[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

computer-go: Extended Zorbrist function



I have looked at the Zorbrist function
and wonder why he choose this way of calculation

To me it is clear that this is an easy way of
calculating, but not the correct way.

Stones do not radiate influence in all directions equal.
The influence in horizontal and vertical direction are
much stronger.

So my porposal is to extend the functions into something like this:

For every point there is 4 values
- a up value
- a down value
- a right value
- and a left value.

influence are calculated in 2 steps:

Step 1.
- 3 equal values in a row are used to calculate the new
value of the middle.
example: 3 right values
+---+---+---+
|0.3| 0 |0.3|
+---+---+---+

Sum the values and divede by 3 : (0.3 + 0 + 0.3)/3=0.2
+---+---+---+
|0.3|0.2|0.3|
+---+---+---+

If a value is 1 or -1 for black and white it stays 1 or -1.

Step 2.
- Calculate the sideways influence.
- in 1 cell are the up, down, right and left values added and
diveded by a "sideway influence number" between 2 and 10000,
depending on how strong the sideways influence has to be.
- the new value is added to the old value.

Now repeat this for as many iteration you want.

Kjeld