.94/14
19s} 24. Rb3 {+1.22/33 9.9s} Bg6 {+1.52/15 18s} 25. b5 {+6.71/28 8.1s} Nd8
{+0.56/15 22s} 26. Nc4 {+7.12/28 8.5s} Rb7 {+0.55/14 12s} 27. e4 {+7.44/31
8.3s} Nf4 {+0.13/14 21s} 28. Bxf4 {+8.67/27 9.3s} gxf4 {-2.23/14 48s} 29.
b6 {+8.70/28 9.4s} f3 {-4.34/15 88s} 30. Bxf3 {+9.55/32 10s} c6 {-5.09/14
46s} 31. Na5 {+10.32/30 11s} Rh7 {-5.29/13 10s} 32. d5 {+12.25/29 19s} Rxb6
{-3.87/12 3.9s} 33. Rxb6 {+12.72/33 16s} Bc5 {-8.32/13 18s} 34. Rxc5
{+16.24/27 19s} Qc7 {-11.59/13 9.3s} 35. Ra6 {+18.60/26 14s} Be8 {-8.84/11
3.3s} 36. Nxc6 {+30.82/27 16s} Bxc6 {-8.96/12 1.6s} 37. dxc6 {+M27/78 16s}
e5 {-11.26/13 2.6s} 38. Qd5 {+M23/67 19s} Kb8 {-12.74/11 0.64s} 39. Be2
{+M21/76 24s} Kc8 {-12.16/10 1.2s} 40. Ra3 {+M19/78 22s} Qf7 {-M18/9 0.91s}
41. Qd2 {+M15/84 12s} Qc7 {-M14/12 16s} 42. Rb3 {+M13/89 13s} Qb6 {-M12/24
15s} 43. Qd6 {+M11/120 12s} Rc7 {-M10/56 17s} 44. Rxb6 {+M9/245 9.1s} axb6
{-M8/54 14s} 45. Ba6+ {+M7/245 0.063s} Kb8 {-M6/54 14s} 46. Qxd8+ {+M5/245
0.006s} Ka7 {-M4/53 14s} 47. Qxc7+ {+M3/245 0.004s} Kxa6 {-M2/52 14s} 48.
Qb7# {+M1/245 0.002s, White mates} 1-0
[/pgn]

