Stockfish 1.5

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

Uri Blass
Posts: 10460
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Stockfish 1.5

Post by Uri Blass »

Tord Romstad wrote:
Hart wrote:Okay. I guess I will have to accept this as the answer. I am sure the real explanation would have gone over my head anyways.
I don't really think so. It could be explained perfectly well in chess-specific, non-technical terms. The problem is that it would require a tremendous lot of words. :)
One last question for you. I am running some tournaments in Arena201, and I am getting several forfeits on time by SF. The exact number is 9 out of 85. I use a time control of 30/1' which on my hardware should be comparable to CCRL 40/4. Is this normal or am I doing something wrong?
It's probably normal; I don't think you are doing anything wrong. Stockfish frequently loses on time when it has only a few seconds left on the clock, is in a desperate situation, and encounters a big fail low. Apparently it thinks it is lost anyway, and decides to use all the remaining time in the hope that it finds some better move at the last moment.

I'm not sure whether this behavior is intended (I did not introduce it), but I never bothered to look at it, because it only happens in hopeless positions (which means that Stockfish doesn't lose any points) and only when there is very little time left (which means that it doesn't waste any time for the user/tester).
big fail low can happen also when you originally believe that you are winning and suddenly discover that you have only a draw.

stockfish clearly evaluate some drawn tablebase positions as +4 and in my endgame tests it lost under Fritz gui the following position with the black pieces on time(time control 1+1)


[D]7Q/8/2K5/5q2/8/7p/5k2/8 b - - 0 76
Hart

Re: Stockfish 1.5

Post by Hart »

After 120 games with the "UCI filter" option checked, I have 0 losses on time. This is a major improvement from 9 losses out of 85 games.

Here are my results so far:

Code: Select all

2 Stockfish_15_x64_ja          12 120.0 ( 63.5 :  56.5)
                                   24.0 (  4.5 :  19.5) Rybka 3                  231
                                   24.0 ( 12.5 :  11.5) Naum                      -2
                                   24.0 ( 13.0 :  11.0) Zappa Mexico II x64      -15
                                   24.0 ( 14.0 :  10.0) MP-x64-Inert---Thinker   -33
                                   24.0 ( 19.5 :   4.5) Bright-0.4a             -193
Looks like a performance of ~3117 at CCRL 40/4 time controls for an improvement of ~50 Elo, not too shabby. More games to come...
mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 9:17 pm

Re: Stockfish 1.5

Post by mcostalba »

Hart wrote:After 120 games with the "UCI filter" option checked, I have 0 losses on time. This is a major improvement from 9 losses out of 85 games.
This demonstate that the timing issue is _all_ on the shoulders of Arena because that option is unkown to Stockfish that changes absolutly nothing with the option checked or unchecked.

BTW it is unkown also to the UCI protocol:

http://wbec-ridderkerk.nl/html/UCIProtocol.html


So should be something totally internal at Arena.
User avatar
George Tsavdaris
Posts: 1627
Joined: Thu Mar 09, 2006 12:35 pm

Re: Stockfish 1.5

Post by George Tsavdaris »

Hart wrote:After 120 games with the "UCI filter" option checked, I have 0 losses on time. This is a major improvement from 9 losses out of 85 games.
Where is this "UCI filter" ? I use the ancient but good Arena 1.1 BTW.
Or it's only on newer Arenas available?

