Interpretting Walk forward testing results?
Posted: May 15 2011
Lets says you run a walk forward test, over 1 year time frame, using 60 days in sample and 20 days out of sample data. The test yields 6 time series of results, with vary levels of performance. Is the last series of results, the only one(or most) relevant, since these results (and inputs) are the closest to the current market conditions.
There maybe other series, where the OOS results were better, than the last one, but are they still considered as relevant, for future performance?
There maybe other series, where the OOS results were better, than the last one, but are they still considered as relevant, for future performance?