Re: [gnugo-devel] reinstating value(100) owl defense moves?

[Evan Daniel]

On Wed, 8 Sep 2004 13:02:00 +0200 (CEST), Arend Bayer
<address@hidden> wrote:
> 
> 
> Evan Daniel wrote:
> 
> > On Thu, 2 Sep 2004 00:00:34 +0200 (CEST), Arend Bayer wrote:
> >
> > > Maybe you could try out how much your handful of patterns saves when
> > > replaying "real life" games. Then we may have a guess how much saving
> > > one could get with, say, some 50 patterns.
> >
> > Over the complete regressions, my small handful of patterns (a few L
> > group patterns + a few tripod group; neith is done) saved 0.2% owl
> > nodes.  Not terribly impressive.  However, it did save a significant
> > percentage on time -- over 10%.  I assume this is an artifact of
> > something; maybe dfa related?  I haven't yet investigated it, though I
> > should have time in the next day or so.
> 
> I strongly suspect that something went wrong with your timing tests. Was
> your machine idle during both runs?

That's entirely possible.  My machine was almost but not completely
idle during that time.

> If it only saves the 0.2% then I have to admit I remain a little
> skeptical. 1% of the owl nodes in the regressions come from ld_owl.tst,
> so the nodes saved outside of that might be negligeable?

It seems that way.  I don't have the detailed numbers any more, though.

Evan Daniel