Psuedo code for a Bayesian Bandit implementation. Bandit is a class.

psuedo_code_bayesian_bandit.cpp

epsilon = .1;
n_trials = 1000;
counter = 0;

function test() {
  //pick a random number between 1 and 10, inclusive
  random_number = rand(1, 10); 
  //if the random number is within the epsilon percentage
  if (random_number <= bandit.epsilon * 10) {
    //simply picks a bandit at random
    showBandit( getRandomBandit() ); 
  } else {
    //picks the best performing bandit
    showBandit( getBestPerformingBandit() ); 
  }
  
  //increment the counter
  counter++;
  
  //indicate somehow that the test has ended if n_trials has been reached
  if (counter == n_trials) {
    end;
  }
}

function showBandit( bandit ) {
  //runs the bandit and collects the success metric 
  //(converted / didn't convert, etc.)
  outcome = run_trial( bandit );

  //add the outcome to an array containing the trial results
  bandit.trials.push( outcome );
  
  //update the bandit's success metric as simply the sum of trials 
  //(since success = 1)
  bandit.successes = sum( bandit.trials );
  
  //the success rate, or probability for success, 
  //is updated with the new information from the most recent outcome
  bandit.success_rate = bandit.successes / bandit.trials.length; 
  
  //update the posterior as a beta distribution
  //assumes the first prior for each bandit is (1, 1)
  a = 1 + bandit.successes - 1;
  b = 1 + bandit.trials.length - bandit.successes - 1;
  bandit.distribution = beta( a, b );
  
}