Play and learn Tic-Tac-Toe Game on R

Install from CRAN

install.packages("tictactoe")

Or you may install the recent development version from github

devtools::install_github("kota7/tictactoe")

To use,

library(tictactoe)

You can play tic-tac-toe on R console.

ttt(ttt_human(), ttt_ai())

This would give you a prompt as below.

    A B C
   ------
 1| . . .
 2| . . .
 3| . . .

 Player 1 (no name) to play
choose move (e.g. A1) > 

Type a move, then the oppoenet will respond. To finish the game in the middle, type "exit".

The default AI player is very week (in fact, he plays randomly). To play against a more sophisticated player, set the level argument (from 0 (weekest) to 5 (strongest)).

ttt(ttt_human(), ttt_ai(level = 4))

You may play as the second mover by ttt(ttt_ai(), ttt_human()). You may watch games between AI players by ttt(ttt_ai(), ttt_ai()).

To conduct a large scale simulation between AI players, use ttt_simulate function. The code below conducts 100 simulation games between random AIs. The result 0, 1, and 2 indicate draw, won by player 1, and won by player 2 respectively.

res <- ttt_simulate(ttt_ai(), ttt_ai(), N = 100, verbose = FALSE)
prop.table(table(res))
#> res
#>    0    1    2 
#> 0.13 0.57 0.30

Q-learning is implemented to train AI players. The code below trains a random AI through Q-learninig of 500 episodes.

p <- ttt_ai()
o <- ttt_qlearn(p, N = 500, verbose = FALSE)

Now this player is much stronger than the random player.

res <- ttt_simulate(ttt_ai(), p, N = 100, verbose = FALSE)
prop.table(table(res))
#> res
#>    0    1    2 
#> 0.15 0.25 0.60

• Sutton, Richard S and Barto, Andrew G. Reinforcement Learning: An Introduction. The MIT Press (1998)