Re: [Bug-gnubg] improvement of cubeful equities in match play bearoffs

[Joseph Heled]

So lets recap:

eq(cubeful) - from the 2 sided database. This was computed for money, 
but assuming cube is 1 or 2 (am I right? otherwise you need to save 
equities for 2,4,8,16 etc)

eq(fully live cube) - from the theoretical model, assuming continuity 
and perfect recubes points.

eq(cubeless) - from the database.


You probably want to check uneven scores, i.e. ones where gammon price 
is very different from money.

-Joseph


Joern Thyssen wrote:
> On Fri, Apr 04, 2003 at 08:21:18AM +1200, Joseph Heled wrote
>
> > Sounds great. Can you tell more about the theoretical side, i.e. how do 
> > you get X from the equities?
>
>
> For owned cube and unavailable cube the Janowski formula is
>
> eq(cubeful) = (1-x) eq(cubeless) + x eq(fully live cube)
>
> For centered cube the above formula is a reasonable approximation, so I
> use it there as well.
>
> I've got eq(cubeful) from the two-sided bearoff database, and
> eq(cubeless) and eq(fully live) can also easily be calculated, so it's
> trivial to calculate X.
>
> I then feed X into the match play formula:
>
> mwc(cubeful) = (1-x) mwc(cubeless) + x mwc(fully live cube).
>
> Jørn
>
>
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/bug-gnubg