Data Snooping is often suspected when effective technical analysis rules are found or presented. It is difficult to tell if a result is due to data snooping, so evaluating technical analysis rules often boils down to detecting data snooping and if it has invalidated the results. Herein we look at several algorithms designed to increase (risk–adjusted) returns for investors, and several techniques for detecting or compensating for data snooping.
We find no easy answer to detecting data snooping. Many of the methods we look at are useful, but there is no known way to get around sparse data and the unrepeatable nature of investment decisions. We conclude that data snooping bias is a persistent risk and it is unlikely that there is any effective single solution to the problem. The best that we can do is be aware of the risk of data snooping and to report how we have dealt with the risk as part of our analysis.
