Komodo v SF Odds Game

[yanquis1972]

didn't realize it till i stumbled on a game the other night but the komodo team is still playing (casual) odds matches against strong masters; latest victim being IM danny rensch (games can be found here; there are also streams available on the chess.com twitch channel)

the latest one, which is relevant here:

[pgn][Event "Live Chess - Odds Chess"]
[Site "Chess.com"]
[Date "2017.08.22"]
[White "DanielRensch"]
[Black "KomodoChess"]
[Result "0-1"]
[ECO "A40"]
[WhiteElo "2448"]
[BlackElo "3368"]
[TimeControl "2700+30"]
[EndTime "15:22:41 PDT"]
[Termination "KomodoChess won by resignation"]
[SetUp "1"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -"]
[Variant "Odds Chess"]
[CurrentPosition "1r4k1/p2p1ppp/2b5/4r3/2Pq4/P3P1P1/1PQ4P/R3KB1R w KQ - 0 19"]

1.d4 {[%clk 0:44:36]} Bb7 {[%clk 0:45:16]} 2.c4 {[%clk 0:44:49]} e5 {[%clk 0:42:40]} 3.Nf3 {[%clk 0:41:47]} exd4 {[%clk 0:40:56]} 4.Nxd4 {[%clk 0:41:29]} Bc5 {[%clk 0:39:48]} 5.Nc3 {[%clk 0:37:12]} Nf6 {[%clk 0:38:09]} 6.Nb3 {[%clk 0:36:30]} Bb4 {[%clk 0:37:26]} 7.Bd2 {[%clk 0:36:43]} O-O {[%clk 0:35:42]} 8.a3 {[%clk 0:34:29]} Bd6 {[%clk 0:33:46]} 9.Nd4 {[%clk 0:26:26]} Nc6 {[%clk 0:30:55]} 10.Nxc6 {[%clk 0:22:03]} Bxc6 {[%clk 0:28:23]} 11.e3 {[%clk 0:22:31]} Rb8 {[%clk 0:26:43]} 12.Qc2 {[%clk 0:21:52]} Re8 {[%clk 0:25:16]} 13.Nb5 {[%clk 0:18:40]} Be5 {[%clk 0:24:38]} 14.Bc3 {[%clk 0:18:46]} Ng4 {[%clk 0:24:31]} 15.Bxe5 {[%clk 0:15:14]} Rxe5 {[%clk 0:23:47]} 16.Nd4 {[%clk 0:12:10]} Qh4 {[%clk 0:23:01]} 17.g3 {[%clk 0:10:59]} Nxe3 {[%clk 0:22:54]} 18.fxe3 {[%clk 0:07:47]} Qxd4 {[%clk 0:18:35]} 0-1[/pgn]


when looking over this game SF was quite insistent (at fairly low depths) that 3.d5 was best; out of curiosity i let it run on 3 threads & gave komodo (with white contempt at -67, else default) 5. both had 5 man TBs. SF is a recent dev version, i think 17 august.

i was generally harsh with SF & lenient with K with regard to time, definitely biased, & the game was generally pretty rapid. neverthless the game & result surprised me so much i gave it another go, reversing to where i thought the key mistake was made. thought i'd share the games partly out of general interest & partly bc i believe it'd be entirely possible to run similar matches automatically, which isn't something i'd thought possible.

Game 1
[pgn]
[Event "Odds Match"]
[Site "?"]
[Date "2017.08.22"]
[Round "?"]
[White "Stockfish"]
[Black "Komodo 11.2"]
[ECO "A40"]
[Result "1/2-1/2"]
[SetUp "1"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -"]

1. d4 Bb7 2. c4 e5 3. d5 Bc5 4. e3 d6 5. Nc3 Nd7 6. Bd3 a5
7. Nge2 Nh6 8. O-O f5 9. b3 O-O 10. Na4 Ba7 11. Ba3 Rf6
{Komodo Ng3 +1.00} 12. Qd2 $6 Ng4 $11 {Komodo (SF +1.45,
d30)} 13. h3 e4 14. Bc2 Rh6 {Komodo 0.00 d28 (Ng3)} 15. Bb2
Qg5 16. Nf4 Nde5 17. Qe2 Nf3+ 18. Kh1 Nge5 19. Rg1 Qg4
20. Bd1 Nxg1 21. Kxg1 Qh4 22. Qc2 Nd3 23. Nxd3 exd3
24. Qxd3 Re8 25. Bf3 Rg6 26. Re1 Qxh3 27. Kf1 Qh2 28. Qxf5
Rf8 29. Qe4 Bc8 30. Ke2 $11 1/2-1/2[/pgn]

"Rewind" Game 2:

[pgn]
[Event "Odds Match"]
[Site "?"]
[Date "2017.08.22"]
[Round "?"]
[White "Stockfish"]
[Black "Komodo 11.2"]
[ECO "A40"]
[Result "1/2-1/2"]
[SetUp "1"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -"]

1. d4 Bb7 2. c4 e5 3. d5 Bc5 4. e3 d6 5. Nc3 Nd7 6. Bd3 a5
7. Nge2 Nh6 8. O-O f5 9. b3 O-O 10. Na4 Ba7 11. Ba3 Rf6
{Komodo Ng3 +1.00} 12. Ng3 Qe7 13. Bc2 Raf8 14. Nc3 Bc8
15. Qe2 Nc5 16. Rae1 Rg6 17. h3 Bb6 18. Nb5 Bd7 19. Bb1 Qf6
20. Bc1 Qh4 21. Qh5 Qd8 22. Rd1 f4 23. exf4 exf4 24. Bxf4
Rxf4 25. Bxg6 hxg6 26. Qxg6 Nf5 27. Nxf5 Rxf5 28. Rde1 Rf6
29. Qg3 Bf5 30. Nd4 Ne4 31. Qe3 {SF +.80 K11 0.00} Bh7
32. Rd1 Rf8 33. Kh2 Qf6 34. f3 Qe5+ 35. f4 Qf6 36. g4 Re8
37. g5 Qe7 38. a3 Qf7 39. Qf3 Bxd4 40. Rxd4 Qe7 41. Qg4 Qa7
42. Qh5 Rb8 43. Rxe4 Bxe4 44. g6 {SF +2.60 K 0.00} Qd4
45. Qh7+ Kf8 46. Qh8+ Ke7 47. Qxb8 Qb2+ 48. Kg3 Qc3+
49. Kh4 Qf6+ 50. Kg4 Qxg6+ 51. Kh4 Qf6+ 52. Kg3 Qc3+
53. Kf2 Qd2+ 54. Kg3 Qe3+ 55. Kh4 Bf5 56. Qc7+ Ke8 57. Kg5
Bxh3 58. Kg6 Bxf1 59. Qxd6 Bd3+ 60. Kxg7 Qg3+ 61. Kf6 Qg6+
62. Ke5 Qe4+ 63. Kf6 Qg6+ $11 1/2-1/2[/pgn]

