Komodo Dragon 3 Released by KomodoChess

[Graham Banks]

Dragon by Komodo 3 64-bit 4CPU Gauntlet for CCRL 40/15

Live - https://tlcv.net/16063

[Krzysztof Grzelak]

Gentlemen of Komodo, when will there be free versions to download Komodo Dragon.

forum3/viewtopic.php?f=2&t=79738&start=10#p925091

"Larry K.: I don't anticipate making any Dragon versions free in the foreseeable future, unless something makes NNUE obsolete."

Thank you for the information Simon.

[lkaufman]

And another unexpected thing - in a 1500 game match Dragon 3 Elo 2600 (which should be accurate as verified by GM Lenderman) lost to Komodo 14.1 level 23 by 209 ELo! Even compressed by 25% it is still over 150 Elo difference. Is it possible that Komodo 14.1 level 23 is 2750 rapid?!

I'm reasonably convinced that Dragon Elo levels are tougher for humans to play against than whatever Komodo skill level scores 50% against them. So if we assume that the 2600 Elo is about right for Dragon 3 vs. humans in Rapid, then whatever Skill level scores 50% against it might only get 2500 or so vs. humans, so level 23 is probably well below 2750 in Rapid. But the equivalent level was stronger than any human, including Hikaru, at 5' + 2' blitz, so perhaps at Rapid it is not too far below 2750 on the human Fide scale. My estimate was 114 x Skill; 23 times 114 is 2642, so maybe that's about right or perhaps it's a bit higher than that. As for myself, so far I'm pretty even with Dragon 3 at 2200 in Rapid, but I haven't had enough free time yet to play many games.

[Fritz 0]

And another unexpected thing - in a 1500 game match Dragon 3 Elo 2600 (which should be accurate as verified by GM Lenderman) lost to Komodo 14.1 level 23 by 209 ELo! Even compressed by 25% it is still over 150 Elo difference. Is it possible that Komodo 14.1 level 23 is 2750 rapid?!

I'm reasonably convinced that Dragon Elo levels are tougher for humans to play against than whatever Komodo skill level scores 50% against them. So if we assume that the 2600 Elo is about right for Dragon 3 vs. humans in Rapid, then whatever Skill level scores 50% against it might only get 2500 or so vs. humans, so level 23 is probably well below 2750 in Rapid. But the equivalent level was stronger than any human, including Hikaru, at 5' + 2' blitz, so perhaps at Rapid it is not too far below 2750 on the human Fide scale. My estimate was 114 x Skill; 23 times 114 is 2642, so maybe that's about right or perhaps it's a bit higher than that. As for myself, so far I'm pretty even with Dragon 3 at 2200 in Rapid, but I haven't had enough free time yet to play many games.

I've done some personal testing on this, which more or less confirm your hypothesis. In a 10.000 games match Dragon 3 Elo 2000 lost to Komodo 13.3 level 19 by 103 Elo. I've still found the latter a bit tougher for me than the former at 15+10. Precisely, I beat the latter 2:0 about ten days ago, then played 1,5:1,5 last week, then lost 0:6 (!!!) yesterday (probably my worst chess day that I can recall), then played 2:2 today. But if we ignored the 0:6 result as atypical and unrealistic, and assumed that I stopped the session after 0:2 or 0:3 realising it wasn't a good idea to make a fool of myself, I would have an even score against Komodo 13.3 level 19, just like vs. Dragon 3 Elo 2000.

[AdminX]

Okay, well I've gotta say I'm impressed.

Edit: @Larry, I forgot to mention, was a tiny bit disappointed that the (much-spoken of, by users) variable delay (or what MByrne calls "sleep" in Honey) the lower Elo levels wasn't added yet. It's unnerving for the engine to respond instantaneously to every move, and kills the "real" experience a bit. Impressed the the overall weakening though, there's def improvement there.

Hi Norman,

OG Malin has a tool called MoveDelay which should work wonders for fast move by Dragon. Give it a try.

https://github.com/OGMalin/Binaries

[bmp1974]

Dragon 3 suffers time loss even in a 3+2 TC. Is there a bug or the default overhead needs to be changed?

[lkaufman]

Dragon 3 suffers time loss even in a 3+2 TC. Is there a bug or the default overhead needs to be changed?

This is the first such report. Are you playing games on a server over the internet, or just on your own computer? The default overhead should be plenty for games played on your own computer, but we do recommend raising it for server games on the internet, at least double it in that case to 100 ms and keep raising it if necessary, until forfeits stop.