[pgn][Event "My Tournament"]
[Site "?"]
[Date "2021.05.14"]
[Round "4"]
[White "stockfish_21051119_x64contempt100"]
[Black "Fruit2.2.1"]
[Result "0-1"]
[TimeControl "40/780:300"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "112"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {-4.71/35 70s} Nf6 {+2.77/15 16s} 2. g3 {-4.77/30 12s} d5 {+3.06/16
36s} 3. Bg2 {-4.63/29 5.3s} Nc6 {+3.10/15 18s} 4. c4 {-4.65/31 16s} dxc4
{+3.06/16 27s} 5. d5 {-4.86/30 23s} Ne5 {+2.88/15 26s} 6. Qd4 {-4.50/31
16s} Ned7 {+3.10/15 17s} 7. Qxc4 {-4.59/32 74s} Nb6 {+2.99/16 19s} 8. Qc2
{-4.59/31 4.8s} Nbxd5 {+3.38/15 22s} 9. O-O {-4.66/32 31s} e6 {+3.34/15
21s} 10. Rd1 {-4.55/29 11s} Bd6 {+3.40/15 18s} 11. Na3 {-5.00/34 79s} O-O
{+3.30/15 17s} 12. b3 {-4.78/30 7.3s} Ng4 {+3.64/13 12s} 13. Nc4 {-5.01/36
46s} Qf6 {+4.50/13 11s} 14. e3 {-4.99/33 4.1s} Qxa1 {+4.08/15 44s} 15. Bb2
{-5.05/34 8.0s} Nb4 {+4.46/16 11s} 16. Qd2 {-5.04/35 9.9s} Qxa2 {+4.48/16
11s} 17. Qc3 {-5.08/35 9.1s} Nf6 {+4.52/16 22s} 18. Ra1 {-5.09/35 8.6s}
Qxa1+ {+4.47/16 21s} 19. Bxa1 {-5.15/32 7.1s} Be7 {+4.50/17 20s} 20. Qc1
{-5.17/30 6.9s} Rd8 {+4.49/15 12s} 21. Na5 {-4.98/34 33s} Nfd5 {+4.66/14
12s} 22. Be4 {-5.12/34 57s} f6 {+4.72/14 20s} 23. Qb1 {-5.22/31 5.3s} h6
{+4.70/15 14s} 24. Bf3 {-5.31/31 7.5s} e5 {+4.87/14 15s} 25. e4 {-5.44/37
31s} Nb6 {+4.76/17 16s} 26. Bc3 {-5.36/31 8.3s} Rb8 {+4.81/17 20s} 27. Kg2
{-5.69/35 49s} Be6 {+4.84/16 15s} 28. Qa1 {-5.63/32 16s} Nc2 {+5.29/16 18s}
29. Qb2 {-5.62/30 10s} Nd4 {+5.37/16 20s} 30. Bh5 {-5.82/27 5.7s} Kh7
{+5.28/15 21s} 31. Qa1 {-5.80/33 35s} c6 {+5.38/14 18s} 32. f4 {-6.17/30
25s} g6 {+5.58/15 17s} 33. Bd1 {-6.69/32 18s} Rd7 {+5.57/15 18s} 34. Nc4
{-6.88/26 11s} Bxc4 {+5.94/14 14s} 35. bxc4 {-6.48/23 1.8s} Nxc4 {+6.00/15
26s} 36. Kh3 {-6.98/26 8.7s} Bc5 {+6.52/16 20s} 37. Bg4 {-7.06/24 3.3s} Rc7
{+6.44/15 26s} 38. Qc1 {-7.37/22 3.1s} b5 {+6.62/15 14s} 39. Qf1 {-7.40/21
0.61s} a5 {+7.09/15 16s} 40. Be2 {-7.70/19 0.12s} Nxe2 {+7.96/16 19s} 41.
Qxe2 {-10.49/25 27s} a4 {+8.16/16 21s} 42. Kg4 {-16.03/28 161s} a3
{+9.06/15 29s} 43. h4 {-17.47/30 129s} Ne3+ {+12.29/17 26s} 44. Kf3
{-17.56/25 7.8s} b4 {+14.45/17 14s} 45. Qe1 {-23.14/26 103s} bxc3
{+22.29/13 9.9s} 46. Qxc3 {-M24/39 36s} a2 {+26.11/13 26s} 47. h5 {-M18/68
4.0s} Rb3 {+24.01/12 25s} 48. hxg6+ {-M18/75 6.0s} Kxg6 {+M19/11 23s} 49.
Qxb3 {-M16/88 5.9s} a1=Q {+M15/13 12s} 50. Qg8+ {-M14/187 7.5s} Rg7
{+M13/52 4.9s} 51. f5+ {-M12/184 11s} Kh5 {+M11/54 13s} 52. Qe8+ {-M10/182
22s} Kg5 {+M9/50 4.8s} 53. Qh5+ {-M8/245 0.009s} Kxh5 {+M7/4 0.001s} 54.
g4+ {-M6/245 0.007s} Rxg4 {+M5/52 13s} 55. Ke2 {-M4/245 0.005s} Qd1+
{+M3/51 19s} 56. Kf2 {-M2/245 0.003s} Qf1# {+M1/50 13s, Black mates} 0-1
[/pgn]

[Rebel]

Match cancelled after 50 games, apparently a full piece behind is needed to get Stockfish on its knees. Results, cross table and download games at my forum.

According to my tests, Komodo Dragon 2 performs around CCRL 2600 level in Rapid when giving knight odds to other engines (except for your Benjamin, which is much better than other similarly rated engines with the knight odds). Although MCTS does better against humans, the standard version does better in giving knight odds to other engines (by roughly the elo gap between the two), so you should use standard. Since only one of your engines is below 2600 (a bit under 2400), and other a bit under 2700, with the rest much higher, I would expect Dragon to score about 1.5 against those ten opponents at knight odds. If you don't already have Dragon 2, we can send it to you. What your tests do show is that a piece is a much bigger handicap than three pawns. It looks like my claim that a knight is worth 4 pawns in the opening is validated.

I have my pawn value at 0.75 since the early 80's

I don't know whether Stockfish 13 or Dragon 2 will do better giving knight odds to other engines, but I know that setting Contempt to 100 will help Dragon when giving knight odds, whereas I don't know about Stockfish. If you really want to compare them at knight odds, I think you need to pick engines rated in the 2500 to 2700 range on the CCRL Rapid list. It's just a waste of time to try to give knight odds in Rapid to engines rated over 3000; knight odds is roughly a thousand elo handicap.