NOTE if any SF moves aren't reproducible, it's probably (unintentional) operator error as this was quite hastily done, but i don't believe, if this was a factor, it was significant.

[Lyudmil Tsvetkov]

nothing to be surprised at: Komodo excels at such positions bearing more open character, while SF is a bit inferior.

obviously, Komodo has better tuned eval parameters and search for such positions, like mobility, piece activity, etc.

[AdminX]

Funny you should post this topic, as I was looking at this position just the other day.

[d]rnbq1bnr/pppppppp/8/4k3/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1

Odds match with Komodo vs Larry Gilden.

[pgn]
[Event "Bethesda odds matches"]
[Site "Bethesda, Maryland, USA"]
[Date "2015.09.16"]
[Round "1"]
[White "Larry Gilden"]
[Black "Komodo 9.2"]
[Result "0-1"]
[Annotator "Summers Jr.,Ted"]
[SetUp "1"]
[FEN "rnbq1bnr/pppppppp/8/4k3/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1"]
[PlyCount "42"]
[EventDate "2015.??.??"]

{[%evp 0,42,32767,32767,32767,32767,32767,32767,32767,32767,32767,27,28,0,58,0,
62,41,49,-68,-32,-83,-5,-85,0,-91,-68,-150,-132,-222,-209,-230,-208,-242,-195,
-311,-298,-338,-286,-316,-280,-462,-435,-464,-437]} {6144MB, 24core} 1. d4+ {
[%emt 0:00:00]} Kf6 {[%emt 0:00:15] eval 38 depth=24} 2. e4 {0:00:51] (Bf4)} g6
{[%emt 0:00:09] eval 1 depth=21} 3. Nc3 {[%emt 0:08:05]} Kg7 {[%emt 0:00:00]
eval 0 depth=32} 4. f4 {0:02:21] (h4)} e6 {[%emt 0:00:46] eval -58 depth=27} 5.
d5 {0:02:47] (Be2)} exd5 {[%emt 0:00:32] eval -52 depth=25} 6. Qd4+ {0:01:29]
(Qxd5).} f6 {[%emt 0:00:14] eval -74 depth=24. Threatening ...dxe4.} 7. exd5 {
0:01:01] (Qxd5)} b6 {[%eval 62,23] [%emt 0:00:46] eval -56 depth=25} (7... c6
$11 {[%eval 0,22]}) 8. f5 $14 {0:10:31] (Bd2). White is slightly better.} Bc5 {
[%emt 0:01:12] eval -53 depth=24} 9. Qg4 {[%eval -68,23] 0:01:39] (Qh4)} (9.
Qf4 $14 {[%eval 49,22]}) 9... Qe8+ {[%emt 0:00:46] eval -123 depth=26} 10. Kd1
{[%emt 0:02:26]} d6 {[%eval -5,24] [%emt 0:00:00] eval -106 depth=30} (10...
Ne7 $17 {[%eval -83,22]}) 11. Bb5 {[%eval -85,23] 0:04:48] (Bd3)} (11. Nf3 $11
{[%eval -5,24]}) 11... c6 {[%eval 0,24] [%emt 0:00:25] eval -160, 25]} ({
Black should play} 11... Qf7 $17 {[%eval -85,23] aiming for ...Ne7.} 12. Qf4 g5
) {[%tqu ]} 12. dxc6 {[%eval -91,25] 0:06:05] (Bd3)} (12. Bd3 $1 $11 {[%eval 0,
24] remains equal.}) 12... Nxc6 $17 {[%emt 0:00:56] eval -163 depth=27} 13. Qe4
{[%eval -150,24] 0:01:08] (Nf3)} ({Better is} 13. Nf3 $15 {[%eval -68,22]} Bxf5
14. Qf4) 13... Nge7 {[%emt 0:00:10][%cal Bg8e7,Be7c6,Bc6e5][%mdl 32] eval -273
depth=23} 14. Bxc6 {[%eval -222,27] 0:05:37] (fxg6)} (14. g4 $17 {[%eval -132,
24]}) 14... Nxc6 $19 {[%emt 0:00:12] eval -284 depth=28} 15. Qxe8 {[%emt 0:02:
46]} Rxe8 {[%emt 0:00:00] eval -294 depth=33} 16. fxg6 {[%emt 0:00:12]} Ne5 {
[%emt 0:00:39] eval -300 depth=30. And now ...Bb7 would win.} (16... Kxg6 17.
Nge2 $17) 17. h3 {[%eval -311,27] 0:02:19] (Nf3)} (17. Nf3 {[%eval -195,24]
was necessary.}) 17... Bb7 {[%emt 0:00:08] eval -379 depth=27} ({Weaker is}
17... Nxg6 18. Nf3 $17) ({Not} 17... Kxg6 18. Nge2 $17) (17... hxg6 18. Nf3 $19
) 18. Nge2 {0:00:38] (gxh7)} Bxg2 {[%emt 0:00:15] eval -391 depth=29} (18...
Nxg6 $2 19. Nf4 $19) ({Resist} 18... hxg6 19. Rh2 $19) 19. Rh2 {[%emt 0:00:39]}
Bf3 {[%emt 0:00:04] eval -390 depth=33. Hoping for ...Bg1.} 20. Ke1 $2 {
[%eval -462,29] [#] 0:01:25] (gxh7)} (20. gxh7 {[%eval -280,24]} Bg1 21. Rg2+
Bxg2 22. Nxg1) 20... Bb7 {[%emt 0:00:15] eval -518 depth=28} 21. Kf1 {[%emt 0:
00:45]} hxg6 {[%emt 0:00:48] eval -557 depth=32} 0-1