[bmp1974]

Dragon 3 suffers time loss even in a 3+2 TC. Is there a bug or the default overhead needs to be changed?

This is the first such report. Are you playing games on a server over the internet, or just on your own computer? The default overhead should be plenty for games played on your own computer, but we do recommend raising it for server games on the internet, at least double it in that case to 100 ms and keep raising it if necessary, until forfeits stop.

The games were played in my computer (Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz 3.60 GHz, Ram=32GB) Threads=8, using Fritz 18.

[lkaufman]

Dragon 3 suffers time loss even in a 3+2 TC. Is there a bug or the default overhead needs to be changed?

This is the first such report. Are you playing games on a server over the internet, or just on your own computer? The default overhead should be plenty for games played on your own computer, but we do recommend raising it for server games on the internet, at least double it in that case to 100 ms and keep raising it if necessary, until forfeits stop.

The games were played in my computer (Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz 3.60 GHz, Ram=32GB) Threads=8, using Fritz 18.

Was it just one game lost on time out of a big number, or multiple losses? We did have one other report of a single game lost on time in a CCC event. We'd like to figure out what is the cause so we can fix it.

[bmp1974]

Dragon 3 suffers time loss even in a 3+2 TC. Is there a bug or the default overhead needs to be changed?

This is the first such report. Are you playing games on a server over the internet, or just on your own computer? The default overhead should be plenty for games played on your own computer, but we do recommend raising it for server games on the internet, at least double it in that case to 100 ms and keep raising it if necessary, until forfeits stop.

The games were played in my computer (Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz 3.60 GHz, Ram=32GB) Threads=8, using Fritz 18.

Was it just one game lost on time out of a big number, or multiple losses? We did have one other report of a single game lost on time in a CCC event. We'd like to figure out what is the cause so we can fix it.

[pgn][Event "Tour4_blitz only, Blitz 3.0min+2.0sec"]
[Site "?"]
[Date "2022.05.11"]
[Round "2.1"]
[White "Dragon 3 by Komodo Chess, 64-bit"]
[Black "Lc0 v0.30.0-dev+git.77b5659"]
[Result "0-1"]
[ECO "C58"]
[Annotator "1.08;0.27"]
[PlyCount "210"]
[EventDate "2022.05.11"]
[EventType "tourn"]
[SourceTitle "Fritz Engine Tournament"]
[Source "BM"]

{Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz 3600 MHz W=33.1 plies; 6,090kN/s; 68,
152 TBAs B=7.5 plies; 12kN/s} 1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 d5 5. exd5
Na5 6. d3 Nxc4 7. dxc4 h6 8. Nf3 e4 9. Nd4 {1.08/24 7} Bg4 {0.27/15 13} 10. Ne2
{0.67/26 11} Bb4+ {0.24/16 7} 11. Bd2 {0.69/26 6} Qe7 {0.23/17 0} 12. h3 {
0.59/29 24} Bh5 {0.20/17 9} 13. Nbc3 {0.60/29 9 (a3)} O-O {0.17/14 7} 14. O-O {
0.56/29 15} c6 {0.16/15 7} 15. Qe1 {0.48/31 6} Rfe8 {0.15/13 0} 16. Ng3 {
0.49/31 6 (a3)} Bg6 {0.12/15 10} 17. Ncxe4 {0.51/33 6 (Nd1)} Bc5 {0.02/15 10
(Nxe4)} 18. Nxc5 {0.46/30 5} Qxc5 {0.02/14 0} 19. Qc1 {0.38/32 6} Qxc4 {
0.02/13 8} 20. dxc6 {0.33/34 8} Qxc6 {0.02/12 0} 21. c3 {0.31/34 6} Nh5 {
0.03/10 11} 22. Nxh5 {0.37/34 5} Bxh5 {0.03/10 3} 23. Be3 {0.32/35 9 (Re1)} Re6
{0.02/9 14 (Rad8)} 24. f3 {0.31/29 12 (Qd2)} Rae8 {0.00/9 9} 25. Bd4 {0.28/30 4
} Qc7 {0.00/9 11 (Re2)} 26. Rf2 {0.40/26 6} Qg3 {0.00/9 5 (Bg6)} 27. Qd2 {
0.30/29 5 (Qf1)} Bg6 {-0.03/8 8} 28. Kh1 {0.28/32 4} Bf5 {-0.01/8 9} 29. Rg1 {
0.16/30 7} a6 {-0.01/7 4 (a5)} 30. Qd1 {0.26/30 3 (a4)} b5 {0.00/7 8 (Qh4)} 31.
a3 {0.17/29 7 (Qf1)} R6e7 {0.00/7 7 (Kh7)} 32. Qf1 {0.16/31 8 (b3)} Be6 {
0.00/7 5} 33. Re2 {0.16/32 3} Bb3 {0.00/9 5} 34. Rxe7 {0.12/36 9 (Qf2)} Rxe7 {
0.00/8 3} 35. Qd3 {0.12/40 3} Re6 {0.00/7 11 (Qf4)} 36. Be3 {0.24/34 4 (Qd2)}
Qe5 {0.01/6 6 (Qg6)} 37. Re1 {0.19/31 2} Qg3 {0.01/6 5 (Qd6)} 38. Qd8+ {
0.17/32 4 (Qd2)} Kh7 {0.01/7 3} 39. Qd2 {0.16/33 8} Qg6 {0.02/6 9 (Qd6)} 40.
Kg1 {0.16/34 6 (Bd4)} a5 {0.03/5 6 (Qg3)} 41. c4 {0.15/38 5} Bxc4 {0.04/8 3}
42. Qxa5 {0.14/37 8} Bd5 {0.05/8 7 (Qd3)} 43. Kf2 {0.23/32 3} Rc6 {0.04/9 3}
44. Rd1 {0.22/33 4} Rc2+ {0.03/10 4} 45. Rd2 {0.18/35 5} Bc4 {0.02/9 5 (Qd3)}
46. b3 {0.30/31 2 (Rxc2)} Bxb3 {0.01/8 6} 47. Qxb5 {0.30/35 7} Bc4 {0.01/8 5}
48. Qb4 {0.29/35 2 (Qe5)} Rc1 {0.00/8 5} 49. Re2 {0.29/34 11} Qd3 {0.01/8 6
(Rc2)} 50. Bxc1 {0.29/34 3} Qxe2+ {0.02/8 1} 51. Kg3 {0.29/32 2} Qc2 {0.01/6 8
(Qe6)} 52. Bd2 {0.30/36 3} Bf1 {0.02/6 4 (Qb3)} 53. Kf2 {0.30/29 3} Bc4 {
0.04/6 4 (Qc4)} 54. a4 {0.36/22 4} f6 {0.05/6 4 (Qb3)} 55. a5 {0.43/33 3} Qd3 {
0.05/6 3 (Qd1)} 56. Kg1 {0.43/34 4 (Bc3)} Kg6 {0.03/6 3 (Qc2)} 57. Kh2 {
