Print this pageThe Battle against Bayesian Amnesia
“The chance of success (COS) on this prospect is 20%, but if we shoot 3-D seismic, it will go up to about 40%. The COS on that prospect is also 20%. If we see an AVO anomaly, it will go up to about 50%; if we don’t, it will stay at 20%.” Comments like these are common in the upstream energy industry, even among people who have experience in value-of-information (VOI) analysis. Unfortunately, the first comment is extremely misleading and the second is mathematically impossible. There are certain laws of probability that simply must be obeyed.
When acquiring seismic or interpreting a direct hydrocarbon indicator (DHI), many companies completely forget about their previous assessments of the COS for a prospect. This results in over-estimating COS and, therefore, under-performing relative to prediction. The confusion is exacerbated when the new information is expected to yield insights on several different uncertainties and/or the company plans to acquire more than one type of information to help with a single uncertainty.
This paper discusses several examples of scenarios like these, and how some basic (and not-so-basic) Bayesian decision trees can help people to see how to revise their uncertainty assessments appropriately in the wake of new data.
AAPG Search and Discovery Article #90090©2009 AAPG Annual Convention and Exhibition, Denver, Colorado, June 7-10, 2009