Komodo strikes first at the TCEC Superfinal

[Uri Blass]

Here is a clip analysis of 5 best lines with K9. There clearly is something wrong with SF and my guess is that there must be a hash bug somewhere.

97: Komodo 9.01 64-bit - Spearfish, Blitz 2m
[d]4n1k1/3q4/p1p1rr2/PpRp1p1p/1P1P1Pp1/2R1P1PP/2Q2B2/6K1 w - - 0 1

Analysis by Komodo 9.01 64-bit:

1. = (-0.03): 38.h4 Nd6 39.Rxc6 Nc4 40.Rc5 Re4 41.Qe2 Rc6 42.Qc2 Rce6 43.Qb1 Kg7 44.Qd3 Re8 45.Qc2 R8e7 46.Qe2 Kh7 47.Qe1 Qe6 48.Rd3 Re8 49.Rc7+ Kg6 50.Qc1 Rc8 51.Rc5 Kf7 52.Rc3 Kg8 53.Qd1 Kg7 54.Qc1 Re8 55.Rc7+ Re7 56.Rc5 Qd7
2. -/+ (-1.40): 38.e4 fxe4 39.h4 Nd6 40.Qc1 Nf5 41.Kg2 Re8 42.Be3 Rc8 43.Qe1 Re6 44.Kg1 Nd6 45.Rc1 Rf6 46.R1c2 Nc4 47.Qc1 Rd6 48.Qe1 Rf8 49.Qe2 Qg7 50.Rc1 Rdf6 51.Rc3 Qd7 52.Rc1 Re8 53.Qe1 Qc7 54.Rc2 Rd6
3. -/+ (-1.50): 38.Kg2 gxh3+ 39.Kxh3 h4 40.gxh4 Rg6 41.Qc1 Rd6 42.Rc2 Qe6 43.Bg3 Qe4 44.Qf1 Ng7 45.R5c3 Nh5 46.Be1 Rde6 47.Rg2 Qxg2+ 48.Qxg2 Rxe3+ 49.Qf3 Nxf4+ 50.Kh2 Rxf3 51.Rxf3 Rg2+ 52.Kh1 Rg4 53.Bf2 Kf7 54.h5 Nxh5 55.Rxf5+ Kg6 56.Rf3 Nf4 57.Rc3 Kf5 58.Rf3 Ke6 59.Kh2 Ne2 60.Re3+ Re4 61.Rd3 Kf6 62.Be3 Rh4+ 63.Kg2 Ke6 64.Rd2 Nc3
4. -/+ (-1.52): 38.Kh2 gxh3 39.Kxh3 h4 40.gxh4 Rg6 41.Qc1 Rd6 42.Rc2 Qe6 43.Bg3 Qe4 44.Qf1 Ng7 45.R5c3 Nh5 46.Be1 Rde6 47.Rg2 Qxg2+ 48.Qxg2 Rxe3+ 49.Qf3 Nxf4+ 50.Kh2 Rxf3 51.Rxf3 Rg2+ 52.Kh1 Rg4 53.Bf2 Kf7 54.h5 Nxh5 55.Rxf5+ Kg6 56.Rf3 Nf4 57.Rc3 Kf5 58.Rf3 Ke6 59.Kh2 Ne2 60.Re3+ Re4 61.Rd3 Kf6 62.Rf3+ Ke7 63.Rd3 Ke6 64.Rd2 Kf5
5. -+ (-1.86): 38.hxg4 hxg4 39.Kf1 Nd6 40.Rxc6 Nc4 41.Rxe6 Qxe6 42.Rb3 Rh6 43.Bg1 Qc8 44.Rc3 Kg7 45.Qg2 Rc6 46.Ke2 Nd6 47.Rc5 Ne4 48.Bf2 Nxc5 49.bxc5 Qe6 50.Be1 Rc8 51.Bd2 Rb8 52.Kd3 b4 53.Kc2 b3+ 54.Kb2 Qc6 55.Qf1 Qa4 56.Qa1 Qxa1+ 57.Kxa1 Kf6 58.c6 Ke6 59.c7 Rc8 60.Kb2 Rxc7 61.Kxb3 Rc8 62.Bb4 Rc1 63.Ba3 Rb1+ 64.Kc3

(Spearfish, 24.11.2015)

Note that Komodo score was 1.30 pawns advantage for black when it played 37...g4 and expected 38.h4 so I guess Komodo also saw something.