OK, will organize such a 2500-2700 elo pool for the Dragon. I leave it up to you if you want Dragon 1 or 2 to play. In round-3 (minus wpe2/d2/c2) Stockfish had a lot more compensation for the 3 pawns than add Night and Bishop odds due to the space advantage. It was able to win against Arasan, Senpai and Texel. So I suggest to include that position too.

[lkaufman]

Match cancelled after 50 games, apparently a full piece behind is needed to get Stockfish on its knees. Results, cross table and download games at my forum.

According to my tests, Komodo Dragon 2 performs around CCRL 2600 level in Rapid when giving knight odds to other engines (except for your Benjamin, which is much better than other similarly rated engines with the knight odds). Although MCTS does better against humans, the standard version does better in giving knight odds to other engines (by roughly the elo gap between the two), so you should use standard. Since only one of your engines is below 2600 (a bit under 2400), and other a bit under 2700, with the rest much higher, I would expect Dragon to score about 1.5 against those ten opponents at knight odds. If you don't already have Dragon 2, we can send it to you. What your tests do show is that a piece is a much bigger handicap than three pawns. It looks like my claim that a knight is worth 4 pawns in the opening is validated.

I have my pawn value at 0.75 since the early 80's

I don't know whether Stockfish 13 or Dragon 2 will do better giving knight odds to other engines, but I know that setting Contempt to 100 will help Dragon when giving knight odds, whereas I don't know about Stockfish. If you really want to compare them at knight odds, I think you need to pick engines rated in the 2500 to 2700 range on the CCRL Rapid list. It's just a waste of time to try to give knight odds in Rapid to engines rated over 3000; knight odds is roughly a thousand elo handicap.

OK, will organize such a 2500-2700 elo pool for the Dragon. I leave it up to you if you want Dragon 1 or 2 to play. In round-3 (minus wpe2/d2/c2) Stockfish had a lot more compensation for the 3 pawns than add Night and Bishop odds due to the space advantage. It was able to win against Arasan, Senpai and Texel. So I suggest to include that position too.

Please use Dragon 2 (let me know if you don't have it), though the difference is small in this context. I don't object to including the three pawn handicap, but it's probably more like a 700 elo handicap than a 1000 elo handicap, so if you do that you might extend the upper limit to 2800 as otherwise Dragon and SF will both score very heavily at that handicap. Regarding knight=four pawns (in opening or middlegame, say with queens on the board), I recall convincing the Spracklens to modify Fidelity to score that way (with queens on) in the late 1980s, but apparently you had the idea even earlier.

[lkaufman]

Some comments:

1)I read rating of the opponents 2400-3200 so I expected average rating near 2800
Practically the opponents were stronger and the median is 3073(I guess Texel1.7 is Texel1.07 from the CCRL because I do not see 1.7 version).

Practically the 9 non buggy opponents are:
Ethereal 12.75 64-bit 3320 +12 −12 46.9% +22.9 60.7% 2004
Arasan 22.2 64-bit 3159 +22 −22 54.0% −23.1 51.9% 592
Vajolet2 2.8 64-bit 3098 +15 −15 48.9% +5.8 53.4% 1355
Komodo 3 64-bit 3089 +14 −14 58.5% −56.6 44.3% 1737
Texel 1.07 64-bit 3073 +11 −11 49.8% −2.0 47.8% 2662
Cheng 4.41 64-bit 2949 +94 −98 35.0% +90.3 56.7% 30 at 40/15(not enough games at 40/15 so I looked also at blitz rating)
Cheng 4.41 64-bit 3058 +73 −71 57.8% −51.5 37.5% 64 at blitz

Senpai 1.0 64-bit 2960 +13 −13 47.7% +14.5 47.3% 2072
ProDeo 3.1 2795 +31 −31 50.6% −3.8 40.1% 344
Fruit 2.1 2695 +25 −25 51.5% −13.9 37.2% 522

2)I believe that latest stockfish is stronger.

