Backtestle

Case studies: 10 years of rule tests on EURUSD

We take a rule you often hear in trading courses, define it precisely, and test it on 10 years of EURUSD data with costs included. The numbers come from the Lab, and the pages update whenever the data updates each week. Not investment advice.

By · Last updated:

How we run the tests

Data

Tests run on 5-minute EURUSD candles. Prices are FXCM bid prices; short gaps are filled from Dukascopy. The data is checked candle by candle against an independent source, and every difference found is written down on the Data page. The data in the tests is the same data you see on the chart: you can open the same day in Backtest and check the same trade yourself.

Period and splits

The data runs from October 2016 to today, about 10 years, and updates every week; the numbers on these pages refresh with it. Every result is split into before and after October 2022 and also shown year by year. That way a rule that only worked in one market regime stands out right away; a rule that isn't positive in both periods doesn't count as robust.

Rule definition

We take an idea you often hear in trading courses and write it so it leaves no room for interpretation: what the signal is, where the entry is, where the stop is, and how many R the target is. Entry happens after the signal candle closes; no information from the future is used. If the stop and target both fall inside the same candle, the trade counts as a stop. The same rule on the same data always produces the same trades, so the result is reproducible.

Spread and slippage

Every trade pays a 0.6 pip cost, which stands for the usual combined spread and slippage on EURUSD. Trades whose risk is tiny compared with that cost are skipped, because in real trading the cost would eat them. Results are given in risk units (R), independent of lot size and account balance. Overnight swap and commission differences are not modeled separately.

Limitations

Past results don't guarantee future results. The tests cover EURUSD only; don't expect the same result on another pair. Spread widening during news, the real order queue, and psychology are not modeled. When many rules are scanned, some can look good by luck; the two-period and year-by-year criteria reduce that risk but don't remove it. These pages are a starting point before you test a rule in your own backtest, not investment advice.

All tests: Lab