#33 — TR Momentum strategy

Hypothesis — what it tests

Whether large True-Range M5 bars continue in their direction (momentum), harvestable with a tight trailing stop — and, in the post-mortem, whether the validated edge survives a realistic fill model.

Description

TR Momentum entered at M5 bar close whenever the bar's True Range exceeded a threshold — the intuition being that large bars continue in their direction — and exited on a tight 2-pip trailing stop from the high-water mark, with no fixed take-profit. The v2 sweep was a model validation exercise: IS-only P90 spread gate, walk-forward with all three IS chunks positive, OOS>0, and a 300-shuffle Monte Carlo sign test — and all 12 pairs passed everything (USD_JPY +94.4 p/d OOS, GBP_JPY +85.1, all mc_p=0.0000). It went live on account 011 and was killed within a day when AUD_USD showed 0/6 wins. Root cause: the backtest filled trail exits at the exact trail price (a stop-order assumption), while the live service submitted market orders at the next M5 bar close — a 3-7 pip gap per trade. With bar-close fills the sweep collapsed to 0/12 walk-forward survivors: the edge was 100% fill model. The follow-up diagnostic showed momentum and fade directions perform identically (EV ≈ −spread), i.e. zero directional signal existed at all.

Key result

Indicators

Algorithms

Code