Re: [gnugo-devel] GNU Go 3.3/3.5

[Gunnar Farneback]

Arend wrote:
>  - Spend a lot of tuning effort on computing the safety value of a group.
>    The result should not be expressed in a finite number of steps (WEAK,
>    ALIVE, STRONGLY_ALIVE...) but directly as a floating point number. This
>    should make it possible to tune the valuation better than currently;
>    e.g. IMHO one of the important reasons why GNU Go fails to recognize
>    weak groups is that a group with high escape value is at least ALIVE
>    regardless of how small the moyo is. This means e.g. in the fuseki that
>    the engine does not see the need to extend a stone. Such things should
>    be more easily tunable with a continuos set of safety status.

This is not as obvious as it may sound. While it's probably true that
a continuous safety measure allows more precise valuation, it's also
more sensitive to errors in the safety evaluation. I tried with a
continuous approach early in the development of the current move
generation scheme, but had to abandon it when it turned out to be
insufficiently robust. Today the surrounding framework is of course
much stronger, so the chances to succeed with such a scheme are much
better than when I tried it. Still one should be aware that the tuning
may be very delicate.

/Gunnar