Re: [Bug-gnubg] Re: Strange FIBS ratings

[kvandoel]

On Mon, 8 Sep 2003, Jim Segrave wrote:

> But if we look at the full figures Albert quoted:
>
> Overall statistics
> Error rate (total)              -2.769 (-29.273%)    -1.134 (-14.183%)
> Error rate (per decision)       -51.28 ( -0.542%)    -20.25 ( -0.253%)
> Equiv. Snowie error rate        -26.37               -10.80
>
> Overall rating                  Awful!               Casual player
>
> Actual result                    +50.00%              -50.00%
> Luck adjusted result             -11.78%              +11.78%
> Luck based FIBS rating diff.    -186.55
> Error based abs. FIBS rating    1465.2               1704.0
> Chequerplay errors rating loss   353.5                344.7
> Cube errors rating loss          231.4                  1.3
>
> The left hand column has an overall error rate of .020
> The Cube errors rating loss is 1.3, chequer play is 344.7, which
> suggests very strongly that cube errors were not a factor here, and
> that the overall error rate comes from chequer errors. The Fibs rating
> is 1700.
>
> Colour me surprised.

One of the  "surprizing" results of my simulations  was that cube errors
have a much smaller effect on rating than previously assumed.

[7pt match]
> Chequerplay statistics
> Error rate (per move)           +0.000                  -23.08 ( -0.332%)
> Cube statistics
> Error rate (per cube decision)  +0.000                  -12.83 ( -0.147%)
> Cube decision rating            Supernatural            Intermediate
> Error based abs. FIBS rating    2050.0                  1752.7
> Chequerplay errors rating loss     0.0                   287.2
> Cube errors rating loss            0.0                    10.1

> Now you're telling me that an average chequer play error of .023, an average
> cube error of .13 an overall error rate of .020 is 1750 level play on FIBs?

12.83/1000 = .013, not .13.


Kees