Higher than expected by me efficiency of Ponder ON

[Laskos]

Since both engines were the same, and "think" in the same way, is it not likely that the move that is guessed during pondering is more likely to be correct than when different engines play each other?

It's my opinion too : the pondering engine guess the played move very often with the same engine.

By the way, I remember you have a strong machine. If you have 24 cores, you can set 16 cores Stockfish Ponder OFF against 8 core Stockfish Ponder ON. If you have 32 cores, you can put 20 core Ponder OFF against 10 core Ponder ON. My prediction is that Ponder ON half the cores will win in these cases. You can set time control 10''+0.1'' and play say 400 games, that will take 3-4 hours.

[Laskos]

Since both engines were the same, and "think" in the same way, is it not likely that the move that is guessed during pondering is more likely to be correct than when different engines play each other?

It's my opinion too : the pondering engine guess the played move very often with the same engine.

By the way, I remember you have a strong machine. If you have 24 cores, you can set 16 cores Stockfish Ponder OFF against 8 core Stockfish Ponder ON. If you have 32 cores, you can put 20 core Ponder OFF against 10 core Ponder ON. My prediction is that Ponder ON half the cores will win in these cases. You can set time control 10''+0.1'' and play say 400 games, that will take 3-4 hours.

Here is my own test: 1 threaded Komodo against Komodo Ponder ON, 2 threaded Komodo and Komodo double time:

Code: Select all

5s+0.05s
Rank Name                               ELO     +/-   Games   Score   Draws
   1 Komodo double time                  65      18     600     59%     58%
   2 Komodo ponder                       47      19     600     57%     55%
   3 Komodo double threads                5      18     600     51%     56%
   4 Komodo single                     -120      19     600     33%     51%
Finished match

Even in this case, Komodo Ponder ON come ahead of 2 threaded Komodo, probably because at my base short time control of 5+0.05, parallel search is not yet efficient. But it shows the high efficiency of Ponder ON, which comes close to double time.

[Nordlandia]

Interesting experiment Kai!

-----

It is possible to launch two computer match in fritz by using engine cload.

The requirements are fritz gui installed on both computers and active playchess subscription.

Computer A host private engine. Computer B select engine from cload.

"Move Overhead" parameter need to be adjusted to compensate for transmission delay & lag.

Personally i've settled for 200ms, after some back and forth testing.

[Luis Babboni]

Hello!

That 34xx ELO of Stockfish, Houdini and Komodo are with ponder off or on?
The matchs against profesionals, for example last year Komodo vs Nakamura, was with ponder off or on?

Thanks.

[Laskos]

Hello!

That 34xx ELO of Stockfish, Houdini and Komodo are with ponder off or on?
The matchs against profesionals, for example last year Komodo vs Nakamura, was with ponder off or on?

Thanks.

CCRL, CEGT, FGRL are using Ponder OFF. IPON Ponder ON. Ponder gives no more than a doubling in time, so the engine will be 3300-3400 with or without Ponder. Komodo vs Nakamura AFAIK was with Ponder ON, as there was no time handicap. But again, Ponder ON vs Ponder OFF is no more than a doubling, the question is what parallelization to use with shared resources.

[Luis Babboni]

Hello!

That 34xx ELO of Stockfish, Houdini and Komodo are with ponder off or on?
The matchs against profesionals, for example last year Komodo vs Nakamura, was with ponder off or on?

Thanks.

CCRL, CEGT, FGRL are using Ponder OFF. IPON Ponder ON. Ponder gives no more than a doubling in time, so the engine will be 3300-3400 with or without Ponder. Komodo vs Nakamura AFAIK was with Ponder ON, as there was no time handicap. But again, Ponder ON vs Ponder OFF is no more than a doubling, the question is what parallelization to use with shared resources.

In real chess (I´m not a good player but I imagine is like this at all levels, or I´m wrong?) use to happens that when one player use a lot of time, the other could use his time to think and, after his move, could move fast again and appears an asimetrical use of time.
Ponder on do not open a place to develope more or less clever time use strategies?!

