Komodo Dragon 2.6 released

[Chessqueen]

This time I set Komodo Dragon UCI_ELO to 2325 versus Dumb rated 2296 and it still lost

In my Rapid testing Dragon set to 2300 is only about an even match with Snowy (rated about 2000 Rapid by ccrl), but around the 2850 level Dragon matches CCRL. This is based on the notion that 100 elo gap on ccrl is only about 60 elo gap in human ratings. Whether that is accurate is the question.

OliThink rated 2846 is giving Komodo Dragon 2.6 UCI_ELO 2850 a very hard time, Out of 10 games it has won 7 games, but Carlsen lost a game to 2600 rated players at Rapid, so there is nothing to worry about

[pgn][Event "Computer chess game"]
[Date "2021.12.29"]
[Round "10"]
[White "Dragon-2.6-64bit-avx2"]
[Black "OliThink5101"]
[Result "0-1"]
[BlackElo "2846"]
[ECO "D01"]
[Opening "Richter-Veresov Attack"]
[Time "16:31:32"]
[Variation "1...d5"]
[WhiteElo "2850"]
[TimeControl "900+10"]
[Termination "normal"]
[PlyCount "119"]
[WhiteType "program"]
[BlackType "program"]

1. Nc3 d5 2. d4 Nf6 3. e3 e6 4. Nf3 a6 5. Bd3 c5 6. O-O c4 7. Be2 b5 8. Ne5
Bb7 9. f4 Nc6 10. Bf3 Be7 11. Ne2 Ne4 12. b3 Qc8 13. Bxe4 dxe4 14. Ng3 O-O
15. bxc4 Nxe5 16. fxe5 Qxc4 17. Nh5 Kh8 18. Bd2 a5 19. Rc1 Bd5 20. Qg4 g6
21. Nf6 Bxf6 22. Rxf6 Qxa2 23. Qh4 Kg8 24. Qh6 b4 25. Rf4 Ra7 26. h4 Bc4
27. h5 Rb8 28. Rf2 Qa4 29. Rf4 Qe8 30. Rxe4 a4 31. Rf4 a3 32. Ra1 a2 33.
Kf2 b3 34. cxb3 Rxb3 35. hxg6 fxg6 36. Qh1 Rab7 37. Rxa2 Rb1 38. Qh4 Rf1+
39. Kg3 Rxf4 40. Qxf4 Bxa2 41. e4 Qa4 42. Be3 Rf7 43. Qg4 Qa3 44. Qe2 Bb1
45. d5 Qc3 46. Qd2 Qxe5+ 47. Bf4 Qxe4 48. dxe6 Qxe6 49. Qd6 Qe1+ 50. Kh3
Qh1+ 51. Kg3 h5 52. Qd2 h4+ 53. Kg4 Bf5+ 54. Kg5 h3 55. Be5 Bd7 56. Qc2
Rf5+ 57. Kxg6 Rxe5 58. Qc4+ Re6+ 59. Qxe6+ Bxe6 60. Kf6 {White resigns} *[/pgn]

[Chessqueen]

Ceibo 0.8 is rated 2650 vs Komodo Dragon UCI_ELO = 2850. After 6 games the score is even against Ceibo 0-8

[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2021.12.29"]
[Round "?"]
[White "Ceibo_0.8"]
[Black "Dragon-2.6-64bit-avx2"]
[Result "1/2-1/2"]
[BlackElo "2850"]
[ECO "C42"]
[Opening "Russian Game"]
[Time "20:55:42"]
[Variation "3.Nxe5"]
[WhiteElo "2650"]
[TimeControl "900+10"]
[Termination "normal"]
[PlyCount "190"]
[WhiteType "program"]
[BlackType "program"]

