We have seen that running a simulation$-$based player, with a fixed simulation policy, from a fixed Go position, gives a Bernoulli experiment with some probability p$.$

Now we change the experiment by adding noise to the simulation policy: at each simulation step, with $10\%$ probability, we select a move uniformly at random. What is the effect on the experiment?