[Bug-gnubg] [bugs #8462] MET G1 parameter

[Joern Thyssen]

Hi Ian,

> I think the original structure and definition of the variables was ok.
> That is, the odd 1a2a entry of 71.25% was a direct result of the 15%
> gammon rate entered for the trailer. 
> Working with the original Zadeh program and definition of G1 I was able
> to create a credible looking MET just by altering "gammon-rate-trailer"
> to a more sensible figure in the 25%-28% range.

I'm still not convinced that I should change anything in the algorithms.
Note that leader and trailer refers to the leader and trailer in the
individual game, e.g., the doubler is assumed win 25% gammons and the
doublee 15%. I think this is a reasonable assumption.

> Is Zadeh's original article avaialable online anywhere? I've been unable
> to access it with a Google search.

I don't think you can find Zadeh's article online. You have to go to
your local university library and make a photo copy.

> As an aside, are there any other MET generating programs available
> that will allow me to experiment? I was wondering if Mec26 was
> available in a format suitable for this. 
> I'm doing a project that
> involves Gnubg rollouts and theoretical MET generation. Playing with
> some source code (or at least the parameters for the code) would be
> very useful to me. 

Yes, I've recently added it to gnubg. In the newest installation
archives there should be a mec.xml you can modify.

Alternatively, you can download mec.c from the gnubg distribution
(http://savannah.gnu.org/cgi-bin/viewcvs/gnubg/gnubg/mec.c?rev=1.2&sortby=date&content-type=text/vnd.viewcvs-markup)
 and compile it yourself with

gcc -DMEC_STANDALONE mec.c -o mec

mec.exe 19 0.26 0.50

should give you mec26 (up to 19 points).

Joern