1. e4 e5 2. Nf3 Nf6 3. Nxe5 Qe7 4. d4 Nc6 5. Nf3 Nxe4 6. Be2 d5 7. O-O Bf5
8. Re1 O-O-O 9. c4 dxc4 10. Be3 Qe6 11. Qc1 Na5 12. Nc3 Bb4 13. Nh4 Rhe8
14. Bf3 Bxc3 15. bxc3 Qf6 16. Nxf5 Qxf5 17. Qb1 Re6 18. Bh6 Rxh6 19. Rxe4
Rb6 20. Bg4 Rxb1+ 21. Rxb1 Qxg4 22. Rxg4 g6 23. Rf4 f5 24. Rf3 Nc6 25. Re3
a6 26. g3 b5 27. a4 Rd5 28. Re8+ Kb7 29. Kg2 Kb6 30. Rg8 Rd6 31. Rg7 h5 32.
h4 Nd8 33. Ra1 Nc6 34. Rc1 Nd8 35. Re1 Nc6 36. axb5 axb5 37. Rc1 Nd8 38.
Re7 Nc6 39. Rf7 Nd8 40. Rg7 Nc6 41. f4 Kb7 42. Rc2 Kb6 43. Kf2 Nd8 44. Rc1
Nc6 45. Ke1 Na5 46. Rg8 Nb3 47. Rc2 Kb7 48. Kf2 Na5 49. Rg7 Kb6 50. Re2 Nb3
51. Ke1 b4 52. Rc2 bxc3 53. Rxc3 Rxd4 54. Rxg6+ Kc5 55. Rg7 Re4+ 56. Kf2
Kd4 57. Rf3 Nc5 58. Rxc7 Nd3+ 59. Kg1 Re1+ 60. Rf1 Re3 61. Kg2 Re2+ 62. Kg1
c3 63. g4 hxg4 64. h5 c2 65. h6 c1=Q 66. Rfxc1 Nxc1 67. Rxc1 g3 68. Rc6 Kd5
69. Rc7 Rh2 70. Re7 Rxh6 71. Re5+ Kd6 72. Kg2 Rh5 73. Kxg3 Kc6 74. Kf3 Kc7
75. Kf2 Kc6 76. Ke3 Kd6 77. Ke2 Rh1 78. Rxf5 Ke6 79. Rc5 Kf6 80. Rb5 Rc1
81. Kf3 Rf1+ 82. Ke3 Re1+ 83. Kf2 Rc1 84. Rg5 Rc4 85. Kf3 Rc1 86. Rh5 Rf1+
87. Ke3 Rc1 88. Ra5 Re1+ 89. Kf2 Rd1 90. Kf3 Rf1+ 91. Ke3 Re1+ 92. Kf2 Rc1
93. Kf3 Rf1+ 94. Kg4 Rg1+ 95. Kf3 Rf1+ {3-fold repetition} 1/2-1/2[/pgn]

[lkaufman]

Ceibo 0.8 is rated 2650 vs Komodo Dragon UCI_ELO = 2850. After 6 games the score is even against Ceibo 0-8

[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2021.12.29"]
[Round "?"]
[White "Ceibo_0.8"]
[Black "Dragon-2.6-64bit-avx2"]
[Result "1/2-1/2"]
[BlackElo "2850"]
[ECO "C42"]
[Opening "Russian Game"]
[Time "20:55:42"]
[Variation "3.Nxe5"]
[WhiteElo "2650"]
[TimeControl "900+10"]
[Termination "normal"]
[PlyCount "190"]
[WhiteType "program"]
[BlackType "program"]

1. e4 e5 2. Nf3 Nf6 3. Nxe5 Qe7 4. d4 Nc6 5. Nf3 Nxe4 6. Be2 d5 7. O-O Bf5
8. Re1 O-O-O 9. c4 dxc4 10. Be3 Qe6 11. Qc1 Na5 12. Nc3 Bb4 13. Nh4 Rhe8
14. Bf3 Bxc3 15. bxc3 Qf6 16. Nxf5 Qxf5 17. Qb1 Re6 18. Bh6 Rxh6 19. Rxe4
Rb6 20. Bg4 Rxb1+ 21. Rxb1 Qxg4 22. Rxg4 g6 23. Rf4 f5 24. Rf3 Nc6 25. Re3
a6 26. g3 b5 27. a4 Rd5 28. Re8+ Kb7 29. Kg2 Kb6 30. Rg8 Rd6 31. Rg7 h5 32.
h4 Nd8 33. Ra1 Nc6 34. Rc1 Nd8 35. Re1 Nc6 36. axb5 axb5 37. Rc1 Nd8 38.
Re7 Nc6 39. Rf7 Nd8 40. Rg7 Nc6 41. f4 Kb7 42. Rc2 Kb6 43. Kf2 Nd8 44. Rc1
Nc6 45. Ke1 Na5 46. Rg8 Nb3 47. Rc2 Kb7 48. Kf2 Na5 49. Rg7 Kb6 50. Re2 Nb3
51. Ke1 b4 52. Rc2 bxc3 53. Rxc3 Rxd4 54. Rxg6+ Kc5 55. Rg7 Re4+ 56. Kf2
Kd4 57. Rf3 Nc5 58. Rxc7 Nd3+ 59. Kg1 Re1+ 60. Rf1 Re3 61. Kg2 Re2+ 62. Kg1
c3 63. g4 hxg4 64. h5 c2 65. h6 c1=Q 66. Rfxc1 Nxc1 67. Rxc1 g3 68. Rc6 Kd5
69. Rc7 Rh2 70. Re7 Rxh6 71. Re5+ Kd6 72. Kg2 Rh5 73. Kxg3 Kc6 74. Kf3 Kc7
75. Kf2 Kc6 76. Ke3 Kd6 77. Ke2 Rh1 78. Rxf5 Ke6 79. Rc5 Kf6 80. Rb5 Rc1
81. Kf3 Rf1+ 82. Ke3 Re1+ 83. Kf2 Rc1 84. Rg5 Rc4 85. Kf3 Rc1 86. Rh5 Rf1+
87. Ke3 Rc1 88. Ra5 Re1+ 89. Kf2 Rd1 90. Kf3 Rf1+ 91. Ke3 Re1+ 92. Kf2 Rc1
93. Kf3 Rf1+ 94. Kg4 Rg1+ 95. Kf3 Rf1+ {3-fold repetition} 1/2-1/2[/pgn]

