In ten years Stockfish or Komodo will beat Nakamura with Knight Odds

[lkaufman]

With Computer 4 times fasters and much better Engines than today

He would have a pretty difficult time even today, especially if it were a blitz match. Computers don't feel time pressure.

I'm nearly even with Dragon at 10' + 5" at knight odds now (MCTS or standard mode doesn't seem to matter much). I'm pretty sure that Nakamura plays better with 2' + 1" than I do with 10' + 5", so I'm pretty sure that he would beat Dragon at 2' + 1" (or 3' with no inc). Maybe 1' + 1" might be close to fair. I'm sure we can find ways to improve its performance under such conditions, but it's hard to imagine it ever coming out ahead in Rapid (as opposed to Blitz). Maybe if we make it Armageddon knight odds, where Dragon only needs to draw a piece down to win, then that becomes a realistic goal.

Larry, there are possibility of seeing similar match in the coming days?

Yes this would be a great idea to see if Komodo Dragon can at least draw = Win Vs GM Naka at 1'+1" Knight Odds

That would indeed be an interesting match, probably a close one. Another idea is to give the GM a big handicap initially, reducing it every time he wins a game. I think it makes sense to have a match whenever we release a new version of Dragon, so roughly quarterly.

Having many games with the same handicap would help to understand their value better. The problem is that I would give komodo as a slight favorite in knight odd in 1 '+ 1 "against GM naka, but if naka started playing many games with this handicap he could find a way to play better with the advantage, How did you do it, Larry, right? Once you have played many games to that handicap you got better results.

I do tend to do better at knight odds with more experience, but not way better. It's not like I duplicate previous games for many moves and improve on them; Komodo with MP has decent variety automatically, and also I try to vary my own opening play, I would consider it cheating to purposely duplicate a previous game for many moves (if that's even possible) and vary at the point where I made a mistake. For a serious match Dragon would use a small knight odds opening book (based on its own analysis) which would provide enough variety to minimize the learning effect. Also alternating between b1 and g1 knight odds reduces learning.

[Chessqueen]

Here's a game between Komodo 1 vs Stockfish 12 given Komodo 1 knight odds @ 10'+5". I am not sure what the rating difference is between the two. I also tried so other games between Komodo Dragon and much older engines but Komodo never got a win, the best it did was a draw against Mephisto Dallas uci version.

[pgn]
[Event "banksia game"]
[Date "2021.02.01"]
[White "Komodo64 1.0 JA"]
[Black "Stockfish 12"]
[Result "1-0"]
[TimeControl "600+5"]
[Time "07:55:59"]
[Termination "mate"]
[FEN "rnbqkb1r/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 8"]
[SetUp "1"]

