What on earth is wrong with Thinker in this game?

Discussion of computer chess matches and engine tournaments.

Moderator: Ras

Martin Thoresen
Posts: 1833
Joined: Thu Jun 22, 2006 12:07 am

What on earth is wrong with Thinker in this game?

Post by Martin Thoresen »

I was watching this game being played in my current tournament.

Thinker 5.4D Inert 64-bit 2CPU vs Scorpio 2.4 JA 64-bit 2CPU.

Thinker is one of my favorite engines, but in this game it blundered
really bad twice. The game evolved slowly in a semi-dry position
after Thinker had sacrificed a Rook for a Knight and a Bishop.

First blunder was making a won position about equal, then second
blunder was straight up losing the game. The two blunders came right
after each other. Never seen anything like this in my games. Well, I
haven't checked them all, of course. But still. Major blunders like this
was not expected from a strong engine like Thinker.

Position after 53... Qe8:
[d]
Thinker here chose 54. Nf7+ ??

Then Scorpio rightfully says "why, thank you" and takes the Knight with the Rook and we get this position after 54... Rxf7:
[d]
Then Thinker strikes again and plays 55. Bxg7+ ?? and Scorpio says "that's very generous of you, yet again" and captures the Bishop with his Rook.

Anyone seen anything like this from Thinker? I wonder if I have somehow managed to misconfigure this engine or something, but it runs fine with proper hash size and proper CPU-utilization.

Final position:
[d]

Full game (engine evals are from own point of view):

Code: Select all

[Event "Tournament 002"]
[Site "Martin Thoresen's Chess Computer"]
[Date "2010.02.05"]
[Round "1"]
[White "Thinker 5.4D Inert 64-bit 2CPU"]
[Black "Scorpio 2.4 JA 64-bit 2CPU"]
[Result "0-1"]
[ECO "C53"]
[Opening "Giuoco Piano"]
[Time "03:21:45"]
[Variation "LaBourdonnais Variation"]
[TimeControl "240+2"]
[PlyCount "145"]

1. e4 e5 2. Nf3 Nc6 3. Bc4 h6 4. O-O Nf6 5. Re1 Bc5 6. c3 d6 7. d4 Bb6 8.
a4 exd4 9. a5 Nxa5 {-0.87/16 18} 10. Rxa5 Bxa5 {-0.87/15 8} 11. e5 O-O
{-0.68/15 8} 12. exf6 Qxf6 {-0.53/15 8} 13. Nxd4 Bb6 {-0.47/15 8} 14. Nd2
Bxd4 {-0.30/16 8} 15. cxd4 Qxd4 {-0.32/16 7} 16. Qb3 Bf5 {-0.20/16 7} 17.
Nf1 Be6 {-0.05/15 6} 18. Bxe6 fxe6 {0.00/14 1} 19. Qxe6+ Rf7 {-0.04/17 7}
20. Re2 Qf6 {+0.06/15 6} 21. Qb3 c6 {+0.10/15 7} 22. h3 Qg6 {0.00/16 7} 23.
Bd2 Kh8 {-0.15/16 6} 24. Bc3 Raf8 {0.00/15 6} 25. Qa4 b5 {0.00/15 6} 26.
Qa6 Qh5 {0.00/16 6} 27. Ng3 Qc5 {0.00/17 6} 28. Ne4 Qc4 {-0.02/15 6} 29.
Re3 Qd5 {-0.07/15 6} 30. b4 Kh7 {-0.27/14 5} 31. Qa1 Kg8 {-0.36/13 5} 32.
Qf1 Qa2 {-0.37/15 5} 33. Qe1 Rd7 {-0.48/15 5} 34. Qd1 d5 {-0.78/15 5} 35.
Qg4 Rdf7 {-1.33/16 13} 36. Qe6 Kh7 {-1.32/16 5} 37. Kh2 Qc2 {-1.91/17 7}
38. Qd6 Kg8 {-0.30/14 5} 39. Nd2 h5 {-0.56/14 4} 40. f3 Rc8 {-0.37/14 8}
41. h4 a6 {-0.30/14 4} 42. f4 Qf5 {-0.60/16 4} 43. Be5 Rd7 {-0.97/16 4} 44.
Qc5 Qf8 {-1.15/16 4} 45. Qc2 c5 {-1.55/14 4} 46. bxc5 Rxc5 {-1.82/14 3} 47.
Qg6 Re7 {-2.30/15 9} 48. Rg3 Qe8 {-2.89/15 6} 49. Qd6 Rc8 {-3.35/14 3} 50.
Qxd5+ Kh8 {-4.09/14 3} 51. Ne4 Rd8 {-5.59/15 6} 52. Nd6 Qg8 {-6.49/15 3}
53. Qd1 Qe8 {-6.65/15 3} 54. Nf7+ Rxf7 {+0.90/13 0} 55. Bxg7+ Rxg7
{+9.01/15 3} 56. Qa1 Qf7 {+9.07/15 3} 57. Qe5 Qe7 {+9.35/16 3} 58. Qxe7
Rxe7 {+9.54/15 0} 59. Rg5 Rg8 {+10.00/16 3} 60. Rxh5+ Rh7 {+11.01/17 3} 61.
Rd5 Rxh4+ {+11.15/14 0} 62. Kg1 Rxf4 {+11.36/15 3} 63. Rd6 a5 {+11.68/14 3}
64. Rd5 Rb4 {+11.83/13 3} 65. Rd3 a4 {+12.25/15 2} 66. Rh3+ Kg7 {+12.47/15
0} 67. Rg3+ Kf7 {+12.74/15 2} 68. Rf3+ Ke6 {+12.96/15 3} 69. Rh3 Rb2
{+14.46/14 3} 70. Re3+ Kf6 {+19.33/15 3} 71. Rf3+ Ke5 {+20.57/16 3} 72. g3
Rb3 {+20.77/15 3} 73. Rf7 ... {White resigns} 0-1
Martin Thoresen
Posts: 1833
Joined: Thu Jun 22, 2006 12:07 am

Re: What on earth is wrong with Thinker in this game?

Post by Martin Thoresen »

If anyone cares: it seems some folks over at the Rybka forum had the same experience with 5.4D. I am going back to 5.4C for now, until a new version is released.