Overnight I ran a 500 game Rapid match between Dragon 2.6 with Elo set to 2700 and Benjamin 1.0 (CCRL rapid rating 2699). Benjamin won the match by 63 elo points. So yes, even at 2700 level Dragon plays below the level of the corresponding CCRL Elo Rapid rating, but I don't think that this means that it would lose a Rapid match to a 2700 human; I'm pretty sure that Benjamin 1.0 would win a Rapid match from a 2700 Elo human. Also, based on my own experience, it is harder for me to score well against a suitably rated Dragon level than for me to score well against a weak CCRL engine that is even with that Dragon level. Probably this is because the extra depth of the classical engine is more helpful against crippled Dragon than against a human oopponent who is relying mostly on general principles and pattern recognition. The weak CCRL engine often makes positional errors that I can exploit, while Dragon searching just a couple plies may make some tactical errors but I may not spot them in a fast game.

[Chessqueen]

Ceibo 0.8 is rated 2650 vs Komodo Dragon UCI_ELO = 2850. After 6 games the score is even against Ceibo 0-8

[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2021.12.29"]
[Round "?"]
[White "Ceibo_0.8"]
[Black "Dragon-2.6-64bit-avx2"]
[Result "1/2-1/2"]
[BlackElo "2850"]
[ECO "C42"]
[Opening "Russian Game"]
[Time "20:55:42"]
[Variation "3.Nxe5"]
[WhiteElo "2650"]
[TimeControl "900+10"]
[Termination "normal"]
[PlyCount "190"]
[WhiteType "program"]
[BlackType "program"]

1. e4 e5 2. Nf3 Nf6 3. Nxe5 Qe7 4. d4 Nc6 5. Nf3 Nxe4 6. Be2 d5 7. O-O Bf5
8. Re1 O-O-O 9. c4 dxc4 10. Be3 Qe6 11. Qc1 Na5 12. Nc3 Bb4 13. Nh4 Rhe8
14. Bf3 Bxc3 15. bxc3 Qf6 16. Nxf5 Qxf5 17. Qb1 Re6 18. Bh6 Rxh6 19. Rxe4
Rb6 20. Bg4 Rxb1+ 21. Rxb1 Qxg4 22. Rxg4 g6 23. Rf4 f5 24. Rf3 Nc6 25. Re3
a6 26. g3 b5 27. a4 Rd5 28. Re8+ Kb7 29. Kg2 Kb6 30. Rg8 Rd6 31. Rg7 h5 32.
h4 Nd8 33. Ra1 Nc6 34. Rc1 Nd8 35. Re1 Nc6 36. axb5 axb5 37. Rc1 Nd8 38.
Re7 Nc6 39. Rf7 Nd8 40. Rg7 Nc6 41. f4 Kb7 42. Rc2 Kb6 43. Kf2 Nd8 44. Rc1
Nc6 45. Ke1 Na5 46. Rg8 Nb3 47. Rc2 Kb7 48. Kf2 Na5 49. Rg7 Kb6 50. Re2 Nb3
51. Ke1 b4 52. Rc2 bxc3 53. Rxc3 Rxd4 54. Rxg6+ Kc5 55. Rg7 Re4+ 56. Kf2
Kd4 57. Rf3 Nc5 58. Rxc7 Nd3+ 59. Kg1 Re1+ 60. Rf1 Re3 61. Kg2 Re2+ 62. Kg1
c3 63. g4 hxg4 64. h5 c2 65. h6 c1=Q 66. Rfxc1 Nxc1 67. Rxc1 g3 68. Rc6 Kd5
69. Rc7 Rh2 70. Re7 Rxh6 71. Re5+ Kd6 72. Kg2 Rh5 73. Kxg3 Kc6 74. Kf3 Kc7
75. Kf2 Kc6 76. Ke3 Kd6 77. Ke2 Rh1 78. Rxf5 Ke6 79. Rc5 Kf6 80. Rb5 Rc1
81. Kf3 Rf1+ 82. Ke3 Re1+ 83. Kf2 Rc1 84. Rg5 Rc4 85. Kf3 Rc1 86. Rh5 Rf1+
87. Ke3 Rc1 88. Ra5 Re1+ 89. Kf2 Rd1 90. Kf3 Rf1+ 91. Ke3 Re1+ 92. Kf2 Rc1
93. Kf3 Rf1+ 94. Kg4 Rg1+ 95. Kf3 Rf1+ {3-fold repetition} 1/2-1/2[/pgn]

