Build Your Own Connect Four AI

Build Your Own Connect Four AI


I just finished the last project of Learn the Basics of Machine Learning course.
The codecademy_evaluate_board() function a bit hard to interpreted.
I had another idea how to create same (or maybe better AI).
My AI vs. CodeCademy AI is win with depth 3/4 tuning too… I am so proud… :blush:
Please take a look to my evaluation function:

def improved_evaluate_board(board):

  board = np.array(board)

  def check_streak(player, row):

    #check triple streak

    double = player * 2

    triple = player * 3

    if  f" {triple}" in row or f"{triple} " in row or f"{player} {double}" in row or f"{double} {player}" in row:

      return 2

    #check double streak

    if  f"  {double}" in row or f" {double} " in row or f"{double}  " in row or f"{player} {player} " in row or f" {player} {player}" in row:

      #print(f"{double} streak found")

      return 1

    return 0

  def get_all_way(arr2d):

    max_col = len(arr2d[0])

    max_row = len(arr2d)

    cols = [[] for _ in range(max_col)]

    rows = [[] for _ in range(max_row)]

    fdiag = [[] for _ in range(max_row + max_col - 1)]

    bdiag = [[] for _ in range(len(fdiag))]

    min_bdiag = -max_row + 1

    for x in range(max_col):

      for y in range(max_row):





    return cols, rows, fdiag, bdiag


  def counter(arr, x_streak, o_streak):

    if len(arr) < 4:

      return x_streak, o_streak

    for row in arr:

      row = "".join(row)

      x_streak += check_streak("X", row)

      o_streak += check_streak("O", row)

    return x_streak, o_streak

  if has_won(board, "X"):

    return float("Inf")

  if has_won(board, "O"):

    return -float("Inf")

  rows, cols, fdiag, bdiag = get_all_way(board)

  x_streak, o_streak = counter(rows, 0, 0)

  x_streak, o_streak = counter(cols, x_streak, o_streak)

  x_streak, o_streak = counter(fdiag, x_streak, o_streak)

  x_streak, o_streak = counter(bdiag, x_streak, o_streak)

  return x_streak - o_streak

The full code you can found here:
Link to Repl
Link to GitHub