#65 — Majors lead crosses (lead-lag)

Hypothesis — what it tests

Whether USD-major pairs lead their crosses one-way, and whether the lagging cross's catch-up move is large and slow enough to trade at retail.

Description

A user research question: do the USD majors lead the crosses (minors), is the lead one-way, and is the lag tradeable? Tested in triangular form — each cross expressed as the product of two majors (EUR_JPY ≈ EUR_USD + USD_JPY, EUR_GBP ≈ EUR_USD − GBP_USD, etc.) — on 2.36 million aligned 5-second bars. The structural hypothesis was confirmed emphatically: majors lead crosses one-way with major→cross IC +0.07..+0.12 (t = +109..+189) versus a weaker cross→major IC, positive asymmetry on all 5 triplets, and contemporaneous R² of 0.81–0.94. But it is un-monetizable at retail: the cross catches up within a single 5s bar, the predictable move (~0.1 pip = IC×σ) is 25–40x smaller than the 2–4 pip cross spread, and it does not accumulate, so every catch-up trade nets about minus one spread (EUR_JPY −2.5p, GBP_JPY −4.0p; win rate 0.1–1.6% IS+OOS at all horizons). An HFT-only edge — sub-second, raw ECN spread, colocation — and another instance of the project's 'magnitude-not-direction' wall.

Key result

Indicators

Algorithms

Code