Two Sudden Turnarounds In Evaluation

[brinan]

[Event "Ladder Match 41 12g max Feb 2011"]
[Site "Intel Core i3-330M"]
[Date "2011.02.19"]
[Round "5"]
[White "IvanHoe 47"]
[Black "Deep Rybka 4"]
[Result "1-0"]
[WhiteElo "3244"]
[BlackElo "3202"]
[TimeControl "5400"]
[SetUp "1"]
[FEN "8/8/5p2/5kb1/P1R1p2p/8/1P6/3K4 w - - 0 38"]

Turnarounds
1. After White's 39th move.
2. After Black's 46th.
[SetUp "1"]
[FEN "8/8/5p2/5kb1/P1R1p2p/8/1P6/3K4 w - - 0 38"]

38. Ke2 {(38.Ke2 h3 39.Rd4 Ke5 40.Rd1 f5 41.Rh1 f4 42.Rxh3 f3+ 43.Kf2 Kf4
44.Rh8 e3+ 45.Kf1 Kg3 46.Rg8 e2+ 47.Ke1 Kf4 48.Rf8+ Ke3 49.Re8+ Kd3 50.Kf2
Bh4+ 51.Kxf3 e1Q 52.Rxe1 Bxe1) 0.00/25 83} h3 {(38. ... h3 39.Kf1 h2 40.Kg2
e3 41.Rc1 Ke4 42.a5 e2 43.Re1 Kd3 44.a6 Be3 45.Kxh2 Kd2 46.Rh1 e1Q 47.Rxe1
Kxe1 48.Kg3 Ke2 49.Kg4 Kd3 50.Kf5 Bd4 51.a7 Bxa7 52.Kxf6 Bd4+ 53.Kf5 Kc4)
0.00/19 40} 39. Rd4 {(39.Rd4 Ke5 40.Rd1 f5 41.Rh1 f4 42.Rxh3 f3+ 43.Kf2 Kf4
44.Rh8 e3+ 45.Kf1 Kg3 46.Rg8 e2+ 47.Ke1 Kf4 48.Rf8+ Ke3 49.Re8+ Kd3 50.Kf2
Bh4+ 51.Kxf3 e1Q 52.Rxe1 Bxe1) 0.00/24 0} h2 {(39. ... h2 40.Rd1 Kg4 41.a5
Bf4 42.b4 Bc7 43.Ke3 f5 44.Rb1 Kg3 45.Kd4 Kg2 46.Rb2+ Kh3 47.Rb1 Bg3 48.Kc4
Bf2 49.b5 e3 50.b6 e2 51.b7 e1Q 52.Rxe1 Bxe1 53.b8Q h1Q 54.Qh8+ Bh4)
+3.26/17 95} 40. Rd1 {(40.Rd1 Kg4 41.a5 Bf4 42.Rb1 f5 43.b4 Bc7 44.Ke3 Kg3
45.Kd4 Kg2 46.Rb2+ Kh3 47.Rb1 Bg3 48.Kc4 Bf2 49.b5 e3 50.b6 e2 51.b7 e1Q
52.Rxe1 Bxe1 53.b8Q h1Q 54.Qh8+ Bh4 55.Kc5 f4 56.a6 Qg1+ 57.Kb5 f3 58.Qc8+
Kg2 59.Qg8+ Bg3 60.Qa8 Qb1+ 61.Kc4 Qc2+ 62.Kd4 Bf2+ 63.Ke5 Qc5+ 64.Ke4 Qc4+
65.Kf5 Be3 66.Qb7 Qc5+ 67.Ke4 Qd4+ 68.Kf5 Qf4+ 69.Ke6 Bc5) -3.39/24 443}
Kg4 {(40. ... Kg4) +3.38/23 0} 41. a5 {(41.a5 Bf4 42.b4 Bc7 43.Ke3 f5
44.Ra1 Kg3 45.Kd4 Kg2 46.Ra2+ Kf3 47.Ra3+ Kf4 48.Rh3 Be5+ 49.Kc5 Kg4 50.Rh7
e3 51.a6 Bb8 52.Rxh2 Bxh2 53.a7 e2 54.a8Q Bg1+ 55.Kb5 e1Q 56.Qg2+ Kf4
57.Qc6 Be3 58.Ka4 Qd1+ 59.Ka5 Qd3 60.Ka4 Qd4 61.Qh1 Ke5 62.Qh8+ Ke6 63.Qe8+
Kd5 64.Qf7+ Ke5 65.Qc7+ Qd6 66.Qg7+ Kf4 67.Qc3 Ke4) -3.35/24 118} Bf4 {(41.
... Bf4 42.b4 Bc7) +3.53/20 0} 42. b4 {(42.b4 Bc7 43.Ke3 f5 44.Rb1 Kg3
45.Kd4 Kf2 46.Rb2+ Kf3 47.Rb3+ Kg2 48.Rb2+ Kh3 49.Rb1 Bg3 50.Kc4 Bf2 51.b5
e3 52.b6 e2 53.b7 e1Q 54.Rxe1 Bxe1 55.b8Q h1Q 56.Qh8+ Bh4 57.Kc5 f4 58.a6
Qg1+ 59.Kb5 f3 60.Qc8+ Kg2 61.Qg8+ Bg3 62.Qa8 Qb1+ 63.Kc4 Qc2+ 64.Kd4 Bf2+
65.Ke5 Qc5+ 66.Ke4 Qc4+ 67.Kf5 Be3 68.Qb7 Qc5+ 69.Ke4 Qd4+ 70.Kf5 Qf4+
71.Ke6 Bc5) -3.43/24 108} Bc7 {(42. ... Bc7 43.Ke3 f5 44.Ra1 Bf4+ 45.Kf2
Be5 46.Rb1 f4 47.b5 Kh3 48.Ke2 Bd4 49.b6 Kg2 50.b7 f3+ 51.Kd2 Ba7 52.Kc2
