How the Strategy Works
We trade 3 currency pairs (EUR/USD, EUR/CHF, USD/JPY) using 6 independent methods. The methods are structurally different from each other, which is why they diversify well.
Why These Survive
The surviving methods trade infrequently enough that transaction costs don't erode the edge. Everything else died to costs.
The spread to enter and exit EUR/USD is about 1.5 pips at retail. Any method that trades frequently needs to earn more than that per round trip just to break even. Most don't.
Risks
- The edge is small. A bad quarter could drop the account 4-5%.
- Broker costs matter. Works at IB (~0.4 pip spread), strained at typical retail.
- 150+ ideas tested, so some survival-bias risk despite walk-forward and placebo testing.
- Markets change. The structural patterns we exploit could weaken or disappear.
Simulation Methodology
All results are backtested on 5-second tick data spanning September 2015 to February 2026 (~35 million data points per pair). Every trade enters and exits at the actual bid or ask price at the 5-second bar where the signal fires. Spread, slippage, and overnight financing are all deducted.
The numbers on this site come from three independent codebases:
- Research framework. The original backtest engine, used for all experiments, parameter sweeps, and validation tests. Bar-level simulation with full bid/ask spreads.
- Live trading engine. A separate codebase written from scratch to execute the strategies against a real broker. Runs in production.
- Audit simulator. Built from the live trading engine's code (same signal generation, same execution logic), but replaying historical tick data instead of live prices. Line-by-line audited against the research framework.
The research framework and audit simulator agree to within 2% on total PnL over the full backtest period. Where they differ, the headline numbers use the more conservative figure.