[Luis Babboni]

...
Ponder on do not open a place to develope more or less clever time use strategies?!

For example an engine could have inside its code other engines code and use it to anticipate the other engine move!!

[Nordlandia]

I don't understand why King captures can ever happen in fritz gui and teleportation. Probably an unidentified bug causes by ponder.
The king was listed as captured piece

[pgn][Event "Lenova-Laptop #2, Rapid 20m+20s"]
[Site "Lenova-Laptop #2"]
[Date "2017.03.09"]
[Round "2"]
[White "Houdini 5.01 Pro x64-pext"]
[Black "Stockfish 090317 64"]
[Result "1-0"]
[Annotator "0.87;0.68"]
[SetUp "1"]
[FEN "1nbqkb1r/r4ppp/p2ppn2/1p4B1/3NPP2/P1N5/1PP3PP/R2QKB1R w KQk - 0 1"]
[PlyCount "307"]
[TimeControl "1200+20"]

1. Ndxb5 {0.87/22 82 Both last book move} axb5 {0.68/26 79} 2. e5 {0.83/24 88}
b4 {0.46/28 22} 3. exf6 {0.82/24 46} gxf6 {0.67/29 0} 4. Qd4 {1.00/24 33} fxg5
{0.75/27 13} 5. Nb5 {0.93/24 68} Ra4 {0.89/29 17} 6. Qxh8 {0.80/24 73} bxa3 {
1.00/31 0} 7. Rxa3 {0.90/23 36} Rxa3 {0.83/29 0} 8. Nxa3 {0.95/24 107 (bxa3)}
gxf4 {0.72/25 30} 9. Qd4 {1.08/22 30} Qa5+ {0.52/27 30} 10. Kd1 {0.73/25 170}
e5 {0.41/32 0} 11. Qe4 {0.69/27 52} Ke7 {0.43/29 59} 12. Bb5 {0.68/27 0} Bg4+ {
0.42/27 17} 13. Kc1 {0.76/23 34} Nd7 {0.66/29 66} 14. Rf1 {0.72/23 6 (Kb1)} Qb6
{0.49/27 31} 15. Kb1 {0.67/24 58 (Qxh7)} h5 {0.47/29 80} 16. Qa8 {0.66/26 46
(g3)} Qb8 {0.38/28 86} 17. Qxb8 {0.61/26 55} Nxb8 {0.37/29 0} 18. Nc4 {0.66/25
34} Ke6 {0.35/29 19} 19. Nd2 {0.50/26 71} d5 {0.47/32 5} 20. Nf3 {0.56/26 49
(c3)} Bc5 {0.32/31 73} 21. c3 {0.50/28 0} Kf6 {0.23/32 82} 22. b4 {0.65/23 23}
Be3 {0.29/32 4} 23. Kb2 {0.56/26 110 (Kc2)} Nd7 {0.31/32 83} 24. Ra1 {0.75/22
24 (h3)} Nb6 {0.20/28 49} 25. Kc2 {0.75/24 41 (Ne1)} h4 {0.45/30 81} 26. h3 {
0.40/25 83} Be6 {0.47/34 41} 27. Ra5 {0.79/24 26} e4 {0.41/35 19} 28. Nd4 {
