Re: Topics on influence

[Compgo123@xxxxxxx]

Representing influence at each position by a number may have some valid point.
Couple programs using this approach obatined quite impressive results. I
thumbed through the book 'Mathematical Go Endgames' sometime ago. The main
idea of the book is that the best score in end games can be calculated from
topological properties of the position, no looking ahead is necessary. When
one assign a number to a point, one is calculating certain topological
properties of the whole position. Thus, it may be possible that the best score
concerning a given position can be calculated directly from this property.
However, this topological property may not be as simple as those used by
different programmers so far. One example is in a ladder capturing situation.
Here a sotne far away is more important (to break the ladder or not) than
stones more near by which are not on the ladder path. One possiblility is that
the best score is calculated from the topological properties of the influence
field. Maybe this is the place Henrik Rydberg's differential equations come
into play. Another approach maybe a discret one. Represent the whole board by
a 19x19 matrix. It's elements are determined by the board position. Certain
mathemtical manipulation of the matrix will give the best score.


Dan Liu