bug-gnubg
[Top][All Lists]

## Re: [Bug-gnubg] cube decision analysis in 'limited' cash games...

 From: Adi Kadmon Subject: Re: [Bug-gnubg] cube decision analysis in 'limited' cash games... Date: Sat, 18 Sep 2010 22:42:49 +0200

Hello,

I'm aware that mine is a rather late comment...

Nevertheless: I do generally play single games to the maximum 8-cube stakes, and I use to tackle the problem pointed to by Christian Anthon thus:
After a group of games had been played, when I "batch analyse" this group, I always eliminate from the group (i.e. I manually delete from "savedgames") those games (very infrequent btw) in which the cube was indeed turned to 8   a n d   in which the owner of the 8-cube could have reasonably redoubled, or in which either side could have reasonably changed one's checker-play in order to "save" or "win" a gammon.
Thereby I'm left only with those 8-cube-turned games in which the inaccuracies involved in the analysis are minor if they exist at all.

I would strongly recommend (as others probably did before me) to include in GNU some settings taking into account (a) the maximum possible stakes in each specific single game; (b) the influences on cube-evaluation of the discrepancies between the exact-2-product doubles and any less-then-2-product maximum stakes relative to the initial sum.

Thanks,

On Mon, Jun 28, 2010 at 10:55 AM, Christian Anthon wrote:
Constructing the right match equity table is quite easy, but we need
to make it right also for fractional stakes. A bit of coding changes
are needed to allow for jacoby in matches. I'm sure that I'll get
around to it some day.

Christian.

On Sun, Jun 27, 2010 at 9:04 AM, Julian Wilson <address@hidden> wrote:
> hello all.
>
> when playing money games online the players usually 'limit' their losses by setting a stake per point but with a maximum loss (e.g. £10 per point £80 maximum loss).   Assuming these parameters  and the cube is at 4 or 8, then I'm concerned that the cube decision analysis is wrong.  My concern is that as the game is limited and the cube (say) at 8 (in the example given) then no player can win a double or treble game and more importantly the redouble option is also of no value.
>
> Is this a real limitation of the cube decision analysis? And if yes then are there any settings options (in GNU) to accurately model cube decisions in limited loss cash games?
>
> Thanks in anticipation.
>
> Julian
>
>
>
> _______________________________________________
> Bug-gnubg mailing list