[Bug-gnubg] change in calculation of cube decision ratings

[Joern Thyssen]

Hi,

I've tried to implement a "unforced cube decisions" to be used for the
ratings analogous to the unforced moves for chequer play. Hopefully,
tihs should give more realistic cube decision and overall ratings.

For example, in the match sent to the list by Peter Orum, GGraccoon has
114 cube decisions.  With an total error rate of -.460, this gives an
error rate of -.004 per move, i.e., ET play.

However, many of the cube decisions are very easy. In fact, if we count
actual cube actions, close cube decisions, missed doubles, and every
too-good decisions there is only 30, yeilding an error rate of -.015 per
move, i.e., advanced play.

A close cube decision is defined as decisions where

abs(no double - double take) < 0.1

or

abs(no double - double pass) < 0.1.

Typically, this means that you're well into the market window.

Any comments regarding the threshold?

An alternative solution would be to calculate the initial double point
and assume a close cube decision if your winning chances are above the
IDP.

It's also possible to go one step further and divide all chequer plays
and cube decisions into several categories (forced, very easy, easy,
medium, etc. etc.).

Jørn

-- 
Joern Thyssen, PhD
Vendsysselgade 3, 3., DK-9000 Aalborg, Denmark
+45 9813 2791 (private) / +45 2077 2689 (mobile) / +45 9633 7036 (work)