Once, back in the high school, my friend introduced me an extended version of Tic-Tac-Toe game. I was very interested, since the basic version is trivial and boring. The key idea is that each cell of the # is replaced with yet another #, which acts like normal Tic-Tac-Toe game. Rules for making a move are following:

1. Each player draws a circle O or X in one of nine nested # (see picture below),

2. Choice of nested # by the player is enforced by the last opponent move in a such way:

   A player must draw character inside the nested # which is located in the same cell in an outer #, 
   than the cell which was marked by the opponent in the previous move in a nested #. 
   For example, if the last move was made inside outer cell NORTH-WEST, and SOUTH-EAST 
   in inner cell then the next move have to be made inside SOUTH-EAST outer cell.
   

3. If a nested # has been already decided (game in the # ended), player can choose any nested # to make a move.

4. In the first move a player can choose any nested #.

In this project you can find implementation of the game, which I generalized to more than two levels, i.e. each nested # can contain another # etc. For that reason, I called the game Fractal Tic-Tac-Toe.

Feel free to play MCTS agent on as many levels as you wish. Below you can see part of a match between two MCTS players.

An example of a board state in three levels game.