Re: [Bug-gnubg] More on use of noise in rollouts

[Christian Anthon]

Hi Bob,

I wont have time to look into it any time soon, so here is just a
couple of suggestions.

a) truncation of the rollouts may help you track down what is going on
b) turn on rollout logs by issuing the commands

set rollout log on
set rollout logfile bob_is_an_intermediate_player

which will give you a .sgf file for each trial.

Combining a) and b) should give you a good idea about what is going on.

Christian.

On Tue, Feb 10, 2009 at 8:42 AM, bob koca <address@hidden> wrote:
>
>   I checked and the previous problem of the noise making a difference even
> though every move is forced is gone now.
>
>   I did a RO of the opening position with RO playing as intermediate vs.
> expert player. I obtained that the intermediate wins about
> 17% of 1 point matches. So something is wrong, possibly just labeling the
> amount of noise given as being intermediate. Note that
> 2200 vs 800 ELO gives about 17% win chance which is more like world
> class vs. beginner.
>
>   Also it seems strange that the opening roll is not worth anything. When i
> gave the intermediate player the opening roll (non doubles)
> the intermediate won 16.6668% of the matches. When the expert got it the
> intermediate won 17.2225% matches. Could it really be
> that the play is so bad that the intermediate does better to not win the
> opening roll?  Or did I mess something up? Thanks for any help,
>
> Bob Koca
>
>      This is with the expert having first roll:
>
>  GNU Backgammon  Position ID: 4HPwATDgc/AIMA
>                     Match ID   : MAElAAAAAAAA
>     +-1--2--3--4--5--6-------7--8--9-10-11-12-+  O: gnubg
>     | X              O |   |    O           X |  0 points
>     | X              O |   |    O           X |  Rolled 21
>     |                O |   |    O           X |
>     |                O |   |                X |
>     |                O |   |                X |
>     |                  |BAR|                  |^ 1 point match (Cube: 1)
>     |                X |   |                  |
>     |                X |   |                O |
>     |                X |   |    X           O |
>     | O              X |   |    X           O |
>     | O              X |   |    X  O        O |  0 points
>     +24-23-22-21-20-19------18-17-16-15-14-13-+  X: user
>
> 1. Rollout          16/13                        MWC:  17.22248%
>        0.172225 0.051257 0.006252 - 0.827775 0.490876 0.128110 CL  17.22248%
>       [0.006618 0.005328 0.001462 - 0.006618 0.009700 0.008089 CL
> 0.66179%]
>         Full cubeless rollout with var.redn.
>         1296 games, rollout as initial position, Mersenne Twister dice
> generator with seed 24000
>         Player 0:
>         Play: 0-ply cubeful, noise 0.04 (d) [intermediate]
>         Cube: 0-ply cubeful, noise 0.04 (d)
>         Player 1:
>         Play: 0-ply cubeful [expert]
>         Cube: 0-ply cubeful, noise 0.04 (d)
>
>
> And This is with the intermediate having the first roll:
>
>      GNU Backgammon  Position ID: 4HPwATDgc/AgMA
>                     Match ID   : cAktAAAAAAAA
>     +24-23-22-21-20-19------18-17-16-15-14-13-+  O: gnubg
>     | X              O |   | X  O           X |  0 points
>     | X              O |   |    O           X |
>     |                O |   |    O           X |
>     |                O |   |                X |
>     |                O |   |                  |
>     |                  |BAR|                  |v 1 point match (Cube: 1)
>     |                X |   |                O |
>     |                X |   |                O |
>     |                X |   |    X           O |
>     | O              X |   |    X           O |  Rolled 23
>     | O              X |   |    X           O |  0 points
>     +-1--2--3--4--5--6-------7--8--9-10-11-12-+  X: user
>
>
>             1. Rollout          18/13                        MWC:  83.33321%
>        0.833332 0.487893 0.118926 - 0.166668 0.041701 0.003942 CL  83.33321%
>       [0.006815 0.009393 0.008023 - 0.006815 0.006001 0.001675 CL
> 0.68155%]
>         Full cubeless rollout with var.redn.
>         1296 games, rollout as initial position, Mersenne Twister dice
> generator with seed 24000
>         Player 0:
>         Play: 0-ply cubeful, noise 0.04 (d) [intermediate]
>         Cube: 0-ply cubeful, noise 0.04 (d)
>         Player 1:
>         Play: 0-ply cubeful [expert]
>         Cube: 0-ply cubeful, noise 0.04 (d)
>
>
>
>
>
> ________________________________
> Windows Live™: Keep your life in sync. Check it out.
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/bug-gnubg
>
>