0.79/26 0} Ke5 {0.41/34 19} 29. Ra6 {0.72/29 39} Nc8 {0.33/36 5} 30. Nxe6 {
0.71/30 60 (Nc6+)} fxe6 {0.33/40 90} 31. Be2 {0.71/32 0} Nb6 {0.33/38 14} 32.
Kb3 {0.71/32 34 (Kd1)} Kd6 {0.33/40 50 (Bf2)} 33. Bg4 {0.71/30 47} Bf2 {
0.33/42 0} 34. Kc2 {0.71/32 59} Ke5 {0.33/40 0} 35. Kd1 {0.71/30 46} Kd6 {
0.33/39 4} 36. Kd2 {0.71/31 32 (Bh5)} Bg1 {0.33/42 62 (Be3+)} 37. Be2 {0.71/31
35 (Ke2)} Be3+ {0.33/42 61} 38. Kd1 {0.71/32 0 (Ke1)} Kc6 {0.33/35 68} 39. Bg4
{0.71/31 10} Kd6 {0.33/37 0} 40. Ke1 {0.71/31 35} e5 {0.33/37 20 (Ke5)} 41. Bh5
{0.63/28 48 (Be2)} Kc6 {0.33/34 63 (Kc7)} 42. Be8+ {0.87/28 105} Kb7 {0.41/35 0
} 43. Ra5 {0.64/28 54} Kc7 {0.70/34 68} 44. Bf7 {0.60/29 0} Kd6 {0.58/31 10}
45. Bh5 {0.58/26 0 (Ke2)} Nc4 {0.33/33 48 (Bc1)} 46. Ra6+ {0.71/29 32} Nb6 {
0.36/34 33} 47. Ke2 {0.71/30 0} Ke7 {0.40/35 72} 48. Kd1 {0.71/31 0} Kd7 {
0.26/35 88 (Kf6)} 49. Be2 {0.66/29 37 (Ke1)} Kd6 {0.29/34 54} 50. Kc2 {0.71/30
22} Kc6 {0.29/36 19} 51. Kb3 {0.55/28 44 (Ra5)} Kd6 {0.26/36 46} 52. Ra2 {
0.46/28 48 (Bg4)} Bf2 {0.26/39 45} 53. Ra5 {0.55/29 0 (Ra6)} Be3 {0.26/39 36}
54. Ra6 {0.46/30 29 (Ra1)} Bf2 {0.26/41 37} 55. Ra2 {0.55/29 0 (Ra5)} Kc6 {
0.26/38 138} 56. c4 {0.51/31 0 (Ra5)} dxc4+ {0.18/29 19} 57. Bxc4 {0.63/29 18}
Bd4 {0.23/33 0} 58. Be2 {0.63/28 6} Kd6 {0.13/31 35 (Nd5)} 59. Ra6 {0.72/26 20}
Ke6 {0.25/33 30} 60. Bg4+ {0.73/26 30 (Bh5)} Kd5 {0.23/33 38} 61. Ra5+ {
0.51/28 62} Kc6 {0.36/36 44 (Kd6)} 62. Bf5 {0.79/30 21 (Ra7)} e3 {0.20/36 34}
63. Bg4 {0.86/31 0 (Bd3)} Kd6 {0.54/34 53} 64. Be2 {0.86/30 0} e4 {0.54/33 9}
65. Ra7 {0.87/24 0} Nd5 {0.58/33 71 (Nd7)} 66. Ra6+ {0.93/29 28} Nb6 {0.76/32 5
} 67. b5 {1.07/28 44 (Kc2)} Kd5 {0.91/31 50} 68. Ra7 {1.07/29 32} Kc5 {0.98/38
0 (Kd6)} 69. Re7 {1.14/29 25 (Rc7+)} Kd5 {1.14/40 25} 70. Rh7 {1.20/29 0} Bf6 {
