A backtest is an argument, not evidence. These are the five things I require before a number is allowed on this site — and what each one is designed to catch.
Symbol, timeframe, period, spread in points, execution model, terminal build and whether the file is the original or a duplicate. A result without its conditions is a screenshot, and screenshots are the currency of this industry for a reason.
Order counts, separated into opened and closed. This is the cheapest filter and it is the one nobody applies. In my first batch of 77 files, 34 never opened a position and 20 opened thousands and closed none. Neither group produced a single verifiable trade, yet both ship with equity curves.
Delete the best 5% of trades by profit and re-sum. If the sign flips, the system is not earning from its logic — it is earning from a handful of outliers that may not recur. Three of the five most profitable EAs fail this test.
The sample is cut into calendar years and each one is checked separately. A system that makes all of its money in a single year is telling you which market condition it needs. That is useful information — but it is not an edge that generalises.
Start lot versus peak lot. An EA that moves from 0.01 to 5.77 lots is not a strategy with a risk setting; it is a strategy whose real size is decided by how long it has been losing. The stop-out is then arithmetic, not probability.
What this method cannot do. It cannot tell you an EA will work in the future. It can only tell you whether the historical result is structurally fragile. A system can pass all five tests and still lose money next month.
| Symbol & timeframe | EURUSD H1 |
| Period | Feb 2024 - Oct 2026 (32 months) |
| Spread | 17 points (1.7 pip), measured on the live feed |
| Execution model | Open prices only |
| Terminal | MT4 build 1492, local machine |
| Starting balance | $10,000 |
| Lot step | 0.01 |
The execution model is open prices only. That is a weakness I state up front: it understates slippage on stop entries and overstates fill quality in fast markets. Results that depend on intrabar precision are marked as such.