f2) +3.89/21 0} 43. Ke3 {(43.Ke3 f5 44.Rb1 Kg3 45.Kd4 Kf2 46.Rb2+ Kg1
47.Rb1+ Kg2 48.Rb2+ Kf3 49.Rb3+ Kg2 50.Rb2+) -3.71/25 276} f5 {(43. ... f5
44.Rb1 Kg3 45.Kd4 f4 46.Kxe4 f3 47.Ke3 Bf4+ 48.Kd4 Kg2) +5.89/20 0} 44. Rb1
{(44.Rb1 Kg3 45.Kd4 f4 46.Kc5 f3 47.a6 Bf4 48.Kd5 Be3 49.b5 Bg1 50.Kc4 Kh3
51.a7 Bxa7 52.Kd5 Kg3 53.Rh1 Kg2) -4.56/24 207} Kg3 {(44. ... Kg3 45.Kd4 f4
46.Kc5 f3 47.a6 Bf4 48.Kd4 Bb8 49.b5 Ba7+ 50.Kc4 Bg1 51.Ra1 Ba7 52.Kc3 e3
53.Kd3 e2 54.Kd2 Bg1) +5.76/15} 45. Kd4 {(45.Kd4 f4 46.Kc5 f3 47.a6 Bf4
48.Kd4 Bb8 49.Rc1 Ba7+ 50.Kd5 Be3 51.a7 Bxc1 52.a8Q h1Q 53.Qb8+ Bf4 54.Qg8+
Kf2 55.Qf7 Bg5 56.Kxe4 Qb1+ 57.Kd4 Qxb4+ 58.Kd3) -7.01/18 155} f4 {(45. ...
f4 46.Kc5 f3) +5.76/14 0} 46. Kxe4 {(46.Kxe4 f3 47.Ke3 Bf4+ 48.Kd4 Bb8
49.Kd3 Ba7 50.Re1 Bg1 51.Re7 h1Q 52.Rg7+ Kf2 53.Rc7 Qh5 54.Kc2 Qh6 55.Kb2
Ke2 56.Rc2+ Ke3 57.Rc3+ Ke4 58.Kc2 f2 59.Rc4+ Kf3 60.Rc3+ Ke2) -8.27/18
116} Bb8 {(46. ... Bb8 47.Rh1 Kg2 48.Rxh2+ Kxh2 49.Kf3 Kh3 50.a6 Kh4 51.b5
Bd6 52.b6 Kg5 53.a7 Kf6 54.a8Q Ke6 55.b7 Ke5 56.b8Q Bxb8 57.Qxb8+ Kd4
58.Qb4+ Kd3 59.Qe4+ Kd2 60.Qd4+ Kc2 61.Kxf4) -10.71/13 224} 47. Rh1
{(47.Rh1 Ba7 48.Rxh2 Bg1 49.Rh8 Ba7 50.Rg8+ Kf2 51.Kxf4 Be3+ 52.Kf5 Ba7
53.a6 Ke3 54.b5 Ke2 55.b6 Bxb6) +190.69/16 14} Ba7 {(47. ... Ba7 48.Rxh2
Be3 49.Rh6 Kf2 50.a6 Ke2 51.Rf6 Kd1 52.a7 Bxa7 53.Rxf4 Kc2 54.b5 Kc3 55.Rf6
Bb8) -M23/15 26} 48. Rxh2 {(48.Rxh2 f3 49.a6 Kg4 50.Ra2 Kg3 51.Ra3 Kf2
52.Rxf3+ Ke2 53.Rf7 Be3 54.a7 Bxa7) +197.09/22 0} Be3 {(48. ... Be3 49.Rh7
Kg4 50.a6 Kg5 51.a7 Kg6 52.Rh1 Bxa7 53.Kxf4 Kf6 54.Rh6+ Kg7 55.Rc6 Bb8+
56.Ke4 Kf7 57.b5 Ke7) -M24/13 19} 49. Rh7 {(49.Rh7 Kg4 50.a6 Kg5 51.a7 Bxa7
52.Rxa7 Kg6) +203.74/21 10} Kg4 {(49. ... Kg4 50.a6 Kg5 51.a7 Bxa7 52.Rxa7
Kg6) -M12/17 2} 50. a6 {(50.a6 Kg5 51.a7 Bxa7 52.Rxa7 Kg6) +203.74/21 1}
Kg5 {(50. ... Kg5 51.a7 Bxa7 52.Rxa7 Kg6 53.b5 Kf6 54.b6 Ke6 55.b7 Kf6
56.b8Q Kg6) -M11/17 13} 51. a7 {(51.a7 Bxa7 52.Rxa7 Kg6) +207.58/22 0} Bxa7
{(51. ... Bxa7 52.Rxa7 Kg6) -M10/20 13} 52. Rxa7 {(52.Rxa7 Kg6) +208.86/26}
Kf6 {(52. ... Kf6 53.b5 Ke6) -M9/59 0} 53. Kxf4 Ke6 {(53. ... Ke6 54.b5
Kd6) -M9/59 0} 54. b5 Kd6 {(54. ... Kd6) -M8/59 0} 55. b6 Kc6 {(55. ...
Kc6) -M7/59 0} 56. b7 Kc7 {(56. ... Kc7 57.Ke5) -M6/59 0} 57. b8=Q+ Kxb8
{(57. ... Kxb8) -M7/59 0} 58. Re7 Ka8 {(58. ... Ka8) -M6/59 0} 59. Rf7 Kb8
{-M5/1 0} 60. Ke5 Ka8 {(60. ... Ka8) -M4/59 0} 61. Kd5 Kb8 {(61. ... Kb8)
-M3/30 0} 62. Kc6 Ka8 {(62. ... Ka8) -M2/59 0} 63. Kb6 Kb8 {(63. ... Kb8)
-M1/17 0} 64. Rf8# 1-0