1.14/39 9} 71. Kc2 {1.32/27 16} Kd4 {1.14/38 4 (Na4)} 72. Rh6 {1.32/31 31 (Rf7)
} Ke5 {1.14/42 34} 73. Kd1 {1.32/32 0} Nd5 {1.14/44 21} 74. Ke1 {1.32/34 16
(Rh5+)} Nb6 {1.14/39 19 (Bd8)} 75. Rh7 {1.32/31 23 (Rh5+)} Nd5 {1.21/40 51} 76.
Ra7 {1.32/34 0 (Rf7)} Bd8 {1.21/39 19 (Be7)} 77. Ra8 {1.32/32 32 (Kf1)} Be7 {
1.21/41 17} 78. Kf1 {1.32/32 27 (Re8)} Kd4 {1.29/41 30} 79. Rc8 {1.28/30 0
(Ra6)} Ke5 {1.36/38 20} 80. Re8 {1.32/30 13 (Rb8)} Kd6 {1.51/33 20} 81. Bd1 {
1.57/26 17} Kd7 {1.30/29 12} 82. Rh8 {1.75/27 28} Kd6 {1.36/34 0} 83. Rh6+ {
1.82/28 36 (Be2)} Kc5 {1.50/35 45} 84. Re6 {1.95/30 0} Kxb5 {2.16/31 23} 85.
Rxe4 {2.15/30 0} Bd6 {2.25/35 20} 86. Be2+ {2.15/29 5} Kb6 {2.48/33 14} 87. Rd4
{2.28/28 18} Kc5 {2.58/35 8} 88. Rc4+ {2.29/25 0} Kb6 {2.66/33 4} 89. Bf3 {
2.29/28 28} Nb4 {2.73/37 21 (Ne7)} 90. Rc8 {2.35/30 42} Be5 {2.58/35 1} 91. Re8
{2.35/30 6} Bd6 {3.02/35 0} 92. Rh8 {2.57/26 20} Be7 {3.02/38 5} 93. Rh7 {
2.73/26 20} Bd8 {3.02/39 0} 94. Rb7+ {2.80/27 43} Kc5 {3.10/40 0} 95. Rf7 {
2.81/29 43} Nc2 {3.10/41 0 (Na2)} 96. Ke2 {2.88/28 47 (Rxf4)} Kd6 {3.55/31 98}
97. Rxf4 {2.95/27 18} Bg5 {3.65/36 31} 98. Rc4 {3.45/24 15 (Rg4)} Na3 {3.65/35
47} 99. Rc6+ {3.68/23 17} Kd7 {3.96/31 30} 100. Rc3 {3.68/25 0 (Rg6)} Nb1 {
3.81/30 19} 101. Rc5 {3.83/25 6} Be7 {3.88/33 26} 102. Rh5 {3.91/24 0} Nd2 {
4.07/31 35} 103. Kxe3 {4.04/24 0} Nf1+ {4.27/27 20} 104. Kd4 {4.08/23 8 (Kf4)}
Ng3 {4.16/26 10} 105. Rh7 {4.30/25 30 (Bg4+)} Ke6 {4.19/22 12} 106. Bg4+ {
4.37/25 11} Kf6 {4.21/22 1} 107. Bd7 {4.44/27 25} Bd8 {4.62/27 27} 108. Rxh4 {
4.58/27 26} Bc7 {5.06/28 21} 109. Rh6+ {4.79/27 20 (Ke3)} Kg5 {6.06/31 46} 110.
Rc6 {5.42/26 27} Bf4 {6.40/29 0} 111. Kd3 {5.94/24 20} Be5 {7.03/32 22 (Bb8)}
112. Rc5 {149.99/64 110 (Ke3)} Kf6 {7.05/28 16 (Kg6)} 113. Rxe5 {149.99/56 57
