Re: [gnugo-devel] Levels and time used.

[Gunnar Farnebäck]

Evan wrote:
> That I definitely agree with. One interesting place to start would
> be investigating why some regression tests FAIL on higher levels.

Without looking into it I don't find it surprising that it happens.
It's a well known fact that GNU Go sometimes makes multiple mistakes
which effectively cancel each other. Fixing one of the mistakes (e.g.
by increasing the level) may leave some other mistake uncancelled. Of
course there might be systematic errors introduced at higher levels as
well but it wouldn't surprise me if an investigation would come up
with lots of accidental failures.

/Gunnar