Here are games from my own knight handicap tournament of later version of Stockfish against fruit and first Rybka.
both default stockfish with 3 cores and stockfish with 3 cores and contempt 100 made result of 1-1 at 40 moves/13 minutes

[pgn][Event "My Tournament"]
[Site "?"]
[Date "2021.05.14"]
[Round "1"]
[White "stockfish_21051119_x64contempt100"]
[Black "Rybka v1.0 Beta.x64"]
[Result "1-0"]
[TimeControl "40/780:300"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "77"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {-4.94/36 58s} Nf6 {+1.42/14 20s} 2. g3 {-4.71/28 4.8s} d5 {+1.58/15
21s} 3. Bg2 {-4.85/30 7.0s} Nc6 {+1.57/14 4.1s} 4. c4 {-4.58/32 22s} dxc4
{+1.83/15 14s} 5. d5 {-4.43/32 49s} Ne5 {+2.40/15 14s} 6. Qd4 {-4.35/32
4.5s} Ned7 {+3.95/15 13s} 7. Nc3 {-4.55/35 85s} Nb6 {+4.12/13 5.1s} 8. Be3
{-4.63/31 14s} Bd7 {+4.28/15 19s} 9. O-O {-4.50/29 21s} a5 {+4.29/13 15s}
10. Rad1 {-3.55/28 12s} Qc8 {+4.35/14 16s} 11. Qf4 {-3.21/30 18s} h6
{+4.30/14 19s} 12. Bd4 {-2.54/28 4.5s} Ra6 {+4.09/14 20s} 13. e4 {-2.62/29
11s} Nh7 {+4.28/14 23s} 14. Qe3 {-2.06/35 35s} Ng5 {+4.33/14 17s} 15. f4
{-0.91/30 12s} Nh7 {+3.48/15 45s} 16. e5 {-0.47/30 19s} Bg4 {+2.02/14 26s}
17. d6 {+1.50/30 4.9s} Na8 {+2.02/14 16s} 18. Ne4 {+1.37/30 9.4s} cxd6
{+1.17/13 13s} 19. exd6 {+1.59/37 7.4s} Rc6 {-0.20/15 171s} 20. Bc5
{+3.06/29 7.7s} e5 {-0.35/14 6.7s} 21. Rd5 {+3.31/30 8.1s} Kd8 {+0.06/13
12s} 22. fxe5 {+3.63/31 12s} Be6 {-0.62/13 11s} 23. Rdd1 {+3.95/31 11s} Qd7
{-0.67/13 12s} 24. Ba3 {+5.80/29 12s} b6 {-0.43/14 10s} 25. Nc3 {+6.73/27
10s} Ke8 {-0.70/13 7.5s} 26. Nb5 {+7.22/28 9.8s} Rc8 {-1.27/14 9.3s} 27.
Nd4 {+7.51/30 11s} Kd8 {-1.34/14 12s} 28. Rd2 {+9.01/26 11s} c3 {-3.18/12
11s} 29. bxc3 {+9.94/28 14s} f6 {-3.65/12 9.1s} 30. exf6 {+12.78/27 18s}
Ng5 {-5.47/12 19s} 31. h4 {+14.37/27 14s} gxf6 {-7.56/14 79s} 32. hxg5
{+18.05/26 19s} Bc4 {-9.10/14 39s} 33. Qf4 {+21.01/25 16s} Qf7 {-9.06/10
5.8s} 34. g6 {+M19/54 16s} Qxg6 {-17.10/11 17s} 35. d7 {+M17/76 18s} Qxg3
{-M384/3 0.031s} 36. Qxf6+ {+M9/245 15s} Kxd7 37. Ne6+ {+M7/245 0.37s} Bd5
38. Qf7+ {+M5/245 0.009s} Kc6 39. Nd4# {+M1/245 0.004s, White mates} 1-0[/pgn]

