Steps.txt

---- Steps ----
1. Construct a representation of the board
2. Write a script that returns all possible moves in vector space - Possible Move Generator (PMG)
3. Write a neural network in TensorFlow that evaluates each move and returns a score from [-1, +1]
4. Write a script that takes the board representation and passes it to the PMG, passes all of the moves through the NN,
    chooses the best move and passes it back to the board representation, doing this for n moves
5. Train the NN!

---- Step specs ----
1. Board representation:
  - 8x8 board with a maximum of 16 pieces
  - Be able to take as input a move by one of the players and update the board state
  - Pieces should be captured and removed when appropriate
  - Capability to return some statement of winning/losing either by all pieces being eliminated or 
    whoever has the most pieces
  - Should store every move for a game from each player
2. Possible moves in vector space:
  - Each spot on the board is represented by (1,0), (0,0), or (0,1)
  - (1,0) represents that your player is present on that spot
  - (0,0) represents that no player is present on a spot
  - (0,1) represents that an opponents player is on a spot
  - A single move is represented by a 64x4 matrix where the first two columns are the 
    current state of the board and the next two are the state of the board after a move.
3. What the neural network does:
  - The neural network should take the 64x4 matrix that represents a move as input
  - Should output a score from [-1, +1]
4. Game playing script
  - Passing board state to PMG should take the 64x2 matrix of the board state and pass it to the PMG
  - PMG then returns all possible moves to the NN
  - NN does a foreward pass on each move and passes the maximum value back to the board state (storing each value for later?)
  - Loop over above three steps for n moves
5. Training the NN to play
  - After a completed game (n number of moves), every move that contributed to the winning side is assigned a response value     of +1 and the losing side is assigned a value of -1
  - Update the NN by comparing the output score to the +1 or -1 value assigned to it.