Leverage Research

Each row is one real branch: a strategy type + leverage combination whose parameters were selected using ONLY training-window data, then re-evaluated on unseen validation bars it never touched. "ROI" below is the average VALIDATION return (out-of-sample) — not the training return, which is exactly the number that looked better and turned out to be misleading (see the training-vs-validation gap). A branch only shows Passed if it beat this AND a leverage-matched Monte Carlo random baseline by a real margin — see the Monte Carlo tab.