1. g1f3 {+3.37/18 20.0 65175292} d7d5 {-4.75/33 106.2 494792324} 2. e2e3 g7g6 {-4.76/33 18.6 90323440}
3. c2c4 {+3.60/19 19.3 44209569} e7e6 {-5.01/31 0.1 92624440} 4. b1c3 {+3.64/18 23.9 54647337} f8g7 {-4.81/30 12.0 59294318}
5. d2d4 {+3.67/19 39.1 113828155} e8g8 {-4.90/32 20.8 283573410} 6. f1d3 {+3.68/19 9.6 67200105} b8c6 {-5.15/29 35.8 179652006}
7. e1g1 {+3.75/19 17.3 118716422} d5c4 {-5.04/34 39.3 280771400} 8. d3c4 {+3.71/17 16.6 35884246} a7a6 {-5.09/31 3.6 99587267}
9. c1d2 {+3.70/18 31.7 80433846} b7b6 {-5.38/32 52.1 266747059} 10. d1c2 {+3.66/18 22.2 169702171} c8b7 {-5.21/30 24.7 130523372}
11. c3e4 {+3.74/17 25.7 121854202} c6e7 {-5.32/30 46.2 233527900} 12. a1c1 {+3.76/15 15.1 36407987} e7f5 {-5.40/31 31.9 241764794}
13. f1d1 {+3.68/15 12.7 28845871} d8e7 {-5.23/31 27.9 140598074} 14. c4d3 {+3.67/16 15.9 36856351} a8c8 {-5.50/29 28.4 145751153}
15. b2b4 {+3.78/17 0.4 67588463} f8d8 {-5.48/26 11.2 60457192} 16. a2a4 {+3.78/16 18.2 42047729} h7h6 {-5.71/42 84.4 535632153}
17. d2c3 {+3.95/17 25.4 59027995} f5d6 {-5.55/31 2.0 142319886} 18. e4d2 {+3.89/16 11.8 25645568} d6f5 {-5.78/28 0.9 67967972}
19. c2b3 {+3.89/15 28.0 63774829} g8h7 {-5.48/28 38.8 206227969} 20. h2h3 {+4.02/15 14.9 33172971} h7g8 {-5.98/33 85.4 447115945}
21. c3b2 {+4.00/15 10.8 23439046} b7d5 {-5.61/28 11.5 59353379} 22. b3a3 {+4.05/18 42.1 96016685} d5b7 {-6.32/31 0.0 233914770}
23. a4a5 {+4.12/17 47.4 108281789} c8b8 {-5.93/25 12.7 66766556} 24. b4b5 {+4.04/15 8.7 18512951} a6b5 {-5.17/27 3.6 19446819}
25. a3e7 {+3.95/19 9.1 34483584} f5e7 {-5.29/30 0.0 53598442} 26. c1c7 {+4.17/18 9.8 25834324} e7d5 {-5.36/28 9.1 54796290}
27. a5b6 {+3.78/17 4.8 33768814} d5b6 {-5.44/29 6.4 55148005} 28. d1a1 {+4.14/18 11.8 44496905} b6a4 {-5.62/31 0.0 68399457}
29. b2a3 {+4.02/18 10.2 29344322} b7f3 {-5.50/28 5.0 32648483} 30. d2f3 {+4.20/20 16.1 58807704} b5b4 {-5.51/25 1.8 10880192}
31. a3c1 {+4.20/20 11.5 38334193} a4b6 {-5.33/31 0.3 70157660} 32. a1a7 {+4.15/18 10.4 29363839} b6c8 {-6.09/32 9.9 126107767}
33. a7b7 {+4.19/19 9.4 27212391} c8d6 {-6.09/29 0.0 59889593} 34. b7b8 {+4.14/19 8.0 21870880} d8b8 {-6.20/28 1.2 58226620}
35. g2g3 {+4.29/18 8.5 25950193} d6b5 {-5.55/28 7.9 46118738} 36. c7c6 {+4.24/18 2.2 25937947} b5a7 {-5.75/29 12.1 72386052}
37. c6c4 {+4.34/19 3.6 41347777} g6g5 {-5.84/30 14.8 89232959} 38. g1g2 {+4.49/17 14.4 35639392} a7c8 {-6.17/30 10.1 63969460}
39. f3d2 {+4.42/17 12.0 31281463} g7f8 {-5.74/26 3.5 22031387} 40. c4c7 {+4.46/18 3.6 19750354} c8b6 {-5.85/25 0.2 23247686}
41. e3e4 {+4.58/19 17.0 46336589} b8e8 {-6.01/28 13.9 88641936} 42. d2b3 {+4.71/19 17.2 48691322} e8b8 {-6.45/31 0.0 113487626}
43. d3e2 {+4.78/19 8.8 25568230} f8d6 {-7.03/28 12.2 79516320} 44. c7a7 {+4.80/20 15.4 75207201} b6c8 {-6.39/24 1.8 12008501}
45. a7a5 {+4.80/18 4.5 17262772} d6f8 {-6.31/21 1.0 5844508} 46. e2d3 {+4.77/18 7.6 21229264} b8b7 {-6.28/26 0.0 48464129}