And if this is the case that means on Arena 1.1 it steals time form engines and you can not do nothing about it? :?
After his son's birth they've asked him:
"Is it a boy or girl?"
YES! He replied.....
Hart

Re: Stockfish 1.5

Post by Hart »

I use 2.0.1 so I cannot comment on 1.1. In 2.0.1 it is under Engine > Mangage > Options
User avatar
George Tsavdaris
Posts: 1627
Joined: Thu Mar 09, 2006 12:35 pm

Re: Stockfish 1.5

Post by George Tsavdaris »

Hart wrote:I use 2.0.1 so I cannot comment on 1.1. In 2.0.1 it is under Engine > Mangage > Options
Oh this. Yes i have it enabled by default and have forgotten about it.
Yes it is on Arena 1.1 also and says:


"UCI Filter" lets Arena ignore useless information that is sent by engines during the first 5 seconds of a search. Important information like mainlines or the best move will of course always be processed though. During the first 5 seconds many engines send a lot of information to Arena, which may become very slow because of this. This has a big impact especially at short time controls. You can restrict "UCI Filter" to be used only in tournaments, if you select this option.
After his son's birth they've asked him:
"Is it a boy or girl?"
YES! He replied.....
Vinvin
Posts: 5239
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Stockfish 1.5

Post by Vinvin »

Tord Romstad wrote:It's probably normal; I don't think you are doing anything wrong. Stockfish frequently loses on time when it has only a few seconds left on the clock, is in a desperate situation, and encounters a big fail low. Apparently it thinks it is lost anyway, and decides to use all the remaining time in the hope that it finds some better move at the last moment.
Larry Kaufman reported a lost on time in a draw endgame : http://rybkaforum.net/cgi-bin/rybkaforu ... #pid188109
nepossiver
Posts: 38
Joined: Wed Sep 03, 2008 4:12 am

Re: Stockfish 1.5

Post by nepossiver »

Larry Kaufman reported a lost on time in a draw endgame : http://rybkaforum.net/cgi-bin/rybkaforu ... #pid188109
I've seen some losses on time in dead draw endgames (Rook+BishopvsRook, if I'm not mistaken), and if I recall it correctly they happened both under chessGUI and arena. Time control was 1+1. This was Stockfish 1.3.1 (or 1.2.1). I think (still from memory) Glaurung 2.1 & 2.2 do not show this behaviour. I am away from my computer, but I can try to find the games if there is interest.
User avatar
George Tsavdaris
Posts: 1627
Joined: Thu Mar 09, 2006 12:35 pm

Re: Stockfish 1.5

Post by George Tsavdaris »

SzG wrote:
George Tsavdaris wrote:
Hart wrote:I use 2.0.1 so I cannot comment on 1.1. In 2.0.1 it is under Engine > Mangage > Options
Oh this. Yes i have it enabled by default and have forgotten about it.
Yes it is on Arena 1.1 also and says:


"UCI Filter" lets Arena ignore useless information that is sent by engines during the first 5 seconds of a search. Important information like mainlines or the best move will of course always be processed though. During the first 5 seconds many engines send a lot of information to Arena, which may become very slow because of this. This has a big impact especially at short time controls. You can restrict "UCI Filter" to be used only in tournaments, if you select this option.
It looks as though it is always advantageous to switch this filter on. But as it is an option, there must be drawbacks. Exactly what?
I guess that needless information that are sent to Arena if it's OFF will not be sent. :D
I always run games with this filter ON(default is OFF) since as you say i notices back in the past that with OFF some engine lose on time.
After his son's birth they've asked him:
"Is it a boy or girl?"
YES! He replied.....
Hart

Re: Stockfish 1.5

Post by Hart »

Here is my first loss on time after 170 games. It looks like the program hung after move 107 with a score to its favor and with two minutes still on the clock in a 30/1' game. So, this could be from something totally separate from the time management issues I had earlier. After playing through this game again in infinite analysis, the last few positions from this game are causing SF to crash. This happens when running 1 thread and 4 but it does not happen if I clear the hash between moves. When it does crash it is not always on the same position.

