[Bug-gnubg] movefilter and pruning problem

[Andrew Cunningham]

Movefilter puzzelment. I think I'm misunderstanding something to do
with the movefilters. As I understood it if a move filter had no
settings (all 0s), then there would be no pruning of the search space
at that ply. This examples illustrates my problem.

This is only looking at 3 ply chequer play.

With move filters for 3 ply:
set evaluation movefilter 3 0 0 8 0.160000
****set evaluation movefilter 3 1 0 8 0.160000**** Note this line is
different from the Normal filter
set evaluation movefilter 3 2 0 2 0.040000

And a hint gives us-

>hint
    1. Cubeful 2-ply    24/16                        Eq.:  -0.215
       0.451 0.099 0.003 - 0.549 0.165 0.005
        2-ply cubeful prune [world class]
    2. Cubeful 2-ply    13/5                         Eq.:  -0.263 ( -0.048)
       0.434 0.116 0.004 - 0.566 0.169 0.009
        2-ply cubeful prune [world class]
    3. Cubeful 2-ply    24/22 13/7                   Eq.:  -0.269 ( -0.054)
       0.433 0.110 0.004 - 0.567 0.171 0.007
        2-ply cubeful prune [world class]
    4. Cubeful 2-ply    13/11 13/7                   Eq.:  -0.285 ( -0.070)
       0.429 0.119 0.004 - 0.571 0.176 0.010
        2-ply cubeful prune [world class]
    5. Cubeful 2-ply    13/7 6/4                     Eq.:  -0.342 ( -0.127)
       0.418 0.106 0.003 - 0.582 0.186 0.013
        2-ply cubeful prune [world class]

However if I used the default Normal setting for the move filter:

      Move filter for 3 ply:
        keep the first 0 0-ply moves and up to 8 more moves within equity 0.16
****   keep the first 0 1-ply moves*** Only difference from above
        keep the first 0 2-ply moves and up to 2 more moves within equity 0.04

>hint
    1. Cubeful 1-ply    24/16                        Eq.:  -0.244
       0.439 0.094 0.002 - 0.561 0.165 0.005
        1-ply cubeful prune
    2. Cubeful 1-ply    24/22 13/7                   Eq.:  -0.310 ( -0.066)
       0.417 0.106 0.003 - 0.583 0.179 0.008
        1-ply cubeful prune
    3. Cubeful 1-ply    13/5                         Eq.:  -0.313 ( -0.069)
       0.416 0.108 0.003 - 0.584 0.176 0.011
        1-ply cubeful prune
    4. Cubeful 1-ply    13/11 13/7                   Eq.:  -0.328 ( -0.084)
       0.412 0.112 0.003 - 0.588 0.181 0.012
        1-ply cubeful prune
    5. Cubeful 1-ply    13/7 6/4                     Eq.:  -0.399 ( -0.155)
       0.396 0.098 0.002 - 0.604 0.196 0.014
        1-ply cubeful prune

My point is that when I set a specific filter for 1ply I get a more
accurate evaluation to 2 ply for the hints. When I have 1ply set as
nothing, i would have thought that since no prunning happens this
would also evaluation the hints up to 2 ply, but as seen from the
example it does not.

This is all based on the command line interface, though I took the
settings for the Normal move filter from the gui.

Is it just me misunderstanding the pruning?

Thanks
Andrew Cunningham