(Rc6+)} Kxe5 {7.20/32 13 (Nh1)} 114. Ke3 {149.99/1 0 (Ke3)} Kxe3 {-9.39/39 16}
115. Xxe3 {149.99/1 0} Xxe3 {-11.27/33 13} 116. Xxe3 {149.99/1 0} Xxe3 {
-13.98/32 16} 117. Xxe3 {149.99/1 0} Xxe3 {-20.40/33 18} 118. Xxe3 {149.99/1 0}
Xxe3 {-42.57/36 14} 119. Xxe3 {149.99/1 0} Xxe3 {-50.81/33 20} 120. Xxe3 {
149.99/1 0} Xxe3 {-51.81/35 15} 121. Xxe3 {149.99/1 0} Xxe3 {-67.64/40 32} 122.
Xxe3 {149.99/1 0} Xxe3 {-85.93/41 40} 123. Xxe3 {149.99/1 0} Xxe3 {-119.08/41
30} 124. Xxe3 {149.99/1 0} Xxe3 {-119.08/33 26} 125. Xxe3 {149.99/1 0} Xxe3 {
-127.93/34 29} 126. Xxe3 {149.99/1 0} Xxe3 {-116.91/33 24} 127. Xxe3 {149.99/1
0} Xxe3 {-127.98/30 12} 128. Xxe3 {149.99/1 0} Xxe3 {-128.18/35 19} 129. Xxe3 {
149.99/1 0} Xxe3 {-126.87/39 29} 130. Xxe3 {149.99/1 0} Xxe3 {-128.20/40 13}
131. Xxe3 {149.99/1 0} Xxe3 {-128.22/40 24} 132. Xxe3 {149.99/1 0} Xxe3 {
-128.25/46 16} 133. Xxe3 {149.99/1 0} Xxe3 {-128.27/48 26} 134. Xxe3 {149.99/1
0} Xxe3 {-128.27/51 21} 135. Xxe3 {149.99/1 0} Xxe3 {-128.28/53 18} 136. Xxe3 {
149.99/1 0} Xxe3 {-128.28/56 17} 137. Xxe3 {149.99/1 0} Xxe3 {-128.29/56 25}
138. Xxe3 {149.99/1 0} Xxe3 {-128.29/50 20} 139. Xxe3 {149.99/1 0} Xxe3 {
-128.29/51 21} 140. Xxe3 {149.99/1 0} Xxe3 {-128.30/53 17} 141. Xxe3 {149.99/1
0} Xxe3 {-128.29/58 23} 142. Xxe3 {149.99/1 0} Xxe3 {-128.30/49 14} 143. Xxe3 {
149.99/1 0} Xxe3 {-128.30/57 17} 144. Xxe3 {149.99/1 0} Xxe3 {-128.30/59 20}
145. Xxe3 {149.99/1 0} Xxe3 {-128.30/61 30} 146. Xxe3 {149.99/1 0} Xxe3 {
-128.30/62 21} 147. Xxe3 {149.99/1 0} Xxe3 {-128.30/61 17} 148. Xxe3 {149.99/1
0} Xxe3 {-128.30/63 23} 149. Xxe3 {149.99/1 0} Xxe3 {-128.30/65 20} 150. Xxe3 {
149.99/1 0} Xxe3 {-128.30/65 21} 151. Xxe3 {149.99/1 0} Xxe3 {-128.30/60 20}
152. Xxe3 {149.99/1 0} Xxe3 {-128.30/66 17} 153. Xxe3 {149.99/1 0} Xxe3 {
-128.30/66 21} 154. Xxe3 {149.99/1 0 adjud.} 1-0

