Re: [research report] Equation Derived -- Re: On Game Space Size (Monte Carlo)

[Mousheng Xu <xu@xxxxxxxxxx>]

At 02:47 AM 4/30/99 +0300, Antti Huima wrote:
>It is a good exercise to derive a simple equation to match given
>statistical data, especially if the data itself contains many variables.
>However, in the case of the game of go, the equation in itself is not very
>valuable because an exact numerical calculation can be performed. However,

	Exact numerical calculation can be performed, but it's hard or technically
impossible for a 19 x 19. That's why Monte Carlo.

>if you could somehow show that the equation stems from the structure of go
>in a way that does not need to relate to the statistical results then that
>would be cool.

	Sure. But that is so hard.

>100 random samples is way too few! This is manifested in your data.

	The beauty of statistics is that it does not care too much about
individual points. If you have some 300 data points, the R^2 is high (which
indicates 100 samples are enough), risidual is random, it is a good stat.
	Go ahead and plot the data in excel in 3-D (2-D won't be good), rotate in
various directions, you will see a VERY beautiful plot, with most of the
points fall into a geometric plane. There is almost no outliers.

	Thanks.

-- Mousheng