RE: [gnugo-devel] More on minimum values

[Portela Fernand]

Evan wrote:

> The equivalent calculation on your data:
> 
> sd = sqrt(400 * .5 * .5) = 10
> 
> 229 - 200 = 29
> 
> 29 = 2.9 * sd

I wish my patch would have got 229 out of 400, but you misread the
results of serie 3. Black (unpatched) won it (and big). So it's
globally 203 out of 400.

Also, one of your assumptions is that GG is exactly as strong with
white stones as with black stones (with a 5.5 komi). Dan already
mentioned that this is not the case (he mentioned a komi of 8) and I'm
convinced current engine is way stronger with black stones when playing
against itself (I think the delta is actually greater than a komi of
8 point). My patch was actually an attempt at finding out the reason(s)
of this discrepancy as well as a possible mean for correcting it.

Also I would like to invite you to make the calculations on each serie
separately, and also globally. What conclusions do you make out of
them ? Are all these p-values consistent ?

Taking sd's into account, my conclusions concerning my patch are
following :
- in the average, it has possibly improved the playing strength
  by about 2~3 more games won out of 100.
- the most optimistic view could be an improvement of 7
- the worst case is a possible degradation of 1~2

Since the worst case is still negative, I'm rejecting my patch in its
current form and trying something else.

/nando