Out-of-sample (OOS)
In short
Out-of-sample testing validates a strategy on data left untouched during development. Only OOS results say anything about real future viability.
The standard: develop rules on one period, validate on a later untouched one. Our lab ticker shows live how many signal variants currently survive this hurdle — usually a small minority.
Why do so few strategies survive OOS testing?
Because most "patterns" are chance plus overfitting. That is exactly what the test is for.
Related terms
Information, not investment advice. More of this? The newsletter.