[pgn][Event "My Tournament"]
[Site "?"]
[Date "2021.05.14"]
[Round "2"]
[White "stockfish_21051119_x64_avx2"]
[Black "Rybka v1.0 Beta.x64"]
[Result "0-1"]
[TimeControl "40/780:300"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "136"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {-5.40/36 60s} Nf6 {+1.42/14 20s} 2. g3 {-5.44/31 5.7s} d5 {+1.58/15
21s} 3. Bg2 {-5.42/32 12s} Nc6 {+1.57/14 4.1s} 4. O-O {-5.13/31 13s} h5
{+1.89/15 15s} 5. c4 {-4.82/26 5.7s} dxc4 {+2.01/14 15s} 6. d5 {-4.88/33
15s} Ne5 {+2.26/15 12s} 7. Qd4 {-4.76/32 7.6s} Qd6 {+2.12/15 17s} 8. Bf4
{-3.66/29 8.1s} Nfd7 {+2.19/16 19s} 9. Nc3 {-3.82/33 40s} c5 {+3.87/16 17s}
10. Qe3 {-3.47/26 6.2s} f6 {+3.52/16 35s} 11. a4 {-3.09/30 17s} g5
{+3.98/14 14s} 12. Nb5 {-3.12/32 8.2s} Qb6 {+3.45/14 4.0s} 13. Bxe5
{-3.66/37 25s} Nxe5 {+3.00/16 37s} 14. d6 {-3.77/36 7.9s} Rb8 {+3.41/15
39s} 15. Rad1 {-2.50/31 9.2s} Bd7 {+3.86/14 19s} 16. f4 {-2.79/35 27s} gxf4
{+4.19/14 9.7s} 17. Nc7+ {-3.00/36 18s} Kd8 {+4.40/15 17s} 18. gxf4
{-2.62/34 11s} exd6 {+2.75/15 41s} 19. Nd5 {-3.78/36 28s} Ng4 {+3.18/15
21s} 20. Qc3 {-4.12/37 34s} Qb3 {+3.36/16 10s} 21. Qa5+ {-4.12/39 14s} b6
{+3.57/17 31s} 22. Qxa7 {-4.16/37 15s} Rc8 {+3.16/16 25s} 23. Nxb6
{-4.25/39 16s} Rc7 {+3.95/15 21s} 24. Qa8+ {-4.35/38 11s} Bc8 {+3.95/15
12s} 25. Nxc8 {-4.55/42 25s} Rxc8 {+3.95/15 23s} 26. Qa6 {-4.28/37 12s} Rh7
{+3.95/14 11s} 27. h3 {-4.04/36 13s} Qe3+ {+3.95/13 3.6s} 28. Kh1 {-5.59/40
40s} Nf2+ {+3.95/13 3.8s} 29. Rxf2 {-5.48/37 12s} Qxf2 {+3.95/14 7.9s} 30.
Bf3 {-5.91/40 57s} Rd7 {+4.18/15 57s} 31. Qxc4 {-6.01/35 11s} Qg3 {+4.24/15
28s} 32. Qe6 {-5.57/34 11s} Qxf4 {+4.27/14 9.2s} 33. a5 {-6.02/39 71s} Rb8
{+4.65/13 12s} 34. a6 {-5.98/36 46s} Qa4 {+4.69/14 15s} 35. Qd5 {-6.25/33
31s} h4 {+4.97/15 61s} 36. Qd3 {-6.23/31 19s} Rb6 {+5.40/13 6.9s} 37. Rg1
{-6.45/29 10s} Qxa6 {+4.71/14 10.0s} 38. Qxa6 {-6.54/27 3.0s} Rxa6
{+5.51/18 13s} 39. Rg4 {-6.65/23 1.8s} Rh7 {+5.57/17 12s} 40. Rf4 {-6.06/18
0.35s} Be7 {+6.14/18 14s} 41. Bd5 {-7.40/37 102s} Ra1+ {+6.15/19 22s} 42.
Kg2 {-7.59/31 14s} Rd1 {+6.33/19 14s} 43. Be6 {-8.18/36 144s} Rd2 {+6.33/18
21s} 44. Kf3 {-7.78/31 22s} Rxb2 {+6.34/17 21s} 45. Bc4 {-7.86/35 40s} Rb4
{+7.05/18 13s} 46. Ba2 {-9.12/35 85s} Rg7 {+8.80/20 24s} 47. Re4 {-12.01/38
87s} Rg3+ {+10.27/19 23s} 48. Kf4 {-23.31/27 13s} Rxe4+ {+10.27/19 12s} 49.
Kxe4 {-39.18/30 22s} Rxh3 {+11.13/20 24s} 50. e3 {-62.49/34 62s} f5+
{+18.77/19 21s} 51. Kxf5 {-M34/60 8.3s} Rxe3 {+M384/19 3.1s} 52. Kf4
{-M32/66 5.5s} Rd3 {+M384/16 0.034s} 53. Ke4 {-M24/71 4.1s} c4 {+M384/11
0.015s} 54. Bxc4 {-M22/72 4.0s} d5+ {+M384/9 0.002s} 55. Bxd5 {-M20/85
4.2s} Rxd5 {+M384/9 0.002s} 56. Kxd5 {-M18/102 4.4s} h3 {+M384/7 0.001s}
57. Kc4 {-M16/77 4.3s} h2 {+M384/7 0.012s} 58. Kd4 {-M14/81 4.3s} h1=Q
{+M384/4 0s} 59. Ke3 {-M12/156 4.4s} Qd5 {+M384/5 0.004s} 60. Kf4 {-M10/245
2.8s} Bc5 {+M384/3 0s} 61. Kg4 {-M10/245 3.6s} Qe4+ {+M384/3 0.003s} 62.
Kg5 {-M8/245 0.22s} Be7+ {0s} 63. Kh6 {-M6/245 0.006s} Bf6 64. Kh5 {-M6/245
0.024s} Qf5+ 65. Kh6 {-M4/245 0.005s} Qg5+ 66. Kh7 {-M2/245 0.003s} Qh5+
67. Kg8 {-M4/245 0.005s} Qg6+ 68. Kf8 {-M2/245 0.003s} Be7# {Black mates}
0-1
[/pgn]

