WinHonte 2.01

ARTIFICIAL INTELLIGENCE

WinHonte uses a combination of several different artificial intelligence (AI) techniques:

Neural nets trained to evaluate:

· shape

· sente/gote (the need for responding to a given play)

· group safety

· territorial potential

Pattern matching:

· joseki (local corner patterns)

· yose (endgame)

Alpha-beta search for local reading problems, including:

· capturing/escaping

· connecting/cutting

· life/death

We have attempted to model the AI of WinHonte from the thought process of human go players, hopefully giving an opponent that is challenging and gives the feeling of playing against a real person. The total effort invested in the program is several thousand hours of programming and testing.