An implementation plan

If I'm going to write a computer go player, I need a plan. Plans drawn at this stage are inevitably only approximations, but here's my current plan:

• Infrastructure – the framework to hang the rest on
  • SGF parser
  • SGF writer (simpler that the parser since it only needs to handle the properties I'm actually using)
  • a graphic goban (so i can play the program)
  • GTP client (so I can play my program against other people and other programs)
  • GTP engine (so i can organise play-offs)
  • test suite (so ease debugging)
• Opening: a classical opening book, assembled from pro and senior-amateur games. Also code to handle symmetry.
• Mid-game: a genetic algorithm prioritising a list of relatively complex template patterns.
• End-game: a theorem prover, perhaps along the lines of gotools, but with incremental additions of knowledge to the database as positions are proven safe/unsafe.

One advantage of have separate beginning, middle and end games is that I can develop, test and debug them separately.