A practical guide to the use of success versus failure statistics in the estimation of prospect risk
A practical guide to the use of success versus failure statistics in the estimation of prospect risk
Statistical data documenting past exploration success and failure can be used to inform the estimate of future chance of success, but this is not appropriate to every situation. Even where appropriate, past frequency is not numerically equivalent to future expectation unless the sample size is very large.
Using the Rule of Succession of Laplace (1774), we calculate the appropriate predicted chance of future success that can be used for smaller sample numbers, typical of exploration data sets, which include both successes and failures. The results, presented as a simple look-up table, show that the error which would result from using simple frequency instead of the appropriately calculated value is particularly severe for small samples (>10% error arising if n< 9).
This error is least if past success rate is close to 0.5 but it increases markedly if the past data consist of mostly failure or mostly success.
We review the conditions in which past frequency can be used as a guide, and the circumstances in which it does not reflect future chance. Past success frequency should only be used as a guide to future chance if the past tests and future opportunities belong to the same play, and are similar as far as the available data allow. They should not be used if the historical tests have selectively sampled the “cream” of the pool of opportunities.
137-150
Peel, Frank J.
ccbb86f8-56a0-4b59-b664-e772a9c4015f
Brooks, John R.V.
9138247b-d5b8-4eba-adda-a4f087d6a31c
February 2016
Peel, Frank J.
ccbb86f8-56a0-4b59-b664-e772a9c4015f
Brooks, John R.V.
9138247b-d5b8-4eba-adda-a4f087d6a31c
Peel, Frank J. and Brooks, John R.V.
(2016)
A practical guide to the use of success versus failure statistics in the estimation of prospect risk.
AAPG Bulletin, 100 (2), .
(doi:10.1306/08101515059).
Abstract
Statistical data documenting past exploration success and failure can be used to inform the estimate of future chance of success, but this is not appropriate to every situation. Even where appropriate, past frequency is not numerically equivalent to future expectation unless the sample size is very large.
Using the Rule of Succession of Laplace (1774), we calculate the appropriate predicted chance of future success that can be used for smaller sample numbers, typical of exploration data sets, which include both successes and failures. The results, presented as a simple look-up table, show that the error which would result from using simple frequency instead of the appropriately calculated value is particularly severe for small samples (>10% error arising if n< 9).
This error is least if past success rate is close to 0.5 but it increases markedly if the past data consist of mostly failure or mostly success.
We review the conditions in which past frequency can be used as a guide, and the circumstances in which it does not reflect future chance. Past success frequency should only be used as a guide to future chance if the past tests and future opportunities belong to the same play, and are similar as far as the available data allow. They should not be used if the historical tests have selectively sampled the “cream” of the pool of opportunities.
Text
A practical guide to the use of successfailure statistics in the estimation of prospect risk Article text with bitmap pictures.docx
- Accepted Manuscript
Text
137.full.pdf
- Version of Record
More information
Accepted/In Press date: July 2015
Published date: February 2016
Organisations:
Marine Geoscience
Identifiers
Local EPrints ID: 378951
URI: http://eprints.soton.ac.uk/id/eprint/378951
ISSN: 0149-1423
PURE UUID: 69987d36-7919-4cd4-bd3c-999fd4b18e0d
Catalogue record
Date deposited: 10 Jul 2015 10:55
Last modified: 14 Mar 2024 20:31
Export record
Altmetrics
Contributors
Author:
Frank J. Peel
Author:
John R.V. Brooks
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics