We study the value of stochastic predictions in online optimal control with random disturbances.