Jonathan Kinsey <[EMAIL PROTECTED]> wrote on 01/03/2007 09:51:46:

> Massimiliano Maini wrote:
> > 
> > The method I have in mind will not tell you that the dice was really
> > random (balanced). 
> > 
> > Technically, it's not a test on the statistical hypothesis of each 
face
> > of a dice having probability equal to 1/6. As you said, this would 
take
> > a lot of moves and, by the way, will not help in verifying that the 
on-
> > line site software hasn't overriden the random number generator on a 
> > specific position in a specific match.
> 
> Wouldn't the luck analysis do this?  Assuming a server was setup to
> favour the under dog, I guess you could find a pattern between the luck
> results and the rating differences in quite a small sample of games?

Partially, but the luck analysis is much more complicate. Also, it relies
on 0ply analysis (that could be changed to 2ply, but would be a bit slow)
and it assumes the bot you use (gnubg, of course) has perfect play.
Finally, where to put the threshold above which you can claim somebody has
on purpose manipulated the dice agains you ? 

MaX.
_______________________________________________
Bug-gnubg mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/bug-gnubg

Reply via email to