Re: [Bug-gnubg] Wrong cube errors categories in analysis statistics panel

[Holger]

At 10:54 30.09.2003, olivier croisille wrote:
> 1/  I think somebody wanted to change the 2 analysis categories 
> 'Missed/Wrong doubles around TG' to rename them 'Missed/Wrong doubles 
> around CP' (cash point) : but at the moment both are shown in the stats 
> panel, 'TG' and 'CP', eventhough they refer to the same thing (cf. stats 
> panel copied at the end of post)

Recently, I introduced the classification into 3 categories as opposed to 2 
before (added "around CP"). A wrong cube decision either consists of a 
missed double or a wrong double. With categories of "around DP", "around 
CP" and "around TG" this makes for 6 possibilities. However, later I 
realised that a wrong double around the cash point should not be possible. 
Together with a related cube question I had asked here for confirmation, 
but unfortunately didn't get any response on this post. (I thought to 
better ask even if it's a stupid question, because I lack a lot of bg 
theory.) (see thread "Missed doubles problem ?")

> BTW, it seems to me the proper categories should be named as follows:
> 'Missed doubles around CP' : when in fact you should have cashed the position
> 'Wrong doubles around TG' : the position was intrisically too good, but 
> you cashed

Not sure I understand this. Those two categories exist. Do you mean to 
remove "Wrong doubles around CP"? I'd agree with this. This raises another 
question: Where should the divider be placed - on the arithmetic mean of DP 
and TG or on CP?

> 2/ In the same money game analysis, GNUBG classified my cube error in the 
> 'Wrong doubles around CP' category. But this isn't right at all, it 
> definitely was a 'Wrong doubles around DP' : cf. game file attached in SGF 
> format, move #8 for French Kiss

That is due to a bug I was asking about in the same mail. The calculation 
of the dividers fails sometimes leading to wrong classification.

From the above mentioned posting:
> What getMatchPoints() returns and what is listed under Market Window 
> differs. Namely, the live values for CP and TG are swapped respective to 
> the players.
>
> (gdb) p aaarPointsMatch
> $1 = {{{0.330667853, 0.257251322}, {0.513904393, 0.513904393},
>        {0.697271645, 0.724131286}, {0.902519166, 0.724131286}},
>       {{0.302728355, 0.275868714}, {0.486095399, 0.486095399},
>         ^^^^^^^^^^^                 ^^^^^^^^^^^
>        {0.669332147, 0.742748678}, {0.880213678, 0.742748678}}}
>         ^^^^^^^^^^^                              ^^^^^^^^^^^
>
> The market window reports (I suppose correctly) 74,275% for player0 and 
> 72,413% for player1.

Does someone know why getMatchPoints() does this and whether this is intended?

I'd also like to repeat another question on this topic:
To categorize cubes gnubg compares the arithmetic means of DP and CP, and 
CP and TG with the winning probabilities.
Could someone please explain me why (and confirm that) this is correct?
I'm wondering about the following: The winning probability (aarOutput[ 0 ][ 
OUTPUT_WIN ]) doesn't include any added value for gammon or bg chances 
(aarOutput[ 0 ][ OUTPUT_WINGAMMON ] and aarOutput[ 0 ][ 
OUTPUT_WINBACKGAMMON ]). Those match points (DP, CP, TG) have an entity of 
cubeful equity, don't they? And I think equities are calculated as the sum 
of winning chances, gammon chances and bg chances, thus include gammon and 
bg value. If all this was correct so far, then I suppose these 2 values 
can't be compared.

Regards,

        Holger