[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-gnubg] blowfish dll vs gnubg
From: |
Joern Thyssen |
Subject: |
Re: [Bug-gnubg] blowfish dll vs gnubg |
Date: |
Fri, 30 Aug 2002 16:21:39 +0000 |
User-agent: |
Mutt/1.4i |
On Sat, Aug 24, 2002 at 10:35:19PM -0700, maareyes wrote
>
> Hi,
>
> I recently played this 9 pt match using the offline gnubg player on my
> FIBS client 3DFiBs. It is using gnubg.dll from Mike Rudman's Blowfish
> program. It was set at 1 ply for checker play and 0-ply cube action
> and the default settings of 8 candidates, .160 tolerance and no noise.
>
> I analyzed the match with the current GnuBG for Windows from Oystein's
> site with the same settings of 1 ply and it flagged very many errors
> for the bot and give it a beginner rating. (this is unusual because it
> consistantly rates ET or world class)
>
> I ran the analysis again with 2 ply and the bot is now rated world
> class
>
> I found this to be rather confusing. How can a player using Blowfish
> DLL (based on gnubg) at 1 ply make moves that 1 ply analyzer does not
> agree with but 2 ply rates world Class.
>
>
> I am using pretty much all default settings 8 candidates, .160
> tolerance.. no noise
>
>
> I still trust the anlayzer more than the Blowfish dll since it has
> made many errors that were obvious and that the analyzer flagged but
> in this case, 2 ply seems to agree with Blowfish dll. Is this just an
> isolated freak case?
>
> Anyway,, they are all ET or close most of the time so for any
> practical applications, they are just fine.. just wondering what
> happened here.. thats all
>
It looks a bit fishy...
I also did an analysis on 0-ply, 1-ply, and 2-ply, and the actual play
didn't match any of those. Do you know which version of eval.c that is
used in the DLL?
If I analyse the attached match on 0-ply I get:
GnuBG You
Moves unmarked 119 127
Moves marked doubtful 6 0
Moves marked bad 4 2
Moves marked very bad 1 0
Error rate (total) -1.649 (-31.725%) -0.462 (-4.903%)
Error rate (pr. move) -0.015 (-0.281%) -0.007 (-0.073%)
Checker play rating Expert World class
Total cube decisions 58 93
Actual or close cube decisions 25 56
Doubles 3 4
Takes 3 3
Pass 0 0
Missed doubles around DP 0 (-0.000, -0.000%) 4 (-0.213, -5.003%)
Missed doubles around TG 0 (-0.000, -0.000%) 1 (-0.003, -0.096%)
Wrong doubles around DP 1 (-0.119, -1.293%) 1 (-0.055, -1.551%)
Wrong doubles around TG 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Wrong takes 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Wrong passes 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Error rate (total) -0.119 (-1.293%) -0.272 (-6.651%)
Error rate (per cube decision) -0.005 (-0.052%) -0.005 (-0.119%)
Cube decision rating Extra-terrestrial
Extra-terrestrial
On 1-ply I get:
GnuBG You
Moves unmarked 100 123
Moves marked doubtful 10 4
Moves marked bad 9 1
Moves marked very bad 11 1
Error rate (total) -4.264 (-98.272%) -0.815 (-11.318%)
Error rate (pr. move) -0.038 (-0.870%) -0.012 (-0.169%)
Checker play rating Beginner Expert
Total cube decisions 58 93
Actual or close cube decisions 26 39
Doubles 3 4
Takes 3 3
Pass 0 0
Missed doubles around DP 2 (-0.041, -0.366%) 1 (-0.220, -4.993%)
Missed doubles around TG 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Wrong doubles around DP 1 (-0.222, -2.410%) 1 (-0.090, -2.529%)
Wrong doubles around TG 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Wrong takes 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Wrong passes 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Error rate (total) -0.263 (-2.776%) -0.310 (-7.522%)
Error rate (per cube decision) -0.010 (-0.107%) -0.008 (-0.193%)
Cube decision rating Expert World class
On 2-ply:
GnuBG You
Moves unmarked 124 125
Moves marked doubtful 4 3
Moves marked bad 0 0
Moves marked very bad 2 1
Error rate (total) -0.943 (-18.398%) -0.648 (-8.851%)
Error rate (pr. move) -0.008 (-0.163%) -0.010 (-0.132%)
Checker play rating World class World class
Total cube decisions 58 93
Actual or close cube decisions 25 55
Doubles 3 4
Takes 3 3
Pass 0 0
Missed doubles around DP 0 (-0.000, -0.000%) 2 (-0.243, -5.528%)
Missed doubles around TG 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Wrong doubles around DP 1 (-0.197, -2.132%) 2 (-0.073, -1.613%)
Wrong doubles around TG 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Wrong takes 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Wrong passes 0 (-0.000, -0.000%) 0 (-0.000, -0.000%)
Error rate (total) -0.197 (-2.132%) -0.316 (-7.140%)
Error rate (per cube decision) -0.008 (-0.085%) -0.006 (-0.130%)
Cube decision rating World class World class
--
Joern Thyssen, PhD
Vendsysselgade 3, 3., DK-9000 Aalborg, Denmark
+45 9813 2791 (private) / +45 2077 2689 (mobile) / +45 9633 7036 (work)