Re: [Bug-gnubg] What's the largest test done between 0 and 1 ply?

[Joern Thyssen]

On Sun, Oct 05, 2003 at 07:13:03PM +0000, Joern Thyssen wrote
> On Sun, Oct 05, 2003 at 02:56:16PM -0400, address@hidden wrote
> > I'm interested in finding out how much stronger 1 ply *checker play* is 
> > versus 0 ply
> > Some small sample work I've done indicates this to be true but wondered if 
> > anyone has done any statistically significant tests.
> 
> Your result below should be statistically significant. 
> 
> > I ran about 50000 cubeless games and 1 ply is about 0.055 ppg stronger 
> > cubeless.
> 
> Based on Zare's post on gammonline the 95% CI should be +/-
> 2.7/sqrt(50000) or +/- 0.12, so 1-ply is clearly better.

Hmm, slight typo that could disturb the meaning of my post: it should be
+/- 0.012 :-)

Jørn

pgppfJL0G4Q13.pgp
Description: PGP signature