[/pgn]

Code: Select all

[Event "Lenova-Laptop #2, Rapid 20m+20s"]
[Site "Lenova-Laptop #2"]
[Date "2017.03.09"]
[Round "2"]
[White "Houdini 5.01 Pro x64-pext"]
[Black "Stockfish 090317 64"]
[Result "1-0"]
[SetUp "1"]
[FEN "1nbqkb1r/r4ppp/p2ppn2/1p4B1/3NPP2/P1N5/1PP3PP/R2QKB1R w KQk - 0 1"]
[PlyCount "307"]

1. Ndxb5 axb5 2. e5 b4 3. exf6 gxf6 4. Qd4 fxg5 5. Nb5 Ra4 6. Qxh8 bxa3 7. Rxa3
Rxa3 8. Nxa3 gxf4 9. Qd4 Qa5+ 10. Kd1 e5 11. Qe4 Ke7 12. Bb5 Bg4+ 13. Kc1 Nd7
14. Rf1 Qb6 15. Kb1 h5 16. Qa8 Qb8 17. Qxb8 Nxb8 18. Nc4 Ke6 19. Nd2 d5 20. Nf3
Bc5 21. c3 Kf6 22. b4 Be3 23. Kb2 Nd7 24. Ra1 Nb6 25. Kc2 h4 26. h3 Be6 27. Ra5
e4 28. Nd4 Ke5 29. Ra6 Nc8 30. Nxe6 fxe6 31. Be2 Nb6 32. Kb3 Kd6 33. Bg4 Bf2
34. Kc2 Ke5 35. Kd1 Kd6 36. Kd2 Bg1 37. Be2 Be3+ 38. Kd1 Kc6 39. Bg4 Kd6 40.
Ke1 e5 41. Bh5 Kc6 42. Be8+ Kb7 43. Ra5 Kc7 44. Bf7 Kd6 45. Bh5 Nc4 46. Ra6+
Nb6 47. Ke2 Ke7 48. Kd1 Kd7 49. Be2 Kd6 50. Kc2 Kc6 51. Kb3 Kd6 52. Ra2 Bf2 53.
Ra5 Be3 54. Ra6 Bf2 55. Ra2 Kc6 56. c4 dxc4+ 57. Bxc4 Bd4 58. Be2 Kd6 59. Ra6
Ke6 60. Bg4+ Kd5 61. Ra5+ Kc6 62. Bf5 e3 63. Bg4 Kd6 64. Be2 e4 65. Ra7 Nd5 66.
Ra6+ Nb6 67. b5 Kd5 68. Ra7 Kc5 69. Re7 Kd5 70. Rh7 Bf6 71. Kc2 Kd4 72. Rh6 Ke5
73. Kd1 Nd5 74. Ke1 Nb6 75. Rh7 Nd5 76. Ra7 Bd8 77. Ra8 Be7 78. Kf1 Kd4 79. Rc8
Ke5 80. Re8 Kd6 81. Bd1 Kd7 82. Rh8 Kd6 83. Rh6+ Kc5 84. Re6 Kxb5 85. Rxe4 Bd6
86. Be2+ Kb6 87. Rd4 Kc5 88. Rc4+ Kb6 89. Bf3 Nb4 90. Rc8 Be5 91. Re8 Bd6 92.
Rh8 Be7 93. Rh7 Bd8 94. Rb7+ Kc5 95. Rf7 Nc2 96. Ke2 Kd6 97. Rxf4 Bg5 98. Rc4
Na3 99. Rc6+ Kd7 100. Rc3 Nb1 101. Rc5 Be7 102. Rh5 Nd2 103. Kxe3 Nf1+ 104. Kd4
Ng3 105. Rh7 Ke6 106. Bg4+ Kf6 107. Bd7 Bd8 108. Rxh4 Bc7 109. Rh6+ Kg5 110.
Rc6 Bf4 111. Kd3 Be5 112. Rc5 Kf6 113. Rxe5 Kxe5 114. Ke3 Kxe3 115. Xxe3 Xxe3
116. Xxe3 Xxe3 117. Xxe3 Xxe3 118. Xxe3 Xxe3 119. Xxe3 Xxe3 120. Xxe3 Xxe3 121.
Xxe3 Xxe3 122. Xxe3 Xxe3 123. Xxe3 Xxe3 124. Xxe3 Xxe3 125. Xxe3 Xxe3 126. Xxe3
Xxe3 127. Xxe3 Xxe3 128. Xxe3 Xxe3 129. Xxe3 Xxe3 130. Xxe3 Xxe3 131. Xxe3 Xxe3
132. Xxe3 Xxe3 133. Xxe3 Xxe3 134. Xxe3 Xxe3 135. Xxe3 Xxe3 136. Xxe3 Xxe3 137.
Xxe3 Xxe3 138. Xxe3 Xxe3 139. Xxe3 Xxe3 140. Xxe3 Xxe3 141. Xxe3 Xxe3 142. Xxe3
Xxe3 143. Xxe3 Xxe3 144. Xxe3 Xxe3 145. Xxe3 Xxe3 146. Xxe3 Xxe3 147. Xxe3 Xxe3
148. Xxe3 Xxe3 149. Xxe3 Xxe3 150. Xxe3 Xxe3 151. Xxe3 Xxe3 152. Xxe3 Xxe3 153.
Xxe3 Xxe3 154. Xxe3 1-0