The Mind Reader Game


Can you beat the machine?

Mind Reader is an online game played against the computer, which is trying to predict the user's next move.
The algorithm is based on an online learning framework that learns the user's behavior in real time.
The game was developed as part of my 6.883 Online Methods in Machine Learning class project.

Game rules:

  • Each turn, you choose either left or right (use keyboard or the buttons below).
  • The computer tries to predict your choice.
  • If the computer successfully predicts your choice it wins the round, otherwise you win the round.
  • The first to win 100 rounds wins the game.
  • You have 3 seconds for each turn, and the extra time is carried to the next round up to 10 seconds. If you don't make your move in time you lose the round.

This game is motivated by Shannon’s "A Mind-reading(?) Machine"(1953) and Hagelbarger's "SEER, A SEquence Extrapolating Robot"(1956). In fact, some of the strategies that are used by the algorithm here are directly derived from Shannon's and Hagelbarger's work.
The algorithm used here is based on the Expert Setting: multiple strategies (or predictors) predict the user's next move, and a meta algorithm aggregates all the strategies into a single decision.
My class project report can be found here.