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

[Joern Thyssen]

On Fri, Oct 03, 2003 at 09:07:57AM +0200, address@hidden wrote
> >>
> >>This seems a good point - I think picking the point needs to be done
> >>in terms of MWC, it's not enough to do equity, since in some match
> >>points, gammons are irre;evant.
> >
> >Ok, I'll compare them with an entity of MWC. But I think this doesn't solve
> >the problem (if my conclusions were correct) that the winning probability
> >doesn't account for gammon and bg values in games where they matter.
> >eq2mwc() gives a MWC, but the winning probability is no equity, is it?
> 
> In the dead-cube model, the match points (DP,TP,CP,TG) are cubeless game
> winning chances, that's why they are compared to the current cubeless
> (game) winning probability. The effect of gammons and backgammons is
> summarized by gammon and backgammon rates. Keeping G and BG rates constant,
> the DP (for example) tells you what's the minimum percentage of games you
> have to win in order to have a correct double. This keeps into account
> gammon and backgammons via the (current) G and BG rates.
> Also, in cases where gammons/backgamomns are irrelevant (match play), you
> have nothing special to do : the way TP,DP,CP,TG are calculated already
> consider these situations (W and L, the average value of a win or a lose,
> depends on the match score, the MET, the cube value, etc).
> So, to resume, (cubeless) equities are used to compute TP,DP,CP,TG (so they
> keep into account everything, beside cube owning), but they are (cubeless)
> GWC.
> 
> Cube effect is taken into account by the live-cube model, so one possible
> question is if we are allowed to compute the points with the live-cube
> model and then compare the current cubeless (!) GWC to these points.
> But reading Rick Janowski article about the live cube model, I think this
> is ok (http://www.msoworld.com/mindzine/news/classic/bg/cubeformulae.html).
> 
> Anybody able to confirm/refute this ? (Maybe we should ask Rick directly).
> 

Remember, for gammonless money game the dead cube take point is 25%
cubeless gwc and the live cube take point is 25% cubeful gwc! It's the
same formulae that is used to calculate them (i.e., risk versus gain) [*].

The 25% cubeful gwc is identical to 20% cubeless gwc in the continous
model for backgammon. It's within this model getMatchPoints and
getMOneyPoints calculate the live cube points. 

In conclusion, the live cube points calculated by gnubg are by
construction intended to be compared with cubeless gwcs!!! They should
neither be compared to MWCs nor equities!

Jørn




[*] Off-topic: it's precisely the same risk/gain formulae I use watching
"Who wants to be a millionaire" on TV in order to calculate when it's
mathematically correct to guess...

pgpt7Dr_kw8Sw.pgp
Description: PGP signature