[jorleif]

Probability theory is great, but there are still environments that are unpredictable enough that experimentation is the only alternative. Think of science, for example; If probability theory combined with past experience was enough, there would be no point in doing experiments.

Probability theory can only give you the best estimate from the past. If the future is in any way different from the past, for instance because the past gives insufficient information about the process in question, the prediction will still often be wrong, and you are back to experimentation rather than thinking.