47. a5a6 {+4.73/16 7.0 18894733} b7b8 {-6.43/27 7.2 46988773} 48. c1b2 {+4.78/18 2.6 27310624} c8b6 {-6.36/27 2.4 15295801}
49. g2f3 {+4.73/15 6.0 15826726} b6d7 {-6.43/29 5.6 37123867} 50. b2c1 {+4.74/16 1.8 19683249} b8b7 {-6.59/26 1.0 18192142}
51. g3g4 {+4.73/17 5.0 15471309} d7b8 {-6.63/26 9.6 62728498} 52. a6a1 {+4.85/19 21.7 81119028} b8c6 {-6.39/23 1.4 9126506}
53. a1a8 {+4.84/20 8.6 21537556} c6e7 {-6.61/32 5.5 89328223} 54. c1d2 {+4.94/18 6.4 16496047} g8g7 {-6.63/31 1.7 58046211}
55. b3c5 {+5.88/20 8.3 23406076} b7c7 {-6.71/34 0.4 67904236} 56. d2b4 {+5.88/20 5.1 15612701} e7g6 {-6.78/32 0.0 40868561}
57. d3b5 {+5.91/20 9.6 28283816} g6h4 {-6.99/33 24.5 203269434} 58. f3g3 {+5.86/21 5.8 17312999} f8e7 {-7.01/27 3.0 25215145}
59. b5a4 {+5.87/19 14.9 40519690} h4g6 {-6.60/26 2.9 22509129} 60. b4a5 {+5.87/20 4.2 20142045} e7c5 {-6.98/31 5.0 76446706}
61. a5c7 {+5.87/16 0.5 1442900} c5d4 {-7.05/27 2.7 27065123} 62. a8d8 {+5.87/20 3.6 18031581} d4e5 {-7.06/28 4.1 34893853}
63. c7e5 {+5.86/24 3.1 21764264} g6e5 {-7.30/33 1.4 41582677} 64. a4b5 {+5.87/24 5.3 19343775} g7f6 {-7.44/30 4.1 38868426}
65. d8c8 {+5.89/23 6.3 18574193} f6g7 {-7.93/35 8.5 81850246} 66. c8a8 {+5.86/23 11.5 35313434} g7g6 {-8.13/32 0.0 109878493}
67. a8b8 {+5.86/25 5.3 16251412} g6f6 {-8.05/34 4.2 39888618} 68. f2f3 {+5.85/22 8.5 25086530} f6g7 {-7.36/32 7.6 71865406}
69. b8c8 {+5.86/25 7.5 22494372} e5g6 {-7.88/33 8.7 82599812} 70. c8c1 {+5.88/26 4.1 37156989} g7f8 {-7.96/36 5.3 50552324}
71. c1c7 {+5.92/22 5.6 16190428} g6e7 {-8.04/38 3.0 29202934} 72. h3h4 {+6.55/21 5.1 15142604} e7g6 {-11.81/35 20.3 206534700}
73. h4h5 g6f4 {-13.33/34 0.0 95245028} 74. c7c8 {+7.94/21 5.2 14067098} f8g7 {-14.25/31 2.6 29019869}
75. c8b8 {+8.06/23 2.6 15140029} g7h7 {-14.69/33 5.9 64248212} 76. b8b7 {+8.73/22 5.4 14827041} h7g8 {-15.09/32 1.1 73264058}
77. b5e8 {+9.28/24 7.6 24789697} f7f6 {-16.48/32 1.7 108316779} 78. e8g6 {+9.75/23 4.5 18087290} f4e2 {-16.44/29 6.5 67032129}
79. g3f2 {+9.61/22 17.5 48030171} e2f4 {-20.49/33 0.0 196010127} 80. b7h7 {+9.20/20 19.1 52471896} f4d3 {-21.06/28 7.9 82473568}
81. f2e3 {+9.97/21 4.6 12611689} d3f4 {-69.98/40 10.8 174725744} 82. h7h6 {+10.96/22 4.5 12610540} f4g2 {-66.89/33 10.6 114944980}
83. e3f2 {+12.99/25 1.7 34884503} g2f4 {-72.55/31 0.8 28765176} 84. h6h7 f4d3 {M-14/77 0.0 321331521}
85. f2e3 {+10.63/22 4.1 11931754} d3e5 {M-13/72 0.0 48613242} 86. h5h6 {+12.74/22 23.2 6659375} e5c4 {M-13/37 2.7 28247400}
87. e3f2 c4e5 {M-11/91 0.0 219216489} 88. h7e7 {M+11/14 0.3 797091} e5g6 {M-10/99 2.8 38222560}
89. e7g7 g8h8 {M-9/105 3.1 43552478} 90. g7g6 h8h7 {M-8/152 3.2 35289737}
91. g6f6 e6e5 {M-7/67 3.4 56028412} 92. f3f4 {M+7/12 0.1 111891} e5f4 {M-6/245 2.3 36437089}
93. e4e5 f4f3 {M-5/245 0.3 4952238} 94. e5e6 h7h8 {M-4/245 0.0 72259}
95. e6e7 h8g8 {M-3/245 0.0 262206} 96. e7e8q {M+3/1 0.0 161} g8h7 {M-2/245 0.0 39672}
97. f6f7 {M+2/1 0.0 363} h7g6 {M-1/245 0.0 2951} 98. e8e6 {M+1/1 0.0 705} 1-0
[/pgn]