[shrapnel]

Hmm... I wonder if there is something significant to the fact that nearly all Komodo wins have been with Black ?
I wonder if using Contempt 0 really suits Komodo especially playing with Black ?
Stockfish seems to play the same, Black or White, while using Contempt 0.
I don't know, but maybe Contempt 0 has different implications for Stockfish and Komodo ?
Komodo with Contempt 0, as Black, seems to sit back and wait calmly for Stockfish to make the slightest mistake....and THEN it pounces like a cat on a mouse !

[MikeB]

Interesting game going on - although K's analysis look drawish, SF pv is more favorable for white (Komodo). On the iPhone and wasn't apparent how to copy game or position from the site.

[AdminX]

Interesting game going on - although K's analysis look drawish, SF pv is more favorable for white (Komodo). On the iPhone and wasn't apparent how to copy game or position from the site.

[d]1r3k2/4qp1p/2np2p1/p1p5/P3N3/2P4P/3Q1PP1/4R1K1 w - - 6 30 29

[M ANSARI]

Komodo was completely winning and game would have lasted a couple of moves more as both engines were showing around +6.0 scores and both evaluation boards showed 100% win ... then somehow the game was restarted due to a server crash and the game was cancelled. I have seen some ridiculous rules but sorry to say that has to trump anything I have ever seen. Hopefully that game is saved somewhere.

I managed to get the game from chessbomb not sure if it is all the way to the end. I actually had spent a few hours watching the game and was pretty pissed off to have to see something like that.

[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[PlyCount "120"]

1. d4 Nf6 2. c4 c5 3. d5 b5 4. Nf3 g6 5. cxb5 a6 6. b6 Qxb6 7. Nc3 d6 8. e4
Nbd7 9. Be2 Bg7 10. O-O O-O 11. Rb1 Bb7 12. Re1 Rfb8 13. Bf4 a5 14. Bb5 Qd8 15.
h3 Nb6 16. a4 Ne8 17. e5 Nc7 18. exd6 exd6 19. Bc6 Bxc6 20. dxc6 Bxc3 21. bxc3
Ncd5 22. Qd2 Nxf4 23. Qxf4 Nd5 24. Qd2 Ne7 25. Rxb8 Rxb8 26. Ng5 Nxc6 27. Qd5
Qc7 28. Ne4 Kf8 29. Qd2 Qe7 30. Nxd6 Qd7 31. Kh2 Kg8 32. Qf4 Rd8 33. Ne4 Qf5
34. Nf6+ Kh8 35. Qxf5 gxf5 36. Nh5 c4 37. f3 Kg8 38. Ng3 Kf8 39. Nxf5 Rd3 40.
Re4 Rxc3 41. Nd6 Kg7 42. Rg4+ Kf8 43. Nxc4 f5 44. Rf4 Ke7 45. g4 fxg4 46. hxg4
Ke6 47. Re4+ Kf6 48. Kg3 h6 49. Ne3 Rb3 50. Rc4 Ne7 51. Kf4 Ke6 52. Ke4 Kd6 53.
f4 Rb8 54. f5 Nc6 55. Kf4 Rb2 56. Re4 Kc5 57. Kg3 Rb4 58. Re8 Rb8 59. Re6 Nd8
60. Rxh6 Rb3 *

[Graham Banks]

Komodo was completely winning and game would have lasted a couple of moves more as both engines were showing around +6.0 scores and both evaluation boards showed 100% win ... then somehow the game was restarted due to a server crash and the game was cancelled. I have seen some ridiculous rules but sorry to say that has to trump anything I have ever seen. Hopefully that game is saved somewhere.

I managed to get the game from chessbomb not sure if it is all the way to the end. I actually had spent a few hours watching the game and was pretty pissed off to have to see something like that.

[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[PlyCount "120"]

1. d4 Nf6 2. c4 c5 3. d5 b5 4. Nf3 g6 5. cxb5 a6 6. b6 Qxb6 7. Nc3 d6 8. e4
Nbd7 9. Be2 Bg7 10. O-O O-O 11. Rb1 Bb7 12. Re1 Rfb8 13. Bf4 a5 14. Bb5 Qd8 15.
h3 Nb6 16. a4 Ne8 17. e5 Nc7 18. exd6 exd6 19. Bc6 Bxc6 20. dxc6 Bxc3 21. bxc3
Ncd5 22. Qd2 Nxf4 23. Qxf4 Nd5 24. Qd2 Ne7 25. Rxb8 Rxb8 26. Ng5 Nxc6 27. Qd5
Qc7 28. Ne4 Kf8 29. Qd2 Qe7 30. Nxd6 Qd7 31. Kh2 Kg8 32. Qf4 Rd8 33. Ne4 Qf5
34. Nf6+ Kh8 35. Qxf5 gxf5 36. Nh5 c4 37. f3 Kg8 38. Ng3 Kf8 39. Nxf5 Rd3 40.
Re4 Rxc3 41. Nd6 Kg7 42. Rg4+ Kf8 43. Nxc4 f5 44. Rf4 Ke7 45. g4 fxg4 46. hxg4
Ke6 47. Re4+ Kf6 48. Kg3 h6 49. Ne3 Rb3 50. Rc4 Ne7 51. Kf4 Ke6 52. Ke4 Kd6 53.
f4 Rb8 54. f5 Nc6 55. Kf4 Rb2 56. Re4 Kc5 57. Kg3 Rb4 58. Re8 Rb8 59. Re6 Nd8
60. Rxh6 Rb3 *