0.43/33 1} Qd5 {0.04/5 3 (Qc2)} 58. Bf4 {0.43/31 2 (Bc3)} Qb5 {0.03/5 3 (Ba6)}
59. Qd2 {0.43/35 2 (Qe1)} Bf1 {0.03/5 3} 60. Qf2 {0.43/35 5 (Kg1)} Bd3 {
0.04/5 3 (Qe2)} 61. Qe1 {0.43/28 1 (Qb6)} Kh7 {0.03/5 3 (Qf5)} 62. Bd6 {
0.43/32 1 (Qe3)} Kg8 {0.03/5 3 (Kg6)} 63. Bb4 {0.43/35 2 (Qe3)} Kh7 {0.04/5 3}
64. Kg1 {0.43/38 1 (Qe7)} Qc4 {0.05/5 2 (Qd5)} 65. Qe3 {0.43/38 2} Kg8 {
0.05/6 2 (Qc2)} 66. Bd2 {0.43/32 1 (Be7)} Kh7 {0.05/5 2} 67. Be1 {0.43/36 2
(Bb4)} Qd5 {0.05/5 2} 68. Qf4 {0.43/35 6 (Bc3)} Qc5+ {0.04/5 2 (Qc4)} 69. Kh1 {
0.43/39 2 (Kh2)} Qc4 {0.03/5 2 (Qc2)} 70. Qd6 {0.43/41 2 (Qe3)} Bf1 {0.04/5 2}
71. Kg1 {0.43/39 3} Bd3 {0.05/6 2} 72. Bd2 {0.42/35 2 (h4)} Kg6 {0.04/6 3 (Qb5)
} 73. Qe7 {0.43/33 2 (Bb4)} Qd4+ {0.03/5 2 (Qa2)} 74. Kh2 {0.43/35 1} Bf1 {
0.04/6 2 (Qd5)} 75. Bb4 {0.43/33 1 (Qe1)} Kh7 {0.05/6 2} 76. Qe1 {0.43/38 2
(Qb7)} Ba6 {0.05/6 2 (Bd3)} 77. Ba3 {0.43/35 2 (Bf8)} Qd5 {0.04/5 2 (Bd3)} 78.
Bf8 {0.43/35 3} Kg8 {0.03/6 2} 79. Bb4 {0.43/36 2 (Qe7)} Qc6 {0.03/5 2 (Qc4)}
80. Qe3 {0.43/34 2 (Qe7)} Qc7+ {0.04/5 2} 81. Kh1 {0.43/37 2} Qc4 {0.04/5 2}
82. Bd2 {0.43/37 3 (Bc3)} Kf7 {0.03/5 2 (Qf1+)} 83. Kh2 {0.43/32 2 (Qb6)} Qc7+
{0.04/5 2 (Qb5)} 84. Kh1 {0.43/37 1 (Kg1)} Qc4 {0.03/5 2} 85. Qb6 {0.43/34 3
(Qa7+)} Kg6 {0.03/5 2} 86. Qd6 {0.43/34 2 (Bb4)} Qd3 {0.02/5 2 (Qf1+)} 87. Qb4
{0.36/32 1 (Qf4)} Qc4 {0.03/5 2 (Kh7)} 88. Qb8 {0.36/32 3 (Qe7)} Qb5 {0.02/5 2
(Qf1+)} 89. Qg3+ {0.35/29 2 (Qf4)} Kh7 {0.04/6 2} 90. Kh2 {0.34/29 2 (Qf4)} Qd3
{0.04/5 2 (Qc4)} 91. Bb4 {0.33/28 1} Qd8 {0.03/5 2 (Qc4)} 92. Bc3 {0.33/32 2
(Qf4)} Qd3 {0.04/5 2 (Qc8)} 93. Be1 {0.33/35 1 (Qe1)} Qc4 {0.03/5 2} 94. h4 {
0.33/37 2 (Qd6)} Qe6 {0.03/5 2 (Bb5)} 95. Bb4 {0.33/32 3} Qc8 {0.04/5 2 (Qc4)}
96. Qe1 {0.33/33 4 (Qd6)} Kh8 {0.04/5 2 (Qc4)} 97. Qe7 {0.33/32 2} Kh7 {
0.04/5 2 (Qc4)} 98. Qd6 {0.33/31 2 (Kg3)} Bf1 {0.05/5 2 (Qc4)} 99. Kg1 {
0.33/28 1 (Qc5)} Ba6 {0.05/6 2} 100. Qf4 {0.33/32 2 (Bc3)} Kg8 {0.05/5 2 (Qc2)}
101. Be1 {0.33/31 2 (Qd6)} Qc5+ {0.04/5 2 (Qc4)} 102. Kh1 {0.33/37 1 (Kh2)} Qc4
{0.03/5 2 (Qb5)} 103. Qb8+ {0.33/33 1 (Qxc4+)} Kh7 {0.03/6 2} 104. Qb1+ {
0.33/32 3 (Kh2)} Kg8 {0.02/6 2 (Qd3)} 105. Qa1 {0.33/36 2 (Kh2)} Kf7 {0.02/5 2
(Kh7) Time} 0-1