[/pgn]

[Lyudmil Tsvetkov]

for some reason, I have incredibly easy time winning against Komodo 10.1, contempt=150, from the positions Rensch played with.

for example, this one:

[pgn][Event "Blitz 1m"]
[Site "Microsoft"]
[Date "2017.08.23"]
[Round "?"]
[White "Lyudmil Tsvetkov, owner"]
[Black "Komodo 10.1 64-bit"]
[Result "1-0"]
[Annotator "owner"]
[SetUp "1"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "119"]
[EventDate "2017.??.??"]
[TimeControl "60"]

{512MB, OWNER-PC} 1. d4 {0} Bb7 {-0.48/16 2} 2. Nf3 {3} Nf6 {-0.28/17 1} 3. g3
{2} d5 {-0.46/17 2} 4. Bg2 {1} e6 {-0.36/17 0} 5. O-O {3} Nbd7 {-0.24/18 2} 6.
c3 {4} Rc8 {-0.57/17 2} 7. Bg5 {2} Be7 {-0.41/16 2} 8. Nbd2 {1} O-O {-0.53/18 2
} 9. Re1 {3} h6 {-0.53/18 2} 10. Bxf6 {3} gxf6 {-0.46/18 1} 11. e4 {0} dxe4 {
-0.44/15 1} 12. Nxe4 {1} f5 {-0.29/17 2} 13. Ned2 {2} Qc7 {-0.25/17 0} 14. Rb1
{0} Rfd8 {-0.36/15 3} 15. Ne5 {3} Bxg2 {-0.40/16 1} 16. Kxg2 {2} Nxe5 {-0.27/
18 0} 17. Rxe5 {2} Bd6 {-0.19/17 1} 18. Re1 {4} Kh7 {-0.15/18 0} 19. Nf3 {4}
Qa5 {0.00/16 3} 20. a3 {6} Qb5 {-0.05/17 4} 21. Qe2 {7} Qc6 {0.06/16 1} 22. Kg1
{0} a5 {0.12/16 1} 23. Ne5 {5} Bxe5 {0.13/16 1} 24. Qxe5 {7} Rd5 {0.00/16 1}
25. Qf6 {3} Qb7 {0.37/17 1} 26. Re5 {12} Rxe5 {0.00/21 1} 27. dxe5 {9} a4 {0.
00/22 1} 28. Rd1 {5} Rg8 {0.00/27 0} 29. Rd2 {0} Qe4 {0.00/23 0} 30. Rd8 {0}
Qe1+ {0.00/28 1} 31. Kg2 {2} Qe4+ {0.00/30 0} 32. Kf1 {3} Qh1+ {0.00/26 1} 33.
Ke2 {2} Qe4+ {0.00/28 0} 34. Kd2 {5} Rxd8+ {0.00/25 1} 35. Qxd8 {1} Qb7 {0.00/
28 0} 36. Kc2 {0} Qe4+ {0.00/27 0} 37. Qd3 {11} Qxe5 {0.00/17 0} 38. b3 {4} Qe1
{-0.31/17 1} 39. bxa4 {13} Qxf2+ {0.00/28 0} 40. Kb3 {2} Qc5 {0.00/26 0} 41.
Qb5 {0} Qd6 {0.00/25 1} 42. a5 {4} Qd1+ {0.12/24 0} 43. Kb4 {3} f4 {0.52/19 0}
44. gxf4 {7} Qd6+ {1.23/21 0} 45. Qc5 {5} Qxf4+ {1.49/21 0} 46. c4 {20} Qb8+ {
2.07/23 0} 47. Qb6 {7} Qxh2 {1.23/15 0} 48. a6 {0} Qb2+ {2.74/19 2} 49. Kc5 {1}
Qxa3+ {2.82/19 0} 50. Kc6 {2} Qf3+ {3.12/20 0} 51. Kc7 {5} Qf4+ {3.76/21 0} 52.
Kd8 {0} Qh4+ {0.90/18 1} 53. Ke8 {0} Qxc4 {3.13/22 1} 54. a7 {5} Qa4+ {3.26/16
0} 55. Kd8 {0} Qe4 {3.30/14 0} 56. Kc7 {14} Qe5+ {5.85/18 0} 57. Kc8 {10} Qe4 {
8.51/17 1} 58. Qb7 {0} Qc4+ {250.00/17 1} 59. Kd8 {0} Kg6 {#27/14 0} 60. a8=Q {
7} 1-0

[/pgn]

this is just on 4 cores and blitz mode(bullet for Komodo), so might partly explain the puzzle.

[yanquis1972]

at least as far as rensch knew, he was playing against larry kaufmann's personal setup (hardware, software, tuning; although he only mentioned super-high-end hardware).

you're using a version quite a bit older + reaching a depth of ~17; safe to say the version he was playing against was routinely over depth 30, possibly easily so.

