Prediction of Parallel Speed-ups for Las Vegas Algorithms