#50 — Price momentum M15+M5 live

Hypothesis — what it tests

Whether unsmoothed multi-lag price momentum across two timeframes (M15+M5) carries a deployable edge that beats single-timeframe SMA momentum, and whether removing the smoothing layer shortens hold times.

Description

The question was whether raw price momentum — simple close-price deltas over multiple lags, with no SMA smoothing — could match or beat the deployed SMA16 momentum strategy and shorten hold times. A sweep of 840 configs (3 timeframe pairs x 56 lag triplets x 5 TP levels x 12 pairs) using a strict 6/6 lag-agreement signal found 4 configs beating the SMA16 baseline; the M15+M5 config with lags (1,3,8) and TP=10p was the only one to pass IS3 walk-forward on all 12 pairs, and cleared 2000-sign-shuffle Monte Carlo at mc_p=0.0000 (+30.4 p/d OOS). It was deployed as fx-price-mom-live on account 011 (initially 50u, cut to 25u after margin pressure). The smoothing removal did NOT shorten holds (mean hold floor ~30h is structural), and the whole TP-only/no-SL momentum book — this strategy included — was later refuted by the finite-margin re-validation (entry 57): a closeout simulation drove the account NAV to −$15.66 on the $17.79 research scale, and the book was drained in June 2026.

Key result

Indicators

Algorithms

Code