Overnight I ran a 500 game Rapid match between Dragon 2.6 with Elo set to 2700 and Benjamin 1.0 (CCRL rapid rating 2699). Benjamin won the match by 63 elo points. So yes, even at 2700 level Dragon plays below the level of the corresponding CCRL Elo Rapid rating, but I don't think that this means that it would lose a Rapid match to a 2700 human; I'm pretty sure that Benjamin 1.0 would win a Rapid match from a 2700 Elo human. Also, based on my own experience, it is harder for me to score well against a suitably rated Dragon level than for me to score well against a weak CCRL engine that is even with that Dragon level. Probably this is because the extra depth of the classical engine is more helpful against crippled Dragon than against a human opponent who is relying mostly on general principles and pattern recognition. The weak CCRL engine often makes positional errors that I can exploit, while Dragon searching just a couple plies may make some tactical errors but I may not spot them in a fast game.

Can you post your best games based on your own experience against a suitably rated Dragon UCI_ELO level, and which games you did not do so good? With some explanations of the key play so we can learn from your games ?

[Chessqueen]

Ceibo 0.8 is rated 2650 vs Komodo Dragon UCI_ELO = 2850. After 6 games the score is even against Ceibo 0-8

[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2021.12.29"]
[Round "?"]
[White "Ceibo_0.8"]
[Black "Dragon-2.6-64bit-avx2"]
[Result "1/2-1/2"]
[BlackElo "2850"]
[ECO "C42"]
[Opening "Russian Game"]
[Time "20:55:42"]
[Variation "3.Nxe5"]
[WhiteElo "2650"]
[TimeControl "900+10"]
[Termination "normal"]
[PlyCount "190"]
[WhiteType "program"]
[BlackType "program"]

1. e4 e5 2. Nf3 Nf6 3. Nxe5 Qe7 4. d4 Nc6 5. Nf3 Nxe4 6. Be2 d5 7. O-O Bf5
8. Re1 O-O-O 9. c4 dxc4 10. Be3 Qe6 11. Qc1 Na5 12. Nc3 Bb4 13. Nh4 Rhe8
14. Bf3 Bxc3 15. bxc3 Qf6 16. Nxf5 Qxf5 17. Qb1 Re6 18. Bh6 Rxh6 19. Rxe4
Rb6 20. Bg4 Rxb1+ 21. Rxb1 Qxg4 22. Rxg4 g6 23. Rf4 f5 24. Rf3 Nc6 25. Re3
a6 26. g3 b5 27. a4 Rd5 28. Re8+ Kb7 29. Kg2 Kb6 30. Rg8 Rd6 31. Rg7 h5 32.
h4 Nd8 33. Ra1 Nc6 34. Rc1 Nd8 35. Re1 Nc6 36. axb5 axb5 37. Rc1 Nd8 38.
Re7 Nc6 39. Rf7 Nd8 40. Rg7 Nc6 41. f4 Kb7 42. Rc2 Kb6 43. Kf2 Nd8 44. Rc1
Nc6 45. Ke1 Na5 46. Rg8 Nb3 47. Rc2 Kb7 48. Kf2 Na5 49. Rg7 Kb6 50. Re2 Nb3
51. Ke1 b4 52. Rc2 bxc3 53. Rxc3 Rxd4 54. Rxg6+ Kc5 55. Rg7 Re4+ 56. Kf2
Kd4 57. Rf3 Nc5 58. Rxc7 Nd3+ 59. Kg1 Re1+ 60. Rf1 Re3 61. Kg2 Re2+ 62. Kg1
c3 63. g4 hxg4 64. h5 c2 65. h6 c1=Q 66. Rfxc1 Nxc1 67. Rxc1 g3 68. Rc6 Kd5
69. Rc7 Rh2 70. Re7 Rxh6 71. Re5+ Kd6 72. Kg2 Rh5 73. Kxg3 Kc6 74. Kf3 Kc7
75. Kf2 Kc6 76. Ke3 Kd6 77. Ke2 Rh1 78. Rxf5 Ke6 79. Rc5 Kf6 80. Rb5 Rc1
81. Kf3 Rf1+ 82. Ke3 Re1+ 83. Kf2 Rc1 84. Rg5 Rc4 85. Kf3 Rc1 86. Rh5 Rf1+
87. Ke3 Rc1 88. Ra5 Re1+ 89. Kf2 Rd1 90. Kf3 Rf1+ 91. Ke3 Re1+ 92. Kf2 Rc1
93. Kf3 Rf1+ 94. Kg4 Rg1+ 95. Kf3 Rf1+ {3-fold repetition} 1/2-1/2[/pgn]

