Roadmap

This is just a place for me to mess around with python, ML, and code performance evaluation right now the rough road map is:

• Get a decent engine with stock python. No numpy, I want to really see what I can push out of stock python:

  • DONE Prototype to get a good hold on the problem
  • DONE Get some base-line performance numbers for for future performance testing
  • DONE Datastructure Definition
    • DONE What the list represents
    • DONE How the graph will be tracked
  • DONE Skeleton suite of unit tests from small_board and board tests
  • DONE Fill-in test-suites and fucntionality
  • DONE Basic engine should be able to at least generate 100k states per second before proceeding to bot building (This code is now able to generate 220k moves/s and visit 7.5k states/s)

• Train an extremely basic chess bot:

  • Prototype a game-playing framework, using basic strategies inspired and copied by work done by tom7: http://tom7.org/chess/weak.pdf
    • random_move
    • same_color
    • swarm
    • generous
    • cccp
    • suicide_king
    • min_oppt_moves
    • pacifist
  • Build an extremely simple strategy that has a score look-up for each square-state combination
  • Play this strategy against itself to perform a gradient descent
  • Create plugins for other engines to add to the evaluation

• Construct addtiional NN-based architecture strategies.