Re: [Bug-gnubg] gnubg.sql - stats from my database

[address@hidden]

Hello Jim,

Thanks for your advice and kind offer to help. How do your 
scripts run? Do you just feed it a folder of *.sgf files?

The 
statistics I would be interested in are pretty much the same as the 
headers from the webpage - 

http://www.capp-sysware.com/analysis/octnov2010-dc-dicestudy.txt
(I've 
pasted an abbreviated version of this page at the bottom of this post.)


Plus a couple of others - it would be good if there was also a column 
for the 'running total all time' for each category.

+ pip-count loss 
from hits / vs gnubg's stats
+ Total pip count per game / per match - 
for me / GNU 
+ Total number of doubles rolled whilst on the bar for me 
/ GNU 
+ number of occasions successfully hit a single blot when within 
1 dice roll range (when you actually want to hit it i.e. not leaving 
something silly open for your opponent to get of the bar and return a 
hit. I know this bit sounds difficult to calculate) / vs gnubg's stats

+ number of occasions successfully hit a single blot when within 2 dice 
rolls range (when you actually want to hit it) / vs gnubg's stats


Finally, I know that you are not the person to implement it, but do you 
think this was an interesting idea, 

> Another
> clarifying feature 
would be, after I lose to gnubg (again), to be able 
> to play the game 
again.
> But we would swap the CPU's dice rolls with my own.
> This 
would clearly show if gnubg would still win when you have his now 
> 
predetermined 'lucky rolls'.

or is it a waste of time for me to record 
the dice rolls and try and reply the games?


P.S. Your other 
suggestion sounds good too but I'm sure your round tuit list is just as 
big as mine!
P.P.S. I played a lot yesterday - lost 10 games to 5 
against gnubg. Today I'm winning 3-2 :)

Cheers,

djskope 



>
>If you 
analyse your matches and save the results as .sgf files, then
>it is 
possible to extract all sorts of per-move information from an
>analysed 
match. I have (when trying to settle an argument with someone
>about 
"usable" doubles) written scripts which can extract almost any

>information you want - doubles, doubles from the bar, pip-count loss

>from hits, time spent dancing on the bar, you name it. 
>

>Specifically, your requests:
>
>1) Given a collecton of sgf files, I 
can easily generate the number of
>doubles for each player (including 
if the player was able move at
>least one piece one time - a usable 
double)
>
>2) I have a script which calculates pip loss from being hit, 
167 - the
>loss = total pip count
>
>3) would be fairly easy to modify 
an existing script to do
>
>4) would be more work, but certainly do-
able. My script for counting
>dancing reconstructs the board for each 
dice roll so it would know if
>there are pieces on the bar
>
>5) The 
most useful (which I will do one of these days when my round
>tuit 
supply is replenished) would be to create a database of mistakes
>- 
cube and checker play with an EMG loss amount (so you can choose
>what 
level of mistake you want to analyse), the gnubg board and match
>ID 
(so you can put the mistake back into gnubg), and an easy way for
>an 
external script to input the board and match ID, allow you to play
>one 
move, then stop and analyse the move afterward, reporting the
>results 
back to the script. It could then tell you what you played
>before, 
what you just played and what gnubg believes the best
>move/cube action 
is for that position and match state.
>
>-- 
>Jim Segrave           
address@hidden
>


======================================================================================

======================================================================================




Statistics for First Roll of a Game (Can't be a Double)

Roll       
Count    Expected(%)   Observed(%)    Difference(%)   Two-Tailed p-value
(%)

======================================================================================

21            368     6.667%    (  6.88880569%,  +0.22213902% )        
51.037%
31            343     6.667%    (  6.42081617%,  -0.24585049% ) 
       49.290%
...

All          5342   100.000%    (100.00000000%,  +0.
00000000% )              -

======================================================================================



Statistics for all Regular Rolls (Excludes First Roll)

Roll       
Count    Expected(%)   Observed(%)    Difference(%)   Two-Tailed p-value
(%)

======================================================================================

21          11303     5.556%    (  5.61171296%,  +0.05615740% )        
27.121%
31          11314     5.556%    (  5.61717423%,  +0.06161868% ) 
       22.732%
...