[White "Zappa Mexico II x64"]
[Black "Stockfish_15_x64_ja"]
[Result "1-0"]
[TimeControl "30/60:30/60:30/60"]
[Termination "time forfeit"]
[PlyCount "215"]

1. e4 e6 2. d4 d5 3. Nc3 dxe4 4. Nxe4 Bd7 5. Nf3 Bc6 6. Bd3 Nd7 7. O-O Ngf6
8. Ng3 Be7 9. c4 Bxf3 10. Qxf3 c6 11. Bf4 {+0.58/13 8} Nf8 {-0.76/14 2} 12.
Rfd1 {+0.58/13 6} Ng6 {-0.68/15 5} 13. Be5 {+0.59/13 5} Bd6 {-0.48/14 1}
14. Bxg6 {+0.65/12 4} hxg6 {-0.48/15 2} 15. Qf4 {+0.61/13 4} Bb8 {-0.36/15
2} 16. Rac1 {+0.59/13 5} Nh5 {-0.28/15 2} 17. Qe3 {+0.71/13 1} Nxg3
{-0.40/15 2} 18. fxg3 {+0.47/13 2} Bxe5 {-0.20/15 2} 19. Qxe5 {+0.36/12 1}
Kf8 {-0.20/15 3} 20. d5 {+0.61/12 1} exd5 {-0.28/15 2} 21. Qd4 {+0.66/13 2}
b6 {-0.32/16 2} 22. cxd5 {+0.82/13 1} c5 {-0.28/15 2} 23. Qf4 {+0.89/12 1}
Qf6 {-0.28/15 2} 24. d6 {+1.02/12 1} Rd8 {-0.48/16 6} 25. Rc3 {+1.02/12 1}
Rh5 {-0.12/16 3} 26. Qxf6 {+0.95/12 1} gxf6 {-0.12/15} 27. Rf3 {+0.73/14 1}
Kg7 {-0.12/17 2} 28. d7 {+0.66/13} Re5 {+0.20/19 3} 29. Kf2 {+0.62/13 1} f5
{+0.28/19 2} 30. Ra3 {+0.69/9} a5 {+0.36/19 1} 31. Rad3 {+0.43/15 4} Kf6
{+0.20/18 2} 32. Rd6+ {+0.38/15 2} Re6 {+0.16/18 3} 33. Kf3 {+0.38/15 1}
Ke7 {+0.28/18 2} 34. R6d3 {+0.38/14 4} a4 {+0.36/17 2} 35. h3 {+0.20/13 4}
Re5 {+0.20/18 2} 36. R3d2 {+0.27/14 1} f6 {+0.08/18 3} 37. Rd6 {+0.16/14 1}
Re1 {+0.40/18 1} 38. R1d2 {+0.11/14 2} Re6 {+0.24/18 3} 39. R6d3 {+0.13/14
1} b5 {+0.20/18 1} 40. a3 {+0.05/14 3} c4 {+0.08/17 1} 41. Rd5 {0.00/14 1}
Rb6 {0.00/18 1} 42. Re2+ {+0.14/13 1} Kf7 {0.00/18 1} 43. Ke3 {+0.10/15 2}
Rb7 {+0.32/20 3} 44. Kd4 {-0.22/15 2} Rbxd7 {+0.24/21 3} 45. Rxd7+
{-0.22/14} Rxd7+ {+0.20/18} 46. Kc5 {-0.22/14} Rd3 {0.00/21 3} 47. Kxb5
{+0.20/14} Rxg3 {+0.24/19 1} 48. Kxc4 {+0.19/14} Rb3 {+0.24/20 1} 49. Kd5
{0.00/15} g5 {+0.28/19 1} 50. Rf2 {-0.02/15 1} f4 {+0.32/17 1} 51. Kd6
{-0.16/16 1} Rg3 {+0.44/18 1} 52. Kd5 {-0.16/14} Kg6 {+0.24/18 1} 53. Kd4
{-0.16/15} f3 {+0.68/19 1} 54. Rxf3 {-0.42/15 3} Rxg2 {+0.92/19 3} 55. b3
