[Computer-go] Oh no, dynamic Komi again
teytaud at lri.fr
Tue Jun 21 01:16:20 PDT 2011
> Hideki: « # Using some feedback mechanism to keep WR around 50% (used in
> FudoGo and pachi?) is an another issue. »
I can have positive results with rules as in Pachi and MFOG (linearly
decreasing dynamic komi), but not yet with rules aimed at preserving a WR
around 50% (e.g. the 42% rule). If someone has published a paper with
detailed implementation I'd like to read this. This idea looks appealing,
and in some cases it worked very well, but not as stable (I try to have
something compliant with various board sizes and various handicap levels).
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Computer-go