Michael Depreli started this thread on BGO.
http://www.bgonline.org/forums/webbbs_config.pl?noframes;read=47533
At first I couldn¹t reproduce what he saw but after it was observed that
pruning alters the output substantially I came up with this test (Its
reproducible on Linux as well, and is not thread related). First what you
need to do is load the GUI and change the evaluation parameters to be
Supremo for Checker and Word Class for Cube. (This should set things to 2ply
cube/2ply checker Pruning ON/cubeful checker evals for cube and checker
decisions and a Large filter. Now save the settings.
Re-launch gnubg from the command line (Similar thing happens in the GUI, the
CLI is just easier to script. Issue these commands:
set evaluation chequer eval plies 4
set evaluation cubedecision eval plies 4
set evaluation chequer eval prune on
set evaluation cubedecision eval prune on
set gnubgid Position ID: 1a4pwABsd7YAAA Match ID: EYERADAAAAAA
hint
set evaluation chequer eval prune off
set evaluation cubedecision eval prune off
set gnubgid Position ID: 1a4pwABsd7YAAA Match ID: EYERADAAAAAA
hint
My output is attached below. Some curiosities are that with Pruning ON the
eval took considerably longer than with it OFF. Most of the pruning off 4ply
results seem to be nearly what the 0 ply evals say. I am assuming this is
due in part to the ³Filter² being used (In this case large) but its
interesting it gives wildly differing results and pruning is much slower. I
think the question is, is this to be expected, and if not is there a bug?
Specifically look at ³Cubeful 4-ply 9/6 9/5 ³ results for both tests
below to see where the results end up being considerably different to get an
idea on how this was discovered. I think its clear pruning on got this right
and pruning off didn¹t. Is it the filter settings? And if so is the filter
being applied properly with prunign on and off?
-------
PRUNING ON:
Setting GNUBG id 1a4pwABsd7YAAA:EYERADAAAAAA
(gnubg) hint
1. Cubeful 4-ply 10/6 5/2 Eq.: +0.558
0.810 0.002 0.000 - 0.190 0.000 0.000
4-ply cubeful prune
2. Cubeful 4-ply 10/3 Eq.: +0.555 ( -0.003)
0.809 0.002 0.000 - 0.191 0.000 0.000
4-ply cubeful prune
3. Cubeful 4-ply 10/6 4/1 Eq.: +0.551 ( -0.007)
0.807 0.002 0.000 - 0.193 0.000 0.000
4-ply cubeful prune
4. Cubeful 4-ply 9/6 9/5 Eq.: +0.541 ( -0.018)
0.800 0.009 0.000 - 0.200 0.002 0.000
4-ply cubeful prune
5. Cubeful 4-ply 10/6 9/6 Eq.: +0.535 ( -0.023)
0.798 0.010 0.001 - 0.202 0.004 0.000
4-ply cubeful prune
6. Cubeful 4-ply 10/6 8/5 Eq.: +0.518 ( -0.040)
0.791 0.008 0.000 - 0.209 0.002 0.000
4-ply cubeful prune
7. Cubeful 2-ply 6/2 5/2 Eq.: +0.515 ( -0.043)
0.789 0.013 0.000 - 0.211 0.003 0.000
2-ply cubeful prune [world class]
8. Cubeful 2-ply 6/3 5/1 Eq.: +0.511 ( -0.048)
0.788 0.010 0.000 - 0.212 0.005 0.000
2-ply cubeful prune [world class]
9. Cubeful 2-ply 9/2 Eq.: +0.510 ( -0.049)
0.786 0.016 0.001 - 0.214 0.009 0.000
2-ply cubeful prune [world class]
10. Cubeful 2-ply 8/5 8/4 Eq.: +0.500 ( -0.058)
0.783 0.011 0.000 - 0.217 0.003 0.000
2-ply cubeful prune [world class]
PRUNING OFF:
Setting GNUBG id 1a4pwABsd7YAAA:EYERADAAAAAA
(gnubg) hint
1. Cubeful 4-ply 10/6 4/1 Eq.: +0.551
0.807 0.005 0.000 - 0.193 0.001 0.000
4-ply cubeful
2. Cubeful 4-ply 10/6 5/2 Eq.: +0.548 ( -0.003)
0.806 0.004 0.000 - 0.194 0.001 0.000
4-ply cubeful
3. Cubeful 4-ply 10/3 Eq.: +0.529 ( -0.022)
0.797 0.004 0.000 - 0.203 0.001 0.000
4-ply cubeful
4. Cubeful 4-ply 6/2 5/2 Eq.: +0.383 ( -0.167)
0.732 0.015 0.000 - 0.268 0.005 0.000
4-ply cubeful
5. Cubeful 4-ply 10/6 9/6 Eq.: +0.376 ( -0.175)
0.728 0.015 0.000 - 0.272 0.003 0.000
4-ply cubeful
6. Cubeful 4-ply 9/6 9/5 Eq.: +0.350 ( -0.201)
0.716 0.020 0.000 - 0.284 0.006 0.000
4-ply cubeful
7. Cubeful 2-ply 6/3 5/1 Eq.: +0.512 ( -0.039)
0.789 0.010 0.000 - 0.211 0.005 0.000
2-ply cubeful
8. Cubeful 2-ply 10/6 8/5 Eq.: +0.511 ( -0.040)
0.788 0.009 0.000 - 0.212 0.002 0.000
2-ply cubeful
9. Cubeful 2-ply 9/2 Eq.: +0.509 ( -0.042)
0.786 0.015 0.001 - 0.214 0.009 0.000
2-ply cubeful
10. Cubeful 2-ply 8/5 8/4 Eq.: +0.499 ( -0.052)
0.783 0.010 0.000 - 0.217 0.003 0.000
2-ply cubeful
_______________________________________________
Bug-gnubg mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/bug-gnubg