Should really count as a Komodo win, but TCEC has its own rules as you say.

[Carlos777]

Komodo was completely winning and game would have lasted a couple of moves more as both engines were showing around +6.0 scores and both evaluation boards showed 100% win ... then somehow the game was restarted due to a server crash and the game was cancelled. I have seen some ridiculous rules but sorry to say that has to trump anything I have ever seen. Hopefully that game is saved somewhere.

I managed to get the game from chessbomb not sure if it is all the way to the end. I actually had spent a few hours watching the game and was pretty pissed off to have to see something like that.

[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[PlyCount "120"]

1. d4 Nf6 2. c4 c5 3. d5 b5 4. Nf3 g6 5. cxb5 a6 6. b6 Qxb6 7. Nc3 d6 8. e4
Nbd7 9. Be2 Bg7 10. O-O O-O 11. Rb1 Bb7 12. Re1 Rfb8 13. Bf4 a5 14. Bb5 Qd8 15.
h3 Nb6 16. a4 Ne8 17. e5 Nc7 18. exd6 exd6 19. Bc6 Bxc6 20. dxc6 Bxc3 21. bxc3
Ncd5 22. Qd2 Nxf4 23. Qxf4 Nd5 24. Qd2 Ne7 25. Rxb8 Rxb8 26. Ng5 Nxc6 27. Qd5
Qc7 28. Ne4 Kf8 29. Qd2 Qe7 30. Nxd6 Qd7 31. Kh2 Kg8 32. Qf4 Rd8 33. Ne4 Qf5
34. Nf6+ Kh8 35. Qxf5 gxf5 36. Nh5 c4 37. f3 Kg8 38. Ng3 Kf8 39. Nxf5 Rd3 40.
Re4 Rxc3 41. Nd6 Kg7 42. Rg4+ Kf8 43. Nxc4 f5 44. Rf4 Ke7 45. g4 fxg4 46. hxg4
Ke6 47. Re4+ Kf6 48. Kg3 h6 49. Ne3 Rb3 50. Rc4 Ne7 51. Kf4 Ke6 52. Ke4 Kd6 53.
f4 Rb8 54. f5 Nc6 55. Kf4 Rb2 56. Re4 Kc5 57. Kg3 Rb4 58. Re8 Rb8 59. Re6 Nd8
60. Rxh6 Rb3 *

Komodo has played 10. Qc2 instead of O-O, so a different game now.

In this case, I would have adjudicated the game to Komodo or at least, let them finish the game from the last position before the crash.

[M ANSARI]

Actually it was a very interesting game with Komodo managing to win a pawn in the middle game and then SF putting incredible resistance in making it almost impossible to convert the win. Then Komodo seemed to have SF tied up trying to protect a passed pawn and all of a sudden Komodo had a very active King and SF had to jettison one more pawn and it was game over. Two connected passed pawns are pretty easy to convert and both evaluation boards showed totally winning lines.

Wanted to see if the replay of the game would be similar but already few moves from the opening Komodo has played 10.Qc2 rather than castle in the original game.

[MikeB]

Actually it was a very interesting game with Komodo managing to win a pawn in the middle game and then SF putting incredible resistance in making it almost impossible to convert the win. Then Komodo seemed to have SF tied up trying to protect a passed pawn and all of a sudden Komodo had a very active King and SF had to jettison one more pawn and it was game over. Two connected passed pawns are pretty easy to convert and both evaluation boards showed totally winning lines.

Wanted to see if the replay of the game would be similar but already few moves from the opening Komodo has played 10.Qc2 rather than castle in the original game.

The sad part is that the game now lost ( being replayed) was one of the best 5 games in the whole match - now gone forever - sort of. Will not be on the official record.

[Leto]

Komodo's eval jumped 1.54 with 42.Kf1 in the replayed game #83. Komodo might just win again, but if Stockfish wins with the reverse color it might just mean that the opening is busted for black.