For your example, since the probabilities are all multiples of 0.01 you could make a list of 100 elements. Set one of them to a, 5 of them to b, 50 of them to c, etc. Then just randomly sample from the table (which is O(1)). Of course if the probabilities can be arbitrary floating point values then this won't work. But if you can live with rounding to 3 or 4 digits this is probably the fastest and easiest approach.
-- http://mail.python.org/mailman/listinfo/python-list
