bug-gnubg
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug-gnubg] Strange evaluation of position for hypergammon game


From: Nikolay Marinov
Subject: [Bug-gnubg] Strange evaluation of position for hypergammon game
Date: Tue, 16 Jan 2007 11:57:33 +0200
User-agent: Mozilla Thunderbird 1.0.2 (Windows/20050317)

Hi All,

I have a problem with gnubg evaluation in hypergammon games. I generated hyper3.bd, but in spite of this in some situtations evaluation of given postions is very bad. I will attach You a game log in ".mat" format in which you can see the last game of match where the correct decision is to remove last checker from opponents field. Instead gnubg evaluation program shows a hint (in 0ply,and 2ply) which keeps this checkers in opponents field, but in the next turn opponent 100% will win with backgammon. Also the coeficients in 0ply looks strange, because win coeficient is 1 in situation where losing game is 100%. If I evaluate all movements in 2ply gnubg shows that there is greater move. Do You have some ideas from where this confusion may come? I use gnubg 0.14 version, in Windows XP SP2.

Regards,
Nikolay.
 9 point match

 Game 1
zero : 0                achilleas : 0
  1) 63: 23/17 17/14                42: 24/20 22/20
  2) 21: 24/22 14/13                52: 23/18 20/18
  3) 65: 22/16 16/11                54: 20/15 15/11
  4) 42: 22/18 11/9                52: 18/13 13/11
  5) 42: 13/9 18/16                21: 11/9* 11/10
  6) 33: 25/22 22/19 19/16* 9/6                51: 25/20 20/19*
  7) 32: 25/22 9/7*                54: 25/21 10/5
  8)  Doubles => 2                 Takes
  9) 43: 7/3 16/13                32: 19/16 16/14
  10) 42: 13/9 22/20*                22: 25/23 23/21 21/19(2)
  11) 41: 9/5 20/19                21: 14/12 12/11
  12) 41: 19/15 15/14*                32: 25/22* 22/20*
  13) 54: 25/20 25/21                63: 20/14 14/11*
  14) 21: 25/23 20/19                51: 19/14 19/18
  15) 33: 21/18 19/16 18/15 16/13                65: 18/12* 14/9
  16) 55: 25/20 23/18 18/13* 13/8                52: 25/20 11/9
  17) 41: 15/11 20/19                42: 20/16 16/14*
  18) 54: 25/20 8/4                42: 14/10 10/8
  19) 64: 20/14 19/15                 Doubles => 4
  20)  Drops                

 Game 2
zero : 0                achilleas : 2
  1)                         51: 24/19 23/22
  2) 33: 22/19 19/16 16/13 13/10                42: 19/15* 22/20
  3) 42: 25/23 24/20                33: 22/19 20/17 19/16 17/14
  4)  Doubles => 2                 Takes
  5) 62: 23/17 17/15                64: 16/10* 14/10
  6) 42: 25/21 21/19                51: 15/10 10/9
  7) 44: 23/19 20/16* 16/12 12/8                64: 25/19 19/15
  8) 31: 19/16 8/7                52: 15/10 10/8
  9) 54: 19/14 14/10                 Doubles => 4
  10)  Takes                32: 10/7 10/8
  11) 61: 16/10 7/6                51: 8/3 7/6
  12) 11: 10/9(2) 9/8(2)                42: 8/6 6/2
  13) 64: 8/2 8/4                31: 3/o 6/5
  14) 43: 4/o 6/3                51: 5/o 2/1
  15) 21: 3/1 2/1                42: 1/o


 Game 3
zero : 0                achilleas : 6
  1)                         42: 24/20 22/20
  2) 32: 24/21 23/21                32: 23/20 20/18
  3) 54: 21/16 16/12                44: 20/16(2) 16/12(2)
  4) 55: 22/17 21/16 17/12 12/7*                51: 25/20 20/19
  5) 66: 16/10 12/6* 10/4 7/1                54: 25/21* 21/16
  6) 21: 25/24 6/4                11: 12/11(2) 11/10(2)
  7) 52: 24/19 19/17                33: 10/7(2) 7/4(2)
  8) 54: 17/12 12/8                21: 16/14 14/13
  9) 61: 8/2 4/3                 Wins 1 points


 Game 4
zero : 1                achilleas : 6
  1)                         65: 23/17 22/17
  2) 54: 23/18 18/14                31: 24/21 21/20
  3) 22: 22/20 20/18 14/12 12/10                62: 20/14 17/15*
  4) 21: 25/24 18/16                31: 17/14 15/14
  5) 51: 16/15 15/10                66: 14/8(3) 8/2
  6) 54: 24/19 24/20                11: 8/7(2) 7/6(2)*
  7) 63: 25/22 20/14                65: 6/o 6/1
  8) 62: 10/4 4/2                11: 2/1 1/o(2)



reply via email to

[Prev in Thread] Current Thread [Next in Thread]