[JJJ]

for some reason, I have incredibly easy time winning against Komodo 10.1, contempt=150, from the positions Rensch played with.

for example, this one:

[pgn][Event "Blitz 1m"]
[Site "Microsoft"]
[Date "2017.08.23"]
[Round "?"]
[White "Lyudmil Tsvetkov, owner"]
[Black "Komodo 10.1 64-bit"]
[Result "1-0"]
[Annotator "owner"]
[SetUp "1"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "119"]
[EventDate "2017.??.??"]
[TimeControl "60"]

{512MB, OWNER-PC} 1. d4 {0} Bb7 {-0.48/16 2} 2. Nf3 {3} Nf6 {-0.28/17 1} 3. g3
{2} d5 {-0.46/17 2} 4. Bg2 {1} e6 {-0.36/17 0} 5. O-O {3} Nbd7 {-0.24/18 2} 6.
c3 {4} Rc8 {-0.57/17 2} 7. Bg5 {2} Be7 {-0.41/16 2} 8. Nbd2 {1} O-O {-0.53/18 2
} 9. Re1 {3} h6 {-0.53/18 2} 10. Bxf6 {3} gxf6 {-0.46/18 1} 11. e4 {0} dxe4 {
-0.44/15 1} 12. Nxe4 {1} f5 {-0.29/17 2} 13. Ned2 {2} Qc7 {-0.25/17 0} 14. Rb1
{0} Rfd8 {-0.36/15 3} 15. Ne5 {3} Bxg2 {-0.40/16 1} 16. Kxg2 {2} Nxe5 {-0.27/
18 0} 17. Rxe5 {2} Bd6 {-0.19/17 1} 18. Re1 {4} Kh7 {-0.15/18 0} 19. Nf3 {4}
Qa5 {0.00/16 3} 20. a3 {6} Qb5 {-0.05/17 4} 21. Qe2 {7} Qc6 {0.06/16 1} 22. Kg1
{0} a5 {0.12/16 1} 23. Ne5 {5} Bxe5 {0.13/16 1} 24. Qxe5 {7} Rd5 {0.00/16 1}
25. Qf6 {3} Qb7 {0.37/17 1} 26. Re5 {12} Rxe5 {0.00/21 1} 27. dxe5 {9} a4 {0.
00/22 1} 28. Rd1 {5} Rg8 {0.00/27 0} 29. Rd2 {0} Qe4 {0.00/23 0} 30. Rd8 {0}
Qe1+ {0.00/28 1} 31. Kg2 {2} Qe4+ {0.00/30 0} 32. Kf1 {3} Qh1+ {0.00/26 1} 33.
Ke2 {2} Qe4+ {0.00/28 0} 34. Kd2 {5} Rxd8+ {0.00/25 1} 35. Qxd8 {1} Qb7 {0.00/
28 0} 36. Kc2 {0} Qe4+ {0.00/27 0} 37. Qd3 {11} Qxe5 {0.00/17 0} 38. b3 {4} Qe1
{-0.31/17 1} 39. bxa4 {13} Qxf2+ {0.00/28 0} 40. Kb3 {2} Qc5 {0.00/26 0} 41.
Qb5 {0} Qd6 {0.00/25 1} 42. a5 {4} Qd1+ {0.12/24 0} 43. Kb4 {3} f4 {0.52/19 0}
44. gxf4 {7} Qd6+ {1.23/21 0} 45. Qc5 {5} Qxf4+ {1.49/21 0} 46. c4 {20} Qb8+ {
2.07/23 0} 47. Qb6 {7} Qxh2 {1.23/15 0} 48. a6 {0} Qb2+ {2.74/19 2} 49. Kc5 {1}
Qxa3+ {2.82/19 0} 50. Kc6 {2} Qf3+ {3.12/20 0} 51. Kc7 {5} Qf4+ {3.76/21 0} 52.
Kd8 {0} Qh4+ {0.90/18 1} 53. Ke8 {0} Qxc4 {3.13/22 1} 54. a7 {5} Qa4+ {3.26/16
0} 55. Kd8 {0} Qe4 {3.30/14 0} 56. Kc7 {14} Qe5+ {5.85/18 0} 57. Kc8 {10} Qe4 {
8.51/17 1} 58. Qb7 {0} Qc4+ {250.00/17 1} 59. Kd8 {0} Kg6 {#27/14 0} 60. a8=Q {
7} 1-0

[/pgn]

this is just on 4 cores and blitz mode(bullet for Komodo), so might partly explain the puzzle.

Anyway dear Lyudmil, I m still waiting for you to play an official handicap match against Komodo. You are probably the best player in the world against computer.

[Lyudmil Tsvetkov]

at least as far as rensch knew, he was playing against larry kaufmann's personal setup (hardware, software, tuning; although he only mentioned super-high-end hardware).

you're using a version quite a bit older + reaching a depth of ~17; safe to say the version he was playing against was routinely over depth 30, possibly easily so.

you are right, but on the other hand, Rensch also had much more time.

do you manage sometimes to beat SF or Komodo like that?

[Lyudmil Tsvetkov]

for some reason, I have incredibly easy time winning against Komodo 10.1, contempt=150, from the positions Rensch played with.

for example, this one:

[pgn][Event "Blitz 1m"]
[Site "Microsoft"]
[Date "2017.08.23"]
[Round "?"]
[White "Lyudmil Tsvetkov, owner"]
[Black "Komodo 10.1 64-bit"]
[Result "1-0"]
[Annotator "owner"]
[SetUp "1"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "119"]
[EventDate "2017.??.??"]
[TimeControl "60"]