[pgn][Event "My Tournament"]
[Site "?"]
[Date "2021.05.14"]
[Round "3"]
[White "stockfish_21051119_x64_avx2"]
[Black "Fruit2.2.1"]
[Result "1-0"]
[TimeControl "40/780:300"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "95"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {-5.42/35 54s} Nf6 {+2.77/15 17s} 2. g3 {-5.32/33 16s} d5 {+3.06/16
39s} 3. Bg2 {-5.44/30 6.8s} Nc6 {+3.10/15 19s} 4. O-O {-5.26/31 18s} e6
{+3.05/16 45s} 5. b3 {-5.05/31 7.1s} Be7 {+3.16/14 23s} 6. e3 {-5.21/36
69s} h5 {+3.27/14 20s} 7. h3 {-4.86/28 8.4s} h4 {+3.63/14 13s} 8. g4
{-4.85/29 9.1s} g5 {+3.38/14 23s} 9. c4 {-3.99/30 16s} b6 {+3.00/14 27s}
10. Bb2 {-3.90/32 28s} Ba6 {+3.03/14 23s} 11. Nd2 {-3.61/32 17s} Qd7
{+3.28/13 15s} 12. a3 {-3.56/32 13s} O-O-O {+3.35/13 12s} 13. Qc2 {-3.79/33
32s} Kb7 {+3.30/13 12s} 14. Rfc1 {-3.17/34 27s} Rh6 {+2.98/14 28s} 15. Rab1
{-2.65/31 10s} Rc8 {+3.03/14 17s} 16. b4 {-1.75/30 5.1s} dxc4 {+2.74/15
14s} 17. Bc3 {-2.36/36 21s} b5 {+2.60/15 22s} 18. a4 {-2.63/41 71s} Ka8
{+3.05/15 12s} 19. Qa2 {-2.22/33 13s} Nd5 {+2.96/15 14s} 20. axb5 {-2.06/34
6.3s} Bxb5 {+2.91/15 21s} 21. Nxc4 {-1.96/33 9.3s} Rb8 {+2.79/13 17s} 22.
Bd2 {-2.21/39 39s} f6 {+2.57/13 21s} 23. Na3 {+0.39/32 5.5s} Bd3 {+1.94/14
19s} 24. Rb3 {+1.22/33 9.9s} Bg6 {+1.52/15 18s} 25. b5 {+6.71/28 8.1s} Nd8
{+0.56/15 22s} 26. Nc4 {+7.12/28 8.5s} Rb7 {+0.55/14 12s} 27. e4 {+7.44/31
8.3s} Nf4 {+0.13/14 21s} 28. Bxf4 {+8.67/27 9.3s} gxf4 {-2.23/14 48s} 29.
b6 {+8.70/28 9.4s} f3 {-4.34/15 88s} 30. Bxf3 {+9.55/32 10s} c6 {-5.09/14
46s} 31. Na5 {+10.32/30 11s} Rh7 {-5.29/13 10s} 32. d5 {+12.25/29 19s} Rxb6
{-3.87/12 3.9s} 33. Rxb6 {+12.72/33 16s} Bc5 {-8.32/13 18s} 34. Rxc5
{+16.24/27 19s} Qc7 {-11.59/13 9.3s} 35. Ra6 {+18.60/26 14s} Be8 {-8.84/11
3.3s} 36. Nxc6 {+30.82/27 16s} Bxc6 {-8.96/12 1.6s} 37. dxc6 {+M27/78 16s}
e5 {-11.26/13 2.6s} 38. Qd5 {+M23/67 19s} Kb8 {-12.74/11 0.64s} 39. Be2
{+M21/76 24s} Kc8 {-12.16/10 1.2s} 40. Ra3 {+M19/78 22s} Qf7 {-M18/9 0.91s}
41. Qd2 {+M15/84 12s} Qc7 {-M14/12 16s} 42. Rb3 {+M13/89 13s} Qb6 {-M12/24
15s} 43. Qd6 {+M11/120 12s} Rc7 {-M10/56 17s} 44. Rxb6 {+M9/245 9.1s} axb6
{-M8/54 14s} 45. Ba6+ {+M7/245 0.063s} Kb8 {-M6/54 14s} 46. Qxd8+ {+M5/245
0.006s} Ka7 {-M4/53 14s} 47. Qxc7+ {+M3/245 0.004s} Kxa6 {-M2/52 14s} 48.
Qb7# {+M1/245 0.002s, White mates} 1-0
[/pgn]

