Re: [Bug-gnubg] Missed doubles problem ?

[Holger]

At 18:18 04.09.2003 +0000, Joern Thyssen wrote:
> On Thu, Sep 04, 2003 at 07:51:08PM +0200, Holger wrote
>
> > I think the doubling window should run from DP to TGP.
>
> True, but consider this simple example:
>
> DP=50%
> CP=70%
> TG=90%
>
> Using the average of DP and TG as originally suggested we would end up
> using 70% as the divider -- that doesn't seem correct. I would call a
> missed double at 71% a "missed double around the DP".
>
> My new suggestion puts the divider at 80% which seems more correct in my
> simple example above.

I'm not sure if I like this. This seems to call for the next user question.

> Alternatively, we can make three dividers:
>
> Missed doubles around DP: 0% - 60%
> Missed doubles around CP: 60% - 80%
> Missed doubles around TG: 80%-100%
>
> Perhaps that makes more sense.

This I like much more. If you don't mind I'll implement this.

> > If I understood
> > it well, those points can vary according to cube availability. I'll
> > have to look how this is determined. Maybe I can always use
> > aaarPointsMatch[ player ][ point_type ][ 1 ].

Btw, do you know why for the window graph the TGP for dead cube is taken? 
Or does it combine live and dead cube?

E.g. in one game I have the following points at 3-away 3-away, centered cube:
        dead            live
TP      35,703% 29,501%
DP      50,477% 50,477%
CP      66,028% 71,384%
TGP     88,047% 71,384%
The window graph starts (green) at 50,477%, turns red at 71,384% and blue 
at 88,047%.

> > And the other question: right now I'm comparing the middle of the doubling
> > window to the no-double equity. I'm not sure whether this is correct.
>
> No, you should compare the value with aarOutput[ 0 ][ OUTPUT_WIN ].

Just to confirm whether I understood: aarOutput[ 0 ][ OUTPUT_WIN ] should 
equal arDouble[ OUTPUT_OPTIMAL ] ?

Regards,

        Holger