{512MB, OWNER-PC} 1. d4 {0} Bb7 {-0.48/16 2} 2. Nf3 {3} Nf6 {-0.28/17 1} 3. g3
{2} d5 {-0.46/17 2} 4. Bg2 {1} e6 {-0.36/17 0} 5. O-O {3} Nbd7 {-0.24/18 2} 6.
c3 {4} Rc8 {-0.57/17 2} 7. Bg5 {2} Be7 {-0.41/16 2} 8. Nbd2 {1} O-O {-0.53/18 2
} 9. Re1 {3} h6 {-0.53/18 2} 10. Bxf6 {3} gxf6 {-0.46/18 1} 11. e4 {0} dxe4 {
-0.44/15 1} 12. Nxe4 {1} f5 {-0.29/17 2} 13. Ned2 {2} Qc7 {-0.25/17 0} 14. Rb1
{0} Rfd8 {-0.36/15 3} 15. Ne5 {3} Bxg2 {-0.40/16 1} 16. Kxg2 {2} Nxe5 {-0.27/
18 0} 17. Rxe5 {2} Bd6 {-0.19/17 1} 18. Re1 {4} Kh7 {-0.15/18 0} 19. Nf3 {4}
Qa5 {0.00/16 3} 20. a3 {6} Qb5 {-0.05/17 4} 21. Qe2 {7} Qc6 {0.06/16 1} 22. Kg1
{0} a5 {0.12/16 1} 23. Ne5 {5} Bxe5 {0.13/16 1} 24. Qxe5 {7} Rd5 {0.00/16 1}
25. Qf6 {3} Qb7 {0.37/17 1} 26. Re5 {12} Rxe5 {0.00/21 1} 27. dxe5 {9} a4 {0.
00/22 1} 28. Rd1 {5} Rg8 {0.00/27 0} 29. Rd2 {0} Qe4 {0.00/23 0} 30. Rd8 {0}
Qe1+ {0.00/28 1} 31. Kg2 {2} Qe4+ {0.00/30 0} 32. Kf1 {3} Qh1+ {0.00/26 1} 33.
Ke2 {2} Qe4+ {0.00/28 0} 34. Kd2 {5} Rxd8+ {0.00/25 1} 35. Qxd8 {1} Qb7 {0.00/
28 0} 36. Kc2 {0} Qe4+ {0.00/27 0} 37. Qd3 {11} Qxe5 {0.00/17 0} 38. b3 {4} Qe1
{-0.31/17 1} 39. bxa4 {13} Qxf2+ {0.00/28 0} 40. Kb3 {2} Qc5 {0.00/26 0} 41.
Qb5 {0} Qd6 {0.00/25 1} 42. a5 {4} Qd1+ {0.12/24 0} 43. Kb4 {3} f4 {0.52/19 0}
44. gxf4 {7} Qd6+ {1.23/21 0} 45. Qc5 {5} Qxf4+ {1.49/21 0} 46. c4 {20} Qb8+ {
2.07/23 0} 47. Qb6 {7} Qxh2 {1.23/15 0} 48. a6 {0} Qb2+ {2.74/19 2} 49. Kc5 {1}
Qxa3+ {2.82/19 0} 50. Kc6 {2} Qf3+ {3.12/20 0} 51. Kc7 {5} Qf4+ {3.76/21 0} 52.
Kd8 {0} Qh4+ {0.90/18 1} 53. Ke8 {0} Qxc4 {3.13/22 1} 54. a7 {5} Qa4+ {3.26/16
0} 55. Kd8 {0} Qe4 {3.30/14 0} 56. Kc7 {14} Qe5+ {5.85/18 0} 57. Kc8 {10} Qe4 {
8.51/17 1} 58. Qb7 {0} Qc4+ {250.00/17 1} 59. Kd8 {0} Kg6 {#27/14 0} 60. a8=Q {
7} 1-0

[/pgn]

this is just on 4 cores and blitz mode(bullet for Komodo), so might partly explain the puzzle.

Anyway dear Lyudmil, I m still waiting for you to play an official handicap match against Komodo. You are probably the best player in the world against computer.

I am the best player in the world against computers, on my own hardware, when I am in good mood, have some luck, and engine commits at least one major inaccuracy.

why do you need a formal match?

everything is informal in our world of today.

[yanquis1972]

at least as far as rensch knew, he was playing against larry kaufmann's personal setup (hardware, software, tuning; although he only mentioned super-high-end hardware).

you're using a version quite a bit older + reaching a depth of ~17; safe to say the version he was playing against was routinely over depth 30, possibly easily so.

you are right, but on the other hand, Rensch also had much more time.

do you manage sometimes to beat SF or Komodo like that?

never tried, but probably never would, since i don't think i could convert any endgame wihtout takebacks. against a current top engine i suppose in an odds match, with very little time (& prob cores) allowed for the computer, i might get lucky enough to achieve a totally dominant position (did this once, with a simple pawnstorm against hiarcs a few years ago -- i think i posted about it here, in fact -- & still consider it 'the time i beat a computer at blitz'), but even at low depth my complete lack of knowledge & poor tactics would always catch me out.

[MikeB]

didn't realize it till i stumbled on a game the other night but the komodo team is still playing (casual) odds matches against strong masters; latest victim being IM danny rensch (games can be found here; there are also streams available on the chess.com twitch channel)

the latest one, which is relevant here:

[pgn][Event "Live Chess - Odds Chess"]
[Site "Chess.com"]
[Date "2017.08.22"]
[White "DanielRensch"]
[Black "KomodoChess"]
[Result "0-1"]
[ECO "A40"]
[WhiteElo "2448"]
[BlackElo "3368"]
[TimeControl "2700+30"]
[EndTime "15:22:41 PDT"]
[Termination "KomodoChess won by resignation"]
[SetUp "1"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -"]
[Variant "Odds Chess"]
[CurrentPosition "1r4k1/p2p1ppp/2b5/4r3/2Pq4/P3P1P1/1PQ4P/R3KB1R w KQ - 0 19"]

