Re: [Bug-gnubg] More gammons than wins?

[nedc]

On Sun, 10 Oct 2004 09:43:10 +0200
 address@hidden wrote:
> Hello,
>
> i just did a rollout with a surprising result. More 
> gammons than wins occur here. 
>
> Newest windows version 1.1645 (Sept 6th, 2004)   
>
>
>    GNU Backgammon  Position ID: zLbBAWDb7AMAIg
>                    Match ID   : cAlgAAAACAAA
>    +24-23-22-21-20-19------18-17-16-15-14-13-+  O: 
> Player 0
>    | X     O  X  O  O | O | O  O             |  0 points
>    |       O     O  O | O | O  O             |  
>    |                  |   |                  |  
>    |                  |   |                  |  
>    |                  |   |                  |  
>    |                  |BAR|                  |v 3 point 
> match (Cube: 1)
>    |                X |   |                  |  
>    |                X |   |                  |  
>    |                X |   |                O |  
>    | X  X  X     X  X |   |                O |  On roll
>    | X  X  X     X  X |   |                O |  1 point
>    +-1--2--3--4--5--6-------7--8--9-10-11-12-+  X: 
> Player 1
>
>
> Cube analysis
> Rollout cubeless equity  +1.0881 (Money:  +0.6556)
>
> Cubeful equities:
> 1. No double            +0.7877
> 2. Double, pass         +1.0000  ( +0.2123)
> 3. Double, take         -0.0727  ( -0.8604)
> Proper cube action: No double, take (80.2%)
> Rollout details:
> Centered 1-cube:
>  0.6436 0.5092 0.0046 - 0.3564 0.0535 0.0919 CL  +1.0881 
> CF  +0.7877
> [0.0279 0.0281 0.0015 - 0.0279 0.0192 0.0816 CL   0.1173 
> CF   0.1174]
> Player Nanome owns 2-cube:
>  0.6185 0.7296 0.0044 - 0.3815 0.0849 0.0102 CL  +0.9044 
> CF  -0.0727      <-------
> [0.0258 0.1872 0.0013 - 0.0258 0.0219 0.0063 CL   0.1465 
> CF   0.2016]
> Full cubeful rollout with var.redn.
> 8 games, Mersenne Twister dice gen. with seed 943883169 
> and quasi-random dice
> Play: 0-ply cubeful [expert]
> Cube: 0-ply cubeful [expert]
>
> Thx in advance,
>
> Joachim Matussek

Not really if you look at the standard error figure. Your 
rollout has only 8 trials and the standard error is still 
19%.

Also the gammons figure after double/take is meaningless 
anyway since at this match score the game becomes DMP. 
The gammon figure for "centered 1-cube" is the one that 
is relevant.

Ned