Re: RE: [Bug-gnubg] Removal of absolute fibs ratings

[Christian Anthon]

This is as it should be. Most (all?) players will observe that they
have lower error rates in short matches without so many cube
descisions and gammons not counting in many games. So since a players
error rate depends on the matchlength so does the translation of error
rates to fibs rating.

Christian.

On 8/21/06, address@hidden
<address@hidden> wrote:
> Another funny thing wrt to FIBS ratings : it seems to me that when playing
> 1-pointers the FBS rating is much more severe (compared to the error rate
> based "overall rating") than when playing longer matches.
>
> Example :
>
> 1) 1-pointer
> Error rate (total) :                -0.111
> Error rate (per decision) :         -5
> Equiv. Snowie error rate  :         -2
> Overall rating :                    World class
> Error based abs. FIBS rating :      1890.8
>
> 2) 15-point match
> Error rate (total) :                -0.695
> Error rate (per decision) :         -5
> Equiv. Snowie error rate  :         -2
> Overall rating :                    World class
> Error based abs. FIBS rating :      2004.3
>
> 3) 1-pointer
> Error rate (total) :                -0.256
> Error rate (per decision) :         -11
> Equiv. Snowie error rate  :         -5
> Overall rating :                    Advanced
> Error based abs. FIBS rating :      1668.3
>
> [I've no example of a long match where I played -11/Advanced, but I'm
> sure that in a 5pt match, when you get Advanced you normally get a FIBS
> rating of 1900 or more].
>
> Maybe there's just a bug in the computations, possibly due to absence
> of cube. Another possibility is that Kees' formula somehow incorporates
> a correlation between chequer errors and cube errors, which would make
> the frmula not adapted to 1-pt matches.
>
> Any thoughts ?
>
> MaX.