[pgn][Event "My Tournament"]
[Site "?"]
[Date "2021.05.14"]
[Round "4"]
[White "stockfish_21051119_x64contempt100"]
[Black "Fruit2.2.1"]
[Result "0-1"]
[TimeControl "40/780:300"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "112"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {-4.71/35 70s} Nf6 {+2.77/15 16s} 2. g3 {-4.77/30 12s} d5 {+3.06/16
36s} 3. Bg2 {-4.63/29 5.3s} Nc6 {+3.10/15 18s} 4. c4 {-4.65/31 16s} dxc4
{+3.06/16 27s} 5. d5 {-4.86/30 23s} Ne5 {+2.88/15 26s} 6. Qd4 {-4.50/31
16s} Ned7 {+3.10/15 17s} 7. Qxc4 {-4.59/32 74s} Nb6 {+2.99/16 19s} 8. Qc2
{-4.59/31 4.8s} Nbxd5 {+3.38/15 22s} 9. O-O {-4.66/32 31s} e6 {+3.34/15
21s} 10. Rd1 {-4.55/29 11s} Bd6 {+3.40/15 18s} 11. Na3 {-5.00/34 79s} O-O
{+3.30/15 17s} 12. b3 {-4.78/30 7.3s} Ng4 {+3.64/13 12s} 13. Nc4 {-5.01/36
46s} Qf6 {+4.50/13 11s} 14. e3 {-4.99/33 4.1s} Qxa1 {+4.08/15 44s} 15. Bb2
{-5.05/34 8.0s} Nb4 {+4.46/16 11s} 16. Qd2 {-5.04/35 9.9s} Qxa2 {+4.48/16
11s} 17. Qc3 {-5.08/35 9.1s} Nf6 {+4.52/16 22s} 18. Ra1 {-5.09/35 8.6s}
Qxa1+ {+4.47/16 21s} 19. Bxa1 {-5.15/32 7.1s} Be7 {+4.50/17 20s} 20. Qc1
{-5.17/30 6.9s} Rd8 {+4.49/15 12s} 21. Na5 {-4.98/34 33s} Nfd5 {+4.66/14
12s} 22. Be4 {-5.12/34 57s} f6 {+4.72/14 20s} 23. Qb1 {-5.22/31 5.3s} h6
{+4.70/15 14s} 24. Bf3 {-5.31/31 7.5s} e5 {+4.87/14 15s} 25. e4 {-5.44/37
31s} Nb6 {+4.76/17 16s} 26. Bc3 {-5.36/31 8.3s} Rb8 {+4.81/17 20s} 27. Kg2
{-5.69/35 49s} Be6 {+4.84/16 15s} 28. Qa1 {-5.63/32 16s} Nc2 {+5.29/16 18s}
29. Qb2 {-5.62/30 10s} Nd4 {+5.37/16 20s} 30. Bh5 {-5.82/27 5.7s} Kh7
{+5.28/15 21s} 31. Qa1 {-5.80/33 35s} c6 {+5.38/14 18s} 32. f4 {-6.17/30
25s} g6 {+5.58/15 17s} 33. Bd1 {-6.69/32 18s} Rd7 {+5.57/15 18s} 34. Nc4
{-6.88/26 11s} Bxc4 {+5.94/14 14s} 35. bxc4 {-6.48/23 1.8s} Nxc4 {+6.00/15
26s} 36. Kh3 {-6.98/26 8.7s} Bc5 {+6.52/16 20s} 37. Bg4 {-7.06/24 3.3s} Rc7
{+6.44/15 26s} 38. Qc1 {-7.37/22 3.1s} b5 {+6.62/15 14s} 39. Qf1 {-7.40/21
0.61s} a5 {+7.09/15 16s} 40. Be2 {-7.70/19 0.12s} Nxe2 {+7.96/16 19s} 41.
Qxe2 {-10.49/25 27s} a4 {+8.16/16 21s} 42. Kg4 {-16.03/28 161s} a3
{+9.06/15 29s} 43. h4 {-17.47/30 129s} Ne3+ {+12.29/17 26s} 44. Kf3
{-17.56/25 7.8s} b4 {+14.45/17 14s} 45. Qe1 {-23.14/26 103s} bxc3
{+22.29/13 9.9s} 46. Qxc3 {-M24/39 36s} a2 {+26.11/13 26s} 47. h5 {-M18/68
4.0s} Rb3 {+24.01/12 25s} 48. hxg6+ {-M18/75 6.0s} Kxg6 {+M19/11 23s} 49.
Qxb3 {-M16/88 5.9s} a1=Q {+M15/13 12s} 50. Qg8+ {-M14/187 7.5s} Rg7
{+M13/52 4.9s} 51. f5+ {-M12/184 11s} Kh5 {+M11/54 13s} 52. Qe8+ {-M10/182
22s} Kg5 {+M9/50 4.8s} 53. Qh5+ {-M8/245 0.009s} Kxh5 {+M7/4 0.001s} 54.
g4+ {-M6/245 0.007s} Rxg4 {+M5/52 13s} 55. Ke2 {-M4/245 0.005s} Qd1+
{+M3/51 19s} 56. Kf2 {-M2/245 0.003s} Qf1# {+M1/50 13s, Black mates} 0-1
[/pgn]

