n independent classifiers each have accuracy p, with n odd. Return the accuracy of their majority vote.

Sum the binomial tail from $\lfloor n/2 \rfloor + 1$ upwards. Above $p = 0.5$ the vote beats any single model and improves with n; below it, the ensemble is worse than one model, because the errors reinforce.

Input

n = 5
p = 0.6

Output

0.68256