{-0.42/15 1} axb3 {+1.09/20 1} 56. Rxb3 {-0.29/11} f5 {+1.05/18} 57. Rb6+
{-0.53/13 1} Kh5 {+1.09/18 1} 58. Rb3 {-0.53/14 1} Kh4 {+1.93/19 1} 59. Rf3
{-1.54/14 4} Rg3 {+2.42/19 1} 60. Rxg3 {-0.90/12 1} Kxg3 {+2.54/14} 61. a4
{-0.90/12} f4 {+3.55/19 3} 62. a5 {-0.90/12} f3 {+3.75/20 2} 63. a6
{-0.90/12} f2 {+3.75/16} 64. a7 {-0.90/12} f1=Q {+3.75/17} 65. a8=Q
{-0.90/11} Kxh3 {+3.71/18 3} 66. Ke5 {0.00/63} g4 {+3.71/18 2} 67. Qc8
{0.00/63} Qf3 {+4.20/19 3} 68. Qh8+ {0.00/62} Kg2 {+4.20/18 2} 69. Kd6
{0.00/28} Qf4+ {+4.28/17 2} 70. Kd7 {0.00/63} g3 {+4.28/16 2} 71. Kc8
{0.00/62} Kf1 {+4.20/18 2} 72. Qa1+ {0.00/63} Kf2 {+4.24/19 1} 73. Qa2+
{0.00/63} Kf3 {+4.40/18 1} 74. Qb3+ {0.00/63} Kg4 {+4.28/18 1} 75. Qg8+
{0.00/63} Qg5 {+4.64/19 1} 76. Qc4+ {0.00/9} Kh5 {+5.01/20 2} 77. Qf7+
{0.00/63} Kh4 {+5.05/21 3} 78. Qh7+ {0.00/63} Qh5 {+5.21/20 2} 79. Qe7+
{0.00/63} Kh3 {+5.17/19 1} 80. Qe3 {0.00/63} Qg4+ {+5.33/18 1} 81. Kb7
{0.00/63} Qg7+ {+5.21/17 1} 82. Kc6 {0.00/63} Qf6+ {+5.97/18 2} 83. Kc7
{0.00/63} Kh2 {+6.50/20 2} 84. Qd2+ {0.00/62} g2 {+6.38/15} 85. Kb7
{0.00/63} Qf7+ {+6.70/19 1} 86. Ka8 {0.00/63} Kg3 {+6.10/19 2} 87. Qe1+
{0.00/63} Qf2 {+6.10/22 1} 88. Qe5+ {0.00/63} Qf4 {+5.97/22 1} 89. Qg7+
{0.00/62} Kf3 {+5.89/22 1} 90. Qc3+ {0.00/63} Qe3 {+5.69/21 1} 91. Qf6+
{0.00/63} Kg3 {+5.69/24 4} 92. Qg6+ {0.00/63} Kf2 {+5.69/23 2} 93. Qc2+
{0.00/13} Kf3 {+5.69/23 2} 94. Qf5+ {0.00/63} Qf4 {+5.65/23 1} 95. Qd3+
{0.00/63} Kf2 {+5.61/24 2} 96. Qc2+ {0.00/63} Kg3 {+5.61/23 1} 97. Qg6+
{0.00/63} Qg4 {+5.61/23 1} 98. Qd6+ {0.00/9} Kh3 {+5.61/23 1} 99. Qh6+
{0.00/63} Qh4 {+5.57/17} 100. Qe6+ {0.00/63} Kh2 {+5.57/23 1} 101. Qd6+
{0.00/63} Qg3 {+5.57/22 1} 102. Qd2 {0.00/63} Qf3+ {+5.57/20 1} 103. Ka7
{0.00/63} Qb3 {+5.57/21 2} 104. Qh6+ {0.00/63} Kg3 {+5.57/22 2} 105. Qg5+
{0.00/63} Kf2 {+5.57/22 1} 106. Qf6+ {0.00/63} Ke2 {+5.57/21 1} 107. Qa6+
{0.00/12} Qd3 {+5.57/22 2} 108. Qa2+ {0.00/11 1-0 Black forfeits on time}
1-0