Rybka 1 does seem to be relatively weak (for its rating) receiving knight odds. I've now run a bunch of 200 game matches between Dragon 2 (Contempt 100) and various engines at CCRL blitz (2' + 1") with Dragon giving knight odds (using the middle 200 positions from ChrisW set so all engines play the same openings), with Dragon performing around 2760 (CCRL blitz scale) against three engines (Wasp 2, Wasp 1.01, Fruit 2.2.1), but much better against Naum 4 (reportedly with eval similar to Rybka) and Rybka 1 and much worse against Benjamin. The difference in performance rating against Naum 4 or Rybka 1 compared to Benjamin is an astonishing 400 elo!! Of course results will be worse at Rapid (15' + 10, roughly CCRL Rapid), but this applies to all the opponents, roughly subtracting 150 or so will predict the Rapid results well. Dragon lost by just 45 elo in Rapid to Fruit 2.2.1 at knight odds in Rapid, and I expect a close match with Rybka 1 in Rapid when I get that match in. I would never have imagined in 2007 that any version of Rybka could ever have a tough time winning a match with knight odds against any future engine at a somewhat serious time control, given that even Rybka 1 on one core was already probably stronger than the best human player in Rapid chess (probably Anand at that time). It is really incredible.

[mar]

Cheng 4.41 64-bit 2949 +94 −98 35.0% +90.3 56.7% 30 at 40/15(not enough games at 40/15 so I looked also at blitz rating)
Cheng 4.41 64-bit 3058 +73 −71 57.8% −51.5 37.5% 64 at blitz

I certainly hope that 2949 won't hold that would mean -12
a crazy bad start at CCRL 40/15 so far, but CEGT's 40/20 looks way better though so I will wait for 1k games to talk rating
not that I can do anything if 4.41 finishes worse than 4.40, just shrug and move on