Can you replay the game with the correct Knight Odds, Vs Mephitos Dallas Uci, and set the Contempt to 150 or 180 for Komodo

[d]rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1

[JohnW]

Ok, I believe I did it correctly this time. Here are two games between Komodo Dragon and Mephisto Dallas @10'+5", one with the king side knight removed and the other with the queen side knight, I wasn't sure if it mattered. Komodo won both of them anyway. For some reason the pgn copied from Banksia gui doesn't display correctly in the "pgn" bracket, it doesn't show the first move as move #1, so I am displaying it raw here.

[Event "banksia game"]
[Date "2021.02.02"]
[White "Dragon by Komodo Chess 64-bit"]
[Black "Mephisto Dallas 32 Bit"]
[Result "1-0"]
[TimeControl "600+5"]
[Time "11:02:06"]
[Termination "mate"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 32"]
[SetUp "1"]

1. d2d4 {-2.78/26 19726 76431730} d7d5 {-4.32/1 21727} 2. e2e3 g8f6 {-4.28/1 19093}
3. a2a3 {-2.87/26 20807 152928880} d8d6 {-4.32/1 18723} 4. c2c4 {-2.60/23 11283 40840717} d5c4 {-4.45/1 18136}
5. f1c4 {-2.90/29 36969 150314904} d6c6 {-4.36/1 17943} 6. c4f1 {-2.96/28 18860 143904121} c8f5 {-4.20/1 17433}
7. b1c3 {-2.26/23 9165 34363239} b8d7 {-3.92/1 17057} 8. f2f3 {-2.17/25 12113 47148704} a7a6 {-3.80/1 16703}
9. e3e4 {-1.88/23 9694 36641103} f5g6 {-4.00/1 16389} 10. c1e3 e8c8 {-4.00/1 16198}
11. f1e2 {-0.64/25 18659 77749192} e7e6 {-3.76/1 15872} 12. e1g1 {-0.24/28 51237 223183537} d7b6 {-3.50/1 15547}
13. d1b3 {+0.00/29 14274 58748849} h7h5 {-3.16/1 15141} 14. f1c1 {+0.38/27 32729 139843306} c6d7 {-2.96/1 14859}
15. a3a4 h5h4 {-3.28/1 14619} 16. a4a5 {+2.68/26 20988 94027624} b6a8 {-3.12/1 14338}
17. c3b5 {+5.37/26 13436 60819692} a6b5 {-5.04/1 14184} 18. a5a6 {+5.71/28 23225 155143819} b7a6 {-3.96/1 13939}
19. a1a6 {+9.96/25 9216 43051798} c8b7 {+1.84/1 13644} 20. e2b5 {M+12/23 17132 82139310} a8b6 {+2.32/1 13253}
21. b5d7 f6d7 {+2.45/1 13093} 22. a6a7 {M+9/20 2105 10218849} b7a7 {+1.00/1 12966}
23. c1c7 a7b8 {+0.68/1 12692} 24. e3f4 e6e5 {+9.99/1 12311}
25. f4e5 {M+6/31 2306 7867722} f8d6 {+9.99/1 12057} 26. e5d6 h8h5 {-0.01/1 11891}
27. c7d7 b8c8 {-0.01/1 189} 28. b3b6 c8d7 {-0.01/1 153}
29. b6c7 d7e8 {-0.01/1 213} 30. c7e7 1-0



[Event "banksia game"]
[Date "2021.02.02"]
[White "Dragon by Komodo Chess 64-bit"]
[Black "Mephisto Dallas 32 Bit"]
[Result "1-0"]
[TimeControl "600+5"]
[Time "11:21:22"]
[Termination "mate"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 64"]
[SetUp "1"]

