RE: [Bug-gnubg] Re: The importance of METs

[Ian Shaw]

address@hidden wrote on Monday, September 08, 2003 9:40 AM

> Now that  the external player  bugs are dead  I tried playing 
>  some 17pt
> matches expert-expert using different met's.
> 
> I  think the  results show  that the  differences between  
> the  mets are
> completely negligible in terms of equivalent rating strength points.
> 
This is very interesting to me, since I've been doing rollouts with a
view to establishing a MET for gnubg at 2-ply play. It takes a long
time, so if I'm wasting my CPU cycles I'd like to know!

I would expect METs to make most difference in shorter matches, over 3,
5 or possibly 7 points. At higher scores I would expect most METs to be
very similar. I would guess a lot of your 17 point matches didn't come
down to the 5-point match stage, so any effect at these scores will have
been diluted.

Can you extract from your data the records when the scores got into 3, 5
or 7-point match range?

It might be useful to do a test of 5-point matches of Woolsey vs.
Snowie, which are probably the most commonly used METs. Or find the two
which diverge the most and compare those, just to see if there is any
noticeable difference.

--
ian