Due to the high number of branches (500 times more than chess), it is impossible to explore every possibility.

Using probabilities, random simulations and backtracking, the

Monte Carlo Tree Search Algorithm provides a way to decrease the number of possibilities to be explored.

Each core is working on its own part of the tree, making the search up to 8 times faster !

For even more processing power, the algorithm can distribute its work among several machines. Using a Master-Worker strategy, each computer added to the cluster increases the odds of finding an advantageous move.

As each worker has its own copy of the tree, there is but few communications between the machines, thus decreasing the amount of inteference and speeding up the search.

As each worker has its own copy of the tree, there is but few communications between the machines, thus decreasing the amount of inteference and speeding up the search.

Prateek Bhatnagar - Baptiste Bignon - Mikaïl Demirdelen

Gabriel Prevosto - Dan Seeruttun--Marie - Benoît Viguier

Supervisors: Nikolaos Parlavantzas & Christian Raymond

Gabriel Prevosto - Dan Seeruttun--Marie - Benoît Viguier

Supervisors: Nikolaos Parlavantzas & Christian Raymond