Re: [Bug-gnubg] [Q] EvaluatePosition(), Utility(), and testing boards

[Jim Segrave]

On Tue 07 Oct 2003 (15:44 -0600), Thomas Hauk wrote:
> I'll just ask the questions outright and see how close what I think the 
> answer is to the real answer. 8)
> 
> What does EvaluatePosition() with a "cubeless 0-ply" evalcontext return?
> What is the range of values the return value can take?

In the strictest sens, 0/-1 are the return values - indicating error
no error. However the array arOutput which is passed as a parameter
will have the elements filled in:

the subscripts are defined in eval.h:

#define OUTPUT_WIN 0
#define OUTPUT_WINGAMMON 1
#define OUTPUT_WINBACKGAMMON 2
#define OUTPUT_LOSEGAMMON 3
#define OUTPUT_LOSEBACKGAMMON 4

Values should be floats betwen 0.0 and 1.0, as they are the
probabilities of the various outcomes for the player on roll.

> What does Utility() return? What range of values can the return value 
> take?

a float representing the cubeless equity after adjusting for the
gammon value (which is less than helpful, because that's exactly what
the comment above the routine says). As far as I know, gammon prices
lie in the range 0.0 to 1.0, and the adjusted equity returned will lie
in the range -1.0 to 1.0
 
> Does anyone know of an existing test bed of board positions used for 
> computer backgammon programs? Perhaps you guys have a gnubg(tm) test bed 
> already? If you don't know of one, any idea where I could find one?

I think Joseph Heled has a set of test positions used for training and
testing the net. Otherwise, no, I don't know of any benchmarks sets.

-- 
Jim Segrave           address@hidden