Validation method

In-Sample and Out-of-Sample Testing

IS and OOS

In-Sample Testing is where parameter choice and scoring happen, on the research sample. Out-of-Sample Testing reads signal edge and attribution on an expanding window.

What these checks look at

In-Sample Testing stays with the research sample. Parameters are chosen there, and the score belongs to that sample.

Out-of-Sample Testing has two readings: signal edge, and attribution. Both use an expanding window. The window grows with the price path. It is not a last slice held back.

Why the window expands

Parameter choice needs a sample that is allowed to teach the score. Signal edge and attribution need a window that keeps lengthening, so the reading is not a single cut of the chart. The research sample remains the place where the choice is made.

How Alphavo runs them

Parameter choice and scoring run on the research sample. Signal edge and attribution run on the expanding window. A Candidate Strategy is still a candidate while these checks run. A candidate is not Alpha.

How to read the result

Read the research-sample score as the result of the choice made there. Read signal edge and attribution from the expanding window. When a strategy is shown, the star is 0 stars, 1 star, 2 stars, or 3 stars.

Limits

• The expanding window is not a last slice held back.
• A score on the research sample does not turn a Candidate Strategy into Alpha.