Overnight I ran a 500 game Rapid match between Dragon 2.6 with Elo set to 2700 and Benjamin 1.0 (CCRL rapid rating 2699). Benjamin won the match by 63 elo points. So yes, even at 2700 level Dragon plays below the level of the corresponding CCRL Elo Rapid rating, but I don't think that this means that it would lose a Rapid match to a 2700 human; I'm pretty sure that Benjamin 1.0 would win a Rapid match from a 2700 Elo human. Also, based on my own experience, it is harder for me to score well against a suitably rated Dragon level than for me to score well against a weak CCRL engine that is even with that Dragon level. Probably this is because the extra depth of the classical engine is more helpful against crippled Dragon than against a human opponent who is relying mostly on general principles and pattern recognition. The weak CCRL engine often makes positional errors that I can exploit, while Dragon searching just a couple plies may make some tactical errors but I may not spot them in a fast game.

Can you post your best games based on your own experience against a suitably rated Dragon UCI_ELO level, and which games you did not do so good? With some explanations of the key play so we can learn from your games ?

NOTE: Probably back in July 0f 2001 you could have won versus Komodo Dragon 2.6 easy at UCI Level 2425, but now 2325 might be a tough challenge for you.

[Odd Gunnar Malin]

Most GUI use UCI_LimitStrength and UCI_Elo and that's what most engines are using . So please fix the missing underscore in future releases

ChessBase GUIs won't work with the underscores so that's why we removed them. Are there any GUIs that won't work without the underscores?

Chessbase GUI recognize UCI_Elo and UCI_LimitStrength, you see this in engines property when they don't show this so it is handled internally. It send these parameters to the engine when you play rated game, sadly it doesn't send more, e.g. ask for moves, so it's this internal engine who play the game. Guess they do it like this to get somewhat stable rating levels.
When you send UCI Elo and UCI LimitStreng you create two new options that Chessbase show under property like all options that isn't standard (and don't understand). With this you also discard all GUI that follow the standard and let the user set the strength of the engine in an easy way without going into the engine property to search for this option, they aren't used to do it here. How many Chessbase users will find this when they are used to just set the elo when starting a training game. Maybe better was to keep the standard and add a new single option for Chessbase users ex. Skill (0-3500) or EloLevel (0-3500) to not confuse old users.

The new Hiarcs gui take advantage of UCI_LimitStrength feature and Dragon will not work with this out of box.
I had to modify the flow with my little utility and a setting like this:

Code: Select all

[InBetween]
CommandLine := dragon-2.6-64bit-avx2.exe

[Client2Server]
UCI_Elo := UCI Elo
UCI_LimitStrength := UCI LimitStrength
OwnBook value false := OwnBook value true

[Server2Client]
UCI Elo type spin default 3500 min 0 := UCI_Elo type spin default 3500 min 1
UCI LimitStrength := UCI_LimitStrength

In addition to add the missing underscore Hiarcs seems to not allow an elo range starting from 0 so I changed it to start from 1.
I can't find anything in the protocol that shouldn't allow 0.

Also, I didn't find a way to use enginebook in Hiarcs so I change the ownbook=false to true and select nobook in Hiarcs to use engine books.
I will report these two to Hiarcs as bug.

[Odd Gunnar Malin]

Forget about the ownbook issue, I get it to work now with selecting Internal Book.
I have no idea what I did wrong when I tested it before. I'm pretty sure I went trough all the options.