All        201418   100.000%    (100.00000000%,  +0.
00000000% )              -
Doubles     33110    16.667%    ( 16.43845138%,  
-0.22821529% )          0.599%
NonDbls    168308    83.333%    ( 
83.56154862%,  +0.22821529% )           0.599%

Average

PipCount              8.167%    (  8.13676533%,  -0.02990133% )         
0.179%

======================================================================================



Statistics for All Regular Rolls from the Bar

Roll       Count    
Expected(%)   Observed(%)    Difference(%)   Two-Tailed p-value(%)

======================================================================================

21           2805     5.556%    (  5.49924520%,  -0.05631035% )        
57.876%
31           2977     5.556%    (  5.83645382%,  +0.28089826% ) 
        0.561%
...

All         51007   100.000%    (100.00000000%,  +0.
00000000% )              -
Doubles      8327    16.667%    ( 16.32521027%,  
-0.34145640% )          3.852%
NonDbls     42680    83.333%    ( 
83.67478973%,  +0.34145640% )           3.852%

Average

PipCount              8.167%    (  8.12498285%,  -0.04168382% )         
2.850%

======================================================================================



Statistics for all Die values (First rolls included)

Roll       
Count    Expected(%)   Observed(%)    Difference(%)   Two-Tailed p-value
(%)

======================================================================================

1           69392    16.667%    ( 16.78080867%,  +0.11414200% )         
4.889%
2           68994    16.667%    ( 16.68456181%,  +0.01789514% ) 
       75.749%
...

All        413520   100.000%    (100.00000000%,  +0.
00000000% )              -

======================================================================================



Statistics for bringing one checker in off the bar

               
Total  # Times  # Times      Observed        Expected      
Difference     Two-Tailed
  Against      Moves   Danced  Success     
Success (%)     Success (%)       (%)         p-value(%)

===================================================================================================

0-pt board        46        0       46    100.00000000%  100.00000000%   
+0.00000000%        -
1-pt board      5961      155     5806     
97.39976514%      97.22222222%   +0.17754292%     43.050%
2-pt board      
8469      918     7551     89.16046759%   88.88888889%   +0.27157870%     
43.658%
...

===================================================================================================



Statistics for bringing two checkers in off the bar

               
Total  # Times  # Times      Observed        Expected      
Difference     Two-Tailed
  Against      Moves   Danced  Success     
Success (%)     Success (%)       (%)         p-value(%)

===================================================================================================

0-pt board         1        0        1    100.00000000%  100.00000000%   
+0.00000000%        -
1-pt board       779      226      553     
70.98844673%      69.44444444%   +1.54400228%     37.106%
2-pt board      
1058      570      488     46.12476371%   44.44444444%   +1.68031926%     
27.895%
...

===================================================================================================



Statistics for consecutive rolls that are doubles

# in a row      
Count      Expected(%)     Observed(%)   Difference(%)   Two-Tailed p-
value(%)

==============================================================================================

1 doubles        33110    16.66666667%  ( 16.43845138%,  -0.22821529% ) 
        0.599%
2 doubles         5414     2.77777778%  (  2.68794249%,  
-0.08983529% )          1.415%
...

==============================================================================================



Statistics for consecutive identical rolls


                            [  2 in a row 
]                                  [  3 in a row 
]                                  [  4 in a row 
]                                  [  5 in a row ]                 

                  Expected    Diff. From    Two-Tailed             
Expected    Diff. From    Two-Tailed             Expected    Diff. 
>From    Two-Tailed             Expected    Diff. From    Two-Tailed

Roll    Count        (%)      Expected(%)   p-value(%)   Count        
(%)      Expected(%)   p-value(%)   Count        (%)      Expected(%)   
p-value(%)   Count        (%)      Expected(%)   p-value(%)

=========================================================================================================================================================================================================

21        641   0.30864198% ( +0.00960168%,  43.724%)       40   
0.01714678% ( +0.00271242%,  35.252%)        4   0.00095260% ( +0.
00103332%,    NeD  )        0   0.00005292% ( -0.00005292%,    NeD  )

31        657   0.30864198% ( +0.01754536%,  15.573%)       38   
0.01714678% ( +0.00171946%,  55.561%)        2   0.00095260% ( +0.
00004036%,    NeD  )        0   0.00005292% ( -0.00005292%,    NeD  )

...

All       868   0.46296296% ( -0.03201836%,   3.428%)       20   
0.01286008% ( -0.00293048%,  24.612%)        0   0.00035722% ( 
-0.00035722%,    NeD  )        0   0.00000992% ( -0.00000992%,    NeD  
)

=========================================================================================================================================================================================================



* NeD = Not Enough Data (npq < 5). Sample size is too small for 
binomial test to be accurate