1.d4 {[%clk 0:44:36]} Bb7 {[%clk 0:45:16]} 2.c4 {[%clk 0:44:49]} e5 {[%clk 0:42:40]} 3.Nf3 {[%clk 0:41:47]} exd4 {[%clk 0:40:56]} 4.Nxd4 {[%clk 0:41:29]} Bc5 {[%clk 0:39:48]} 5.Nc3 {[%clk 0:37:12]} Nf6 {[%clk 0:38:09]} 6.Nb3 {[%clk 0:36:30]} Bb4 {[%clk 0:37:26]} 7.Bd2 {[%clk 0:36:43]} O-O {[%clk 0:35:42]} 8.a3 {[%clk 0:34:29]} Bd6 {[%clk 0:33:46]} 9.Nd4 {[%clk 0:26:26]} Nc6 {[%clk 0:30:55]} 10.Nxc6 {[%clk 0:22:03]} Bxc6 {[%clk 0:28:23]} 11.e3 {[%clk 0:22:31]} Rb8 {[%clk 0:26:43]} 12.Qc2 {[%clk 0:21:52]} Re8 {[%clk 0:25:16]} 13.Nb5 {[%clk 0:18:40]} Be5 {[%clk 0:24:38]} 14.Bc3 {[%clk 0:18:46]} Ng4 {[%clk 0:24:31]} 15.Bxe5 {[%clk 0:15:14]} Rxe5 {[%clk 0:23:47]} 16.Nd4 {[%clk 0:12:10]} Qh4 {[%clk 0:23:01]} 17.g3 {[%clk 0:10:59]} Nxe3 {[%clk 0:22:54]} 18.fxe3 {[%clk 0:07:47]} Qxd4 {[%clk 0:18:35]} 0-1[/pgn]


when looking over this game SF was quite insistent (at fairly low depths) that 3.d5 was best; out of curiosity i let it run on 3 threads & gave komodo (with white contempt at -67, else default) 5. both had 5 man TBs. SF is a recent dev version, i think 17 august.

i was generally harsh with SF & lenient with K with regard to time, definitely biased, & the game was generally pretty rapid. neverthless the game & result surprised me so much i gave it another go, reversing to where i thought the key mistake was made. thought i'd share the games partly out of general interest & partly bc i believe it'd be entirely possible to run similar matches automatically, which isn't something i'd thought possible.

Game 1
[pgn]
[Event "Odds Match"]
[Site "?"]
[Date "2017.08.22"]
[Round "?"]
[White "Stockfish"]
[Black "Komodo 11.2"]
[ECO "A40"]
[Result "1/2-1/2"]
[SetUp "1"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -"]

1. d4 Bb7 2. c4 e5 3. d5 Bc5 4. e3 d6 5. Nc3 Nd7 6. Bd3 a5
7. Nge2 Nh6 8. O-O f5 9. b3 O-O 10. Na4 Ba7 11. Ba3 Rf6
{Komodo Ng3 +1.00} 12. Qd2 $6 Ng4 $11 {Komodo (SF +1.45,
d30)} 13. h3 e4 14. Bc2 Rh6 {Komodo 0.00 d28 (Ng3)} 15. Bb2
Qg5 16. Nf4 Nde5 17. Qe2 Nf3+ 18. Kh1 Nge5 19. Rg1 Qg4
20. Bd1 Nxg1 21. Kxg1 Qh4 22. Qc2 Nd3 23. Nxd3 exd3
24. Qxd3 Re8 25. Bf3 Rg6 26. Re1 Qxh3 27. Kf1 Qh2 28. Qxf5
Rf8 29. Qe4 Bc8 30. Ke2 $11 1/2-1/2[/pgn]

"Rewind" Game 2:

[pgn]
[Event "Odds Match"]
[Site "?"]
[Date "2017.08.22"]
[Round "?"]
[White "Stockfish"]
[Black "Komodo 11.2"]
[ECO "A40"]
[Result "1/2-1/2"]
[SetUp "1"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -"]

1. d4 Bb7 2. c4 e5 3. d5 Bc5 4. e3 d6 5. Nc3 Nd7 6. Bd3 a5
7. Nge2 Nh6 8. O-O f5 9. b3 O-O 10. Na4 Ba7 11. Ba3 Rf6
{Komodo Ng3 +1.00} 12. Ng3 Qe7 13. Bc2 Raf8 14. Nc3 Bc8
15. Qe2 Nc5 16. Rae1 Rg6 17. h3 Bb6 18. Nb5 Bd7 19. Bb1 Qf6
20. Bc1 Qh4 21. Qh5 Qd8 22. Rd1 f4 23. exf4 exf4 24. Bxf4
Rxf4 25. Bxg6 hxg6 26. Qxg6 Nf5 27. Nxf5 Rxf5 28. Rde1 Rf6
29. Qg3 Bf5 30. Nd4 Ne4 31. Qe3 {SF +.80 K11 0.00} Bh7
32. Rd1 Rf8 33. Kh2 Qf6 34. f3 Qe5+ 35. f4 Qf6 36. g4 Re8
37. g5 Qe7 38. a3 Qf7 39. Qf3 Bxd4 40. Rxd4 Qe7 41. Qg4 Qa7
42. Qh5 Rb8 43. Rxe4 Bxe4 44. g6 {SF +2.60 K 0.00} Qd4
45. Qh7+ Kf8 46. Qh8+ Ke7 47. Qxb8 Qb2+ 48. Kg3 Qc3+
49. Kh4 Qf6+ 50. Kg4 Qxg6+ 51. Kh4 Qf6+ 52. Kg3 Qc3+
53. Kf2 Qd2+ 54. Kg3 Qe3+ 55. Kh4 Bf5 56. Qc7+ Ke8 57. Kg5
Bxh3 58. Kg6 Bxf1 59. Qxd6 Bd3+ 60. Kxg7 Qg3+ 61. Kf6 Qg6+
62. Ke5 Qe4+ 63. Kf6 Qg6+ $11 1/2-1/2[/pgn]

NOTE if any SF moves aren't reproducible, it's probably (unintentional) operator error as this was quite hastily done, but i don't believe, if this was a factor, it was significant.