[/pgn]

[pgn][Event "Tour4_blitz only, Blitz 3.0min+2.0sec"]
[Site "?"]
[Date "2022.05.12"]
[Round "6.1"]
[White "Dragon 3 by Komodo Chess, 64-bit"]
[Black "Lc0 v0.30.0-dev+git.77b5659"]
[Result "0-1"]
[ECO "C14"]
[Annotator "1.28;0.44"]
[PlyCount "124"]
[EventDate "2022.05.11"]
[EventType "tourn"]
[SourceTitle "Fritz Engine Tournament"]
[Source "BM"]

{Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz 3600 MHz W=26.4 plies; 5,470kN/s; 81,
508 TBAs B=19.9 plies; 29kN/s; 720 TBAs} 1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. Bg5
Be7 5. e5 Nfd7 6. h4 f6 7. Bd3 c5 8. Qh5+ Kf8 9. exf6 {1.28/26 5} Nxf6 {
0.44/19 12} 10. Bxf6 {1.30/29 5} Bxf6 {0.44/18 0} 11. dxc5 {1.14/28 9} Nd7 {
0.42/19 6} 12. Nf3 {1.06/29 6} Nxc5 {0.42/20 0} 13. O-O {1.13/30 11} Bd7 {
0.42/20 0} 14. Rae1 {1.10/30 21 (Rfe1)} Be8 {0.43/19 16} 15. Qg4 {1.03/28 4} h5
{0.43/22 5} 16. Qf4 {1.08/29 4} Kg8 {0.41/23 14 (Rh6)} 17. Ng5 {1.39/24 4} Rh6
{0.41/23 0} 18. Nxe6 {1.28/29 7} Nxe6 {0.42/22 0} 19. Rxe6 {1.28/30 7} Bf7 {
0.41/20 0} 20. Re2 {1.14/29 26 (Qf5)} Bxh4 {0.46/18 16} 21. Rfe1 {0.96/27 4}
Bg5 {0.52/18 13 (Qf6)} 22. Qf5 {1.28/28 3} Qf6 {0.53/23 0} 23. Re5 {1.21/31 8}
Bd2 {0.53/20 0} 24. Nxd5 {1.30/31 6} Qxf5 {0.56/20 0} 25. Ne7+ {1.38/31 4} Kf8
{0.57/21 11} 26. Nxf5 {1.33/34 12} Rf6 {0.52/21 26} 27. R1e2 {1.03/30 13} Bf4 {
0.51/24 0} 28. Re7 {0.99/26 4} g6 {0.50/26 0 (b6)} 29. Nd4 {1.35/30 4} Rd8 {
0.53/24 7} 30. Nf3 {1.25/32 4} Bd5 {0.56/23 4} 31. Rh7 {1.26/33 5} Kg8 {
0.57/22 1} 32. Ree7 {1.19/34 4} Bf7 {0.56/24 4} 33. Rhxf7 {1.29/35 8} Rxf7 {
0.57/23 0} 34. Bc4 {1.29/34 5} Rd1+ {0.56/22 0} 35. Ne1 {1.25/33 7} Bd6 {
0.54/22 12} 36. Bxf7+ {1.34/34 6} Kf8 {0.56/22 19 (Kg7)} 37. Re4 {1.27/34 22}
Kxf7 {0.57/20 0} 38. Kf1 {1.24/31 9} Rb1 {0.57/21 0} 39. b3 {1.22/30 7} a5 {
0.56/20 0 (Rc1)} 40. Ke2 {1.34/29 2} Rc1 {0.53/20 10} 41. Kd2 {1.34/28 3} Ra1 {
0.54/19 4} 42. a4 {1.39/28 2} Bb4+ {0.54/19 0} 43. c3 {1.21/29 11} Bd6 {
0.52/16 11 (Bc5)} 44. Kd3 {1.33/26 5} Ra2 {0.50/16 9} 45. Rd4 {1.45/26 2 (Re2)}
Ke7 {0.51/15 13 (Be7)} 46. Nc2 {1.12/23 3 (Re4+)} Rb2 {0.50/16 7} 47. b4 {
1.29/22 2} axb4 {0.51/16 5} 48. cxb4 {1.13/24 1} Ke6 {0.49/14 4} 49. Re4+ {
1.30/22 2} Kd7 {0.50/13 6} 50. f3 {1.33/21 3} g5 {0.52/14 7} 51. Kc3 {1.25/20 2
} Rb1 {0.53/16 3} 52. a5 {1.29/19 2 (Ne3)} h4 {0.57/13 10} 53. Ne3 {1.09/16 2}
Bf8 {0.51/16 6 (Rh1)} 54. f4 {1.05/17 2} gxf4 {0.52/21 0} 55. Rxf4 {1.06/14 1}
Bg7+ {0.48/17 12 (Bh6)} 56. Kc4 {1.03/17 2} Bh6 {0.49/17 0} 57. Rf7+ {1.51/14 1
} Ke6 {0.48/16 1} 58. Rf3 {1.24/18 4} Bg5 {0.48/14 2} 59. Nc2 {1.47/18 2} Be7 {
0.49/12 0} 60. Kb5 {1.32/17 2} Rb2 {0.36/14 5} 61. Rc3 {1.23/18 2} Kd7 {
0.34/14 2} 62. Rc4 {1.22/19 1} Bd8 {0.32/13 3 Time} 0-1

[/pgn]

2 games loss on time in a 6 game span