Computer Go programming

Unlike for chess, there are no "good" computer programs yet to play the game of Go, although its simple rules seem perfect for programming. The effords invested in go-programming are comparable to those of chess, so the reasons have to lie in the game itself. Two points are always mentioned:

• the high branching factor, e.g 361 possible moves at the beginning
• no simple evaluation function

The missing evaluation function is considered the biggest problem, as it makes impossible all brute-force-algorithms used for chess. Because of this, compouter go even fails for smaller boards like 9x9 until now.

Human players use not just logical analysis for playing, but also their creativity, intuition and cognition. Because of this and the fail of traditional algorithms, Go has become an interesting research field in Artificial Intelligence.

Topics to cover:

• Basic philosophies of program design for the game of Go (board game)
  • Knowledge-based systems
  • Machine learning
  • Monte-Carlo methods
  • Genetic algorithms
  • Neural networks
• State representation
• Zobrist hashing
• Pattern matching
• Tactical search
• Problems that arise in computer-computer games
• Life and death
  • Eye spaces
  • Escape and confinement
  • Capturing races
• Global vs Local reading
• Fuseki and opening books

See also

• List of free Go programs
• Computer Go

External link

• computer-go mailing list
• Computer Go and Computer Go Programming pages at Sensei's Library
• senseis: Gnugo
• senseis: GnuGoDevelopment
• senseis: GnuGoDevelopmentImplementationDetails
• The Computer Go Room on the Kiseido Go Server (KGS) for online discussion and running "bots".