Moving-Average Crossovers
Buy when a fast moving average crosses above a slow one, sell on the reverse (e.g., golden/death cross).
Early in-sample results did not survive data-snooping corrections or out-of-sample testing; our own 50/200 test on FX majors and gold returned PF 0.94.
- Why it fails
- The early in-sample evidence (Brock-Lakonishok-LeBaron 1992) did not survive data-snooping corrections and out-of-sample testing (Sullivan-Timmermann-White 1999); our own 50/200 test on FX majors plus gold returned a profit factor of 0.94.
- When / how it stopped
- Weak-to-negative out-of-sample once data-snooping is accounted for.
The moving-average crossover is the canonical trend rule: go long when a fast average crosses above a slow one, exit or reverse when it crosses back. The 50/200 “golden cross” and “death cross” are its best-known form.
Unlike astrology or Gann, this method has a genuine empirical kernel. Brock, Lakonishok & LeBaron (1992) reported that simple moving-average rules carried information in early stock data, and that paper is still cited as evidence the approach can work.
The decay is well documented. Sullivan, Timmermann & White (1999) re-examined those rules with explicit data-snooping corrections — accounting for the universe of rules that were searched — and found the apparent edge largely evaporated out-of-sample. What looked like signal was substantially the product of testing many rules and reporting the best.
Our own work is consistent with that. A 50/200 crossover across FX majors plus gold returned a profit factor of 0.94 — below the 1.0 breakeven line before you even weight the drawdown. The rule is honest and well-defined; the edge simply is not there after costs once data-snooping is controlled for.
The full gate scorecard and equity curves are in our validation write-up.
→ Read our full validation report: /strategy/d1-trend-follower
Sources
- Brock, Lakonishok & LeBaron (1992), "Simple Technical Trading Rules and the Stochastic Properties of Stock Returns", Journal of Finance
- Sullivan, Timmermann & White (1999), "Data-Snooping, Technical Trading Rule Performance, and the Bootstrap", Journal of Finance
- Our validation report — D1 Trend Follower (50/200 on FX majors + gold), PF 0.94
Frequently asked
Do moving-average crossovers work in 2026?
The honest answer is weak-to-negative out-of-sample once data-snooping is accounted for. The famous early result (Brock, Lakonishok & LeBaron 1992) looked promising in-sample, but Sullivan, Timmermann & White (1999) showed it did not survive proper data-snooping corrections and out-of-sample testing. Our own 50/200 crossover test across FX majors plus gold returned a profit factor of 0.94 — below breakeven after costs.
Is the golden cross or death cross profitable?
Not reliably, after costs, in our testing. Crossover rules can ride a strong trend when one exists, but on FX majors over recent years the trends were not there, and the signal lagged the turns. The crossover is a real, well-defined rule — it just does not clear the evidentiary bar out-of-sample once the snooping that flattered early studies is removed.
Not investment advice — your mileage may vary, but the burden of proof is on the person claiming an edge. This entry describes general research and published evidence (or its absence), not a recommendation. See the full disclaimer.