[BB+]

Interesting. Passed pawns are tricky,especially with races. Did either have tablebases? From the scores for IvanHoe, it looks like RobboBases were being used.

[brinan]

[...] Did either have tablebases? [...]

Both used tablebases.

After the game, got DR4 to analyse position prior to Bb8 and it immediately found f3 (as did Houdini, IvanHoe and Critter).

In the game, DR4 had 15' 14" on the clock and took 3' 44". Furthermore, its own evaluation of Bb8 was that of a losing move.

Any programming theory as to what might have gone wrong?

[BB+]

Hmm.. maybe Bb8 looked OK at a low depth, then starting failing low, Rybka spent more and more time looking at it to determine how bad it was, and time ran out? Maybe not the best use of time, but it's what could happen.

Rybka also blitzed many of the moves in this game snippet:

40... Kg4 (40. ... Kg4) +3.38/23 0  
41... Bf4 (41. ... Bf4 42.b4 Bc7) +3.53/20 0 
42... Bc7 (42. ... Bc7 43.Ke3 f5 44.Ra1 Bf4+ 45.Kf2 Be5 46.Rb1 f4 47.b5 Kh3 48.Ke2 Bd4 49.b6 Kg2 50.b7 f3+ 51.Kd2 Ba7 52.Kc2 f2) +3.89/21 0 
43... f5 (43. ... f5 44.Rb1 Kg3 45.Kd4 f4 46.Kxe4 f3 47.Ke3 Bf4+ 48.Kd4 Kg2) +5.89/20 0  
44... Kg3 (44. ... Kg3 45.Kd4 f4 46.Kc5 f3 47.a6 Bf4 48.Kd4 Bb8 49.b5 Ba7+ 50.Kc4 Bg1 51.Ra1 Ba7 52.Kc3 e3 53.Kd3 e2 54.Kd2 Bg1) +5.76/15  [no time token?]
45... f4 (45. ... f4 46.Kc5 f3) +5.76/14 0  
46... Bb8 (46. ... Bb8 47.Rh1 Kg2 48.Rxh2+ Kxh2 49.Kf3 Kh3 50.a6 Kh4 51.b5 Bd6 52.b6 Kg5 53.a7 Kf6 54.a8Q Ke6 55.b7 Ke5 56.b8Q Bxb8 57.Qxb8+ Kd4 58.Qb4+ Kd3 59.Qe4+ Kd2 60.Qd4+ Kc2 61.Kxf4) -10.71/13 224

[BB+]

It might also be that elusive "stall" bug that R4 is purported to have, preventing it from changing to a different move in this case.