Re: [Bug-gnubg] FIBS rating and error rates

[kvandoel]

On Wed, 20 Aug 2003 address@hidden wrote:

> > After I include it it's probably better if you can make further changes
> > directly by modifying the texi source.
>
> Can you wait  a few days with that  so I can finalize it  there are some
> small things to  update like typo's? I'll email you when  I am done with
> revisions.

Hi Joern,

I'm currently running simulations trying  to find a bilinear fit to move
and cube errors  separately. It works very well  and really improves the
estimates. So far  I've finished collecting data for 3  and 5 pt matches
and will continue.

When done I'll do a writeup and  if you agree with the results I suggest
implementing the  new formula.  It  will improve the estimate  for games
where  the cube  handling is  really  realy bad,  otherwise the  current
approximation is fine.

I estimate about 150 hrs of cpu time (gimme a week) to finish this.


Kees