From your 3.d5 position tc 3 min 3 sec inc with 3 cores - i don't think SF would have a problem with this at all. The second is from the starting position.
[pgn]
[Event "Mac Pro x5690 3.46 Ghz 18 CPU"]
[Site "Mac-Pro.local"]
[Date "2017.08.23"]
[Round "-"]
[White "SF-McB 210817 64 POPCNT"]
[Black "Komodo 11.2.2 64-bit"]
[Result "1-0"]
[TimeControl "180+3"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1"]
[SetUp "1"]


1. d4 Bb7 2. c4 e5 3. d5 {+2.36/24 +175} Bc5 {-1.70/22 +161} 4. Nc3
{+2.60/24 +175} d6 {-1.80/23 +151} 5. e3 {+2.60/25 +172} Nd7
{-1.76/23 +146} 6. Bd3 {+2.72/24 +170} Bb6 {-1.86/23 +131} 7. Nge2
{+2.73/25 +161} Ngf6 {-1.84/22 +124} 8. O-O {+2.72/22 +163} Rc8
{-1.87/22 +113} 9. b3 {+2.79/23 +159} O-O {-1.88/22 +111} 10. Bf5
{+2.92/24 +150} Rb8 {-1.93/21 +90} 11. Na4 {+2.96/24 +146} Ba5
{-1.97/20 +89} 12. Bc2 {+3.05/25 +144} Qe7 {-1.89/21 +78} 13. e4
{+3.28/23 +139} Nb6 {-2.09/22 +72} 14. Nb2 {+3.32/26 +132} Rfc8
{-2.12/22 +66} 15. Ng3 {+3.30/26 +123} g6 {-2.22/24 +59} 16. Bg5
{+3.26/23 +125} Nbd7 {-2.21/24 +52} 17. h3 {+3.29/25 +113} Qf8
{-2.27/23 +47} 18. Na4 {+3.26/24 +109} Ra8 {-2.29/22 +42} 19. Rb1
{+3.37/24 +107} Bd8 {-2.47/22 +37} 20. Qd2 {+3.27/26 +98} Kh8
{-2.44/22 +34} 21. Be3 {+3.43/24 +85} Nb6 {-2.46/19 +35} 22. Nc3
{+3.47/23 +87} Nfd7 {-2.45/21 +32} 23. Nge2 {+3.54/23 +84} Qe7
{-2.74/22 +31} 24. f4 {+3.42/26 +76} exf4 {-2.66/22 +30} 25. Bxf4
{+3.51/26 +73} Ne5 {-2.49/21 +30} 26. Rf2 {+3.55/22 +74} Nbd7
{-2.67/19 +31} 27. Na4 {+3.67/23 +74} Qf8 {-2.86/23 +23} 28. Rbf1
{+3.95/24 +71} Bh4 {-3.07/23 +18} 29. g3 {+3.89/25 +67} Be7 {-2.98/23 +16}
30. Bh6 {+4.07/22 +69} Qe8 {-3.14/24 +12} 31. Qc3 {+4.37/27 +55} f6
{-3.05/21 +14} 32. Nd4 {+4.37/25 +51} Nf7 {-3.12/21 +14} 33. Bd2
{+4.51/22 +50} Kg8 {-3.04/20 +14} 34. g4 {+4.50/23 +44} Nfe5 {-3.10/19 +12}
35. Qg3 {+4.94/21 +41} Rab8 {-3.47/19 +12} 36. Bh6 {+4.94/23 +36} Bd8
{-3.56/20 +9} 37. Ne6 {+5.87/20 +37} g5 {-3.63/21 +10} 38. h4
{+7.06/22 +34} Be7 {-5.07/21 +7} 39. hxg5 {+8.05/21 +33} fxg5 {-5.02/22 +8}
40. Rf5 {+7.32/19 +35} Qg6 {-5.09/23 +8} 41. Bxg5 {+9.41/23 +35} Re8
{-5.14/22 +10} 42. Qh4 {+10.37/25 +32} Bf8 {-6.87/22 +4} 43. Nf4
{+10.31/24 +34} Qg7 {-6.96/21 +3} 44. Nh5 {+10.60/26 +32} Qg6 {-7.50/19 +5}
45. Nf4 {+10.58/27 +28} Qg7 {-7.08/22 +3} 46. Nh5 {+10.96/26 +27} Qg6
{-7.65/21 +3} 47. Bf6 {+11.10/20 +29} h6 {-8.13/19 +5} 48. g5
{+11.12/25 +28} Bc8 {-8.57/21 +6} 49. Nf4 {+11.19/23 +30} Qh7 {-8.79/23 +6}
50. gxh6 {+11.49/26 +26} Nxf6 {-8.99/22 +6} 51. Qg5+ {+11.51/17 +29} Bg7
{-9.24/23 +5} 52. Rxf6 {+11.62/24 +29} Bg4 {-9.36/22 +6} 53. Rxd6
{+11.56/21 +31} Nf7 {-9.60/21 +3} 54. Qxg4 {+12.26/23 +33} Nxd6
{-10.47/21 +4} 55. hxg7 {+12.50/20 +35} Qxg7 {-11.23/21 +5} 56. Qxg7+
{+12.81/22 +35} Kxg7 {-7.98/5 +8} 57. Nc5 {+13.30/22 +36} Re5
{-11.22/20 +6} 58. Nd7 {+14.06/23 +36} Rg5+ {-11.54/23 +6} 59. Kf2
{+14.23/21 +38} Re8 {-12.19/23 +7} 60. Ne6+ {+14.78/24 +38} Rxe6
{-12.32/22 +7} 61. dxe6 {+15.06/22 +41} Rg6 {-13.64/25 +5} 62. e5
{+25.51/23 +41} Rxe6 {-10.03/21 +7} 63. exd6 {+24.22/20 +44} Rxd6
{-10.44/21 +9} 64. Bf5 {+51.91/23 +44} Kf7 {-17.67/22 +6} 65. Rh1
{+132.65/27 +45} Ke8 {-250.00/22 +5} 66. Ke3 {+132.70/34 +41} Rc6
{-250.00/22 +8} 67. Rh8+ {+1000.14/32 +41} Ke7 {-250.00/23 +11} 68. Rh7+
{+1000.13/35 +40} Kd8 {-250.00/24 +13} 69. Ne5 {+1000.11/36 +39} Rc7
{-250.00/23 +15} 70. Bd7 {+1000.08/39 +39} Rc5 {-1000.08/23 +14} 71. Kd4
{+1000.07/39 +41} Rxe5 {-1000.07/25 +17} 72. Kxe5 {+1000.06/49 +41} Kc7
{-1000.05/27 +19} 73. Kd5 {+1000.05/58 +43} Kb7 {-1000.04/99 +22} 74. Bc6+
{+1000.04/127 +44} Kb8 {-1000.03/99 +25} 75. Kc5 {+1000.03/127 +47} a6
{-1000.02/99 +28} 76. Kb6 {+1000.02/127 +50} Kc8 {-1000.01/99 +31} 77. Rh8#
{+1000.01/127 +53}
{White mates} 1-0