1. g1f3 {-2.73/23 7784 28424587} d7d5 {-4.36/1 21761} 2. e2e3 b8c6 {-4.28/1 19076}
3. d2d4 {-2.85/25 13603 51116265} g8f6 {-4.45/1 18650} 4. f1e2 d8d6 {-4.40/1 18317}
5. e1g1 {-2.99/23 17901 67300979} c8f5 {-4.64/1 17920} 6. f3h4 {-2.84/26 37800 145555890} f5e4 {-4.60/1 17421}
7. f2f3 {-2.70/24 13192 46934648} e4g6 {-4.72/1 17207} 8. c2c3 e7e6 {-4.72/1 16866}
9. a2a4 {-2.86/24 23411 84075534} f8e7 {-4.64/1 16533} 10. g2g3 {-2.87/25 16344 58411615} e8c8 {-4.88/1 16194}
11. b2b4 {-2.01/22 12072 43947653} c8b8 {-4.20/1 16074} 12. a4a5 {-1.93/23 32915 126419948} a7a6 {-3.80/1 15391}
13. b4b5 {-1.19/26 20035 93410961} a6b5 {-3.32/1 15066} 14. e2b5 {+0.00/32 63710 293304493} f6g8 {-2.84/1 14783}
15. h4g6 {+4.62/27 44850 216402400} h7g6 {-2.60/1 14571} 16. c1a3 {+4.30/27 496 74654786} d6d7 {-2.72/1 14396}
17. a5a6 {+4.72/32 28783 212230317} d7c8 {-1.36/1 14039} 18. b5c6 {M+11/23 38107 211733086} b7c6 {-1.12/1 13671}
19. d1a4 {M+9/21 2033 12231802} b8a7 {-0.92/1 13462} 20. f1b1 {M+7/25 2303 14591317} e7a3 {+2.36/1 13321}
21. a1a3 d8d6 {-0.01/1 13108} 22. a4b3 {M+4/99 413 2268684} g6g5
23. b3b7 {M+3/99 29 114713} c8b7 {-0.01/1 160} 24. a6b7 a7b8 {-0.01/1 223}
25. a3a8 1-0

[Chessqueen]

[quote=JohnW post_id=881251 time=1612283119 user_id=6645]
Ok, I believe I did it correctly this time. Here are two games between Komodo Dragon and Mephisto Dallas @10'+5", one with the king side knight removed and the other with the queen side knight, I wasn't sure if it mattered. Komodo won both of them anyway. For some reason the pgn copied from Banksia gui doesn't display correctly in the "pgn" bracket, it doesn't show the first move as move #1, so I am displaying it raw here.

Great games, What is the estimate ratings of Mephito Dalla Uci

[JohnW]

Looking at an old swedish ratings list, it had the Mephisto Dallas 68000 with a rating of 1970

[lkaufman]

Looking at an old swedish ratings list, it had the Mephisto Dallas 68000 with a rating of 1970

Those ratings were tied to national rating lists of Sweden and perhaps Germany and Holland, which had much lower ratings than FIDE and more so than US. So for example the Mephisto Dallas 68000 is rated only a little lower than Fidelity Mach III on the SSDF list, but the Fidelity Mach III got a USCF rating (based on 48 tournament games under strict conditions with prize money for wins and draws) of 2265, which should mean about 2165 FIDE. So perhaps the Mephisto Dallas 68000 was something like 2240 USCF or 2140 FIDE level at the time.

[JohnW]

Looking at an old swedish ratings list, it had the Mephisto Dallas 68000 with a rating of 1970

Those ratings were tied to national rating lists of Sweden and perhaps Germany and Holland, which had much lower ratings than FIDE and more so than US. So for example the Mephisto Dallas 68000 is rated only a little lower than Fidelity Mach III on the SSDF list, but the Fidelity Mach III got a USCF rating (based on 48 tournament games under strict conditions with prize money for wins and draws) of 2265, which should mean about 2165 FIDE. So perhaps the Mephisto Dallas 68000 was something like 2240 USCF or 2140 FIDE level at the time.

Makes me wonder what the rating would be now running on my PC (i7-6700K\16 GB Ram) which obviously is a lot faster than a 68000 processor.

[rcmaddox]

With Computer 4 times fasters and much better Engines than today

I remember a discussion about this five years ago ---->> http://talkchess.com/forum3/viewtopic.p ... 49#p655549