Comparing probabilistic predictions with outcomes in petroleum exploration prospect appraisal

In the oil industry, uncertainty about hydrocarbon volumes in undrilled prospects generally is expressed as an expectation curve. The curve indicates the probability of exceeding a given amount.After drilling a number of prospects conclusively, that is, we know the amount of reserves in the targets, if any, the question arises about the validity of the prediction. Since the prediction was in the form of a probability distribution, the comparison with a single actual outcome of the process is not straightforward.I propose a specific combination of mainly well-known tests that can be applied in this hindsight analysis to address the following: (1) the measure of location or expectation, (2) the probability of success (3) the shape of the distribution of the nonzero outcomes or success cases, and (4) a measure of rank correlation between predictions and outcomes. Even small numbers of drilled structures may suffice for obtaining conclusive results. Such statistical analysis provides useful feedback for those concerned with the maintenance and control of the prediction system.