[Event "Mac Pro x5690 3.46 Ghz 18 CPU"]
[Site "Mac-Pro.local"]
[Date "2017.08.23"]
[Round "-"]
[White "SF-McB 210817 64 POPCNT"]
[Black "Komodo 11.2.2 64-bit"]
[Result "1-0"]
[TimeControl "180+3"]
[FEN "rnbqkbnr/p2ppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1"]
[SetUp "1"]


1. d4 {+2.26/24 +175} Nf6 {-1.74/22 +162} 2. c4 {+2.39/24 +174} Ba6
{-1.90/21 +156} 3. e3 {+2.44/24 +172} e6 {-1.97/23 +143} 4. Nf3
{+2.48/25 +171} Be7 {-2.03/23 +138} 5. Qa4 {+2.53/25 +165} O-O
{-1.92/23 +134} 6. b4 {+2.72/21 +166} Qc7 {-1.96/23 +123} 7. Nbd2
{+2.73/25 +162} Bb7 {-2.18/21 +119} 8. Bd3 {+2.78/24 +157} d6
{-2.38/21 +110} 9. Bb2 {+2.73/24 +152} Nbd7 {-2.41/21 +107} 10. Qc2
{+2.84/23 +151} Rac8 {-2.46/21 +94} 11. O-O {+2.78/23 +144} Rfd8
{-2.48/21 +85} 12. Rfc1 {+3.00/21 +144} e5 {-2.31/20 +83} 13. d5
{+3.04/23 +140} h6 {-2.49/21 +78} 14. e4 {+3.45/22 +140} Nh5 {-2.49/22 +69}
15. a4 {+3.80/24 +123} Qb8 {-2.81/23 +47} 16. Bc3 {+3.87/22 +117} Re8
{-2.96/20 +42} 17. a5 {+4.02/25 +113} Ba6 {-2.57/22 +42} 18. Rab1
{+3.92/28 +107} Qc7 {-2.61/22 +42} 19. b5 {+4.04/25 +101} Bb7
{-3.08/21 +38} 20. Nb3 {+4.01/25 +96} Bf8 {-3.19/22 +32} 21. a6
{+4.37/25 +93} Ba8 {-3.37/24 +30} 22. g3 {+4.35/24 +94} Nhf6 {-3.53/23 +27}
23. Bf1 {+4.37/27 +92} g6 {-3.44/20 +22} 24. Bb4 {+4.37/27 +87} Rb8
{-3.18/21 +21} 25. Qe2 {+4.47/24 +79} h5 {-3.17/22 +18} 26. h4
{+4.57/26 +74} Nc5 {-3.37/23 +16} 27. Nxc5 {+5.33/28 +65} dxc5
{-3.65/23 +16} 28. Bd2 {+5.36/27 +59} Qd7 {-3.80/22 +16} 29. Ng5
{+5.38/27 +54} Qc7 {-3.87/23 +14} 30. Bh3 {+5.43/27 +47} Bd6 {-4.13/24 +10}
31. Rf1 {+6.03/26 +45} Nh7 {-4.10/22 +10} 32. f4 {+6.28/26 +40} exf4
{-4.70/21 +8} 33. gxf4 {+6.32/20 +42} f5 {-4.83/20 +9} 34. Nxh7
{+6.54/26 +42} Qxh7 {-4.32/21 +10} 35. e5 {+6.72/22 +44} Qf7 {-5.04/20 +8}
36. Kh2 {+8.16/22 +42} Rb6 {-5.80/20 +9} 37. Rg1 {+8.37/20 +44} Bc7
{-7.34/20 +5} 38. Qxh5 {+9.49/22 +44} Qg7 {-7.14/18 +6} 39. Qf3
{+10.20/21 +43} Rxe5 {-7.20/19 +7} 40. fxe5 {+11.46/22 +43} Qxe5+
{-8.29/20 +8} 41. Rg3 {+11.96/19 +45} Qd4 {-8.35/21 +3} 42. Qe2
{+12.79/22 +43} Bxg3+ {-9.78/20 +4} 43. Kxg3 {+16.60/24 +41} Bxd5
{-10.77/20 +5} 44. cxd5 {+19.08/22 +40} Qxd5 {-11.06/23 +6} 45. Bg2
{+23.95/22 +40} Qf7 {-13.00/24 +5} 46. Qe5 {+30.82/22 +39} Re6
{-15.18/24 +3} 47. Qxc5 {+65.61/21 +38} Qe7 {-16.17/23 +5} 48. Qxe7
{+132.73/26 +36} Rxe7 {-250.00/24 +6} 49. Bd5+ {+1000.11/26 +37} Kh7
{-23.41/24 +3} 50. b6 {+1000.10/31 +37} f4+ {-250.00/23 +6} 51. Kg4
{+1000.08/34 +37} axb6 {-1000.07/19 +8} 52. Rxb6 {+1000.07/33 +40} Kh6
{-1000.06/22 +11} 53. Rb7 {+1000.06/38 +39} Rxb7 {-1000.05/32 +14} 54. axb7
{+1000.05/42 +42} Kg7 {-1000.04/99 +17} 55. b8=Q {+1000.04/66 +42} Kf6
{-1000.03/99 +19} 56. Bxf4 {+1000.03/127 +45} g5 {-1000.02/99 +22} 57. Qf8+
{+1000.02/127 +48} Kg6 {-1000.01/5 +25} 58. Be4# {+1000.01/127 +51}
{White mates} 1-0[/pgn]