Same EA, Same Dates, Two Brokers: Profitable on One Feed, Losing on the Other

Here is the most expensive lesson from our testing year, in one sentence: a strategy that has only been proven on one broker's data has not been proven at all.

We had a promising trend system. On one well-known clean tick archive it printed a profit factor above 2 with a single-digit drawdown across nine years. We were ready to ship it. As a final gate we re-ran the identical settings, identical dates, on a real broker's tick history.

It lost money. Profit factor below 1, drawdown three times larger. Same market, same years, same code.

Tick-level entries are sensitive to things that differ between feeds: spread behavior around session opens, the exact sequence of ticks inside a bar, gaps, and how the archive was reconstructed. A signal that fires on a smooth reconstructed feed can fire one tick later — or not at all — on a rougher real feed. If the strategy's edge is thin, that difference is the whole edge.

The nastier version: an optimizer will happily tune parameters to exploit artifacts of one specific feed. The result looks like an edge and is actually a description of that archive's noise.

Every product must be profitable on at least two independent broker feeds over the same window, including a stress window with 2020 and 2022 in it. Strategies that only survive on one feed get archived, no matter how beautiful the curve. Several did not make the cut, including some we liked a lot.

1. Ask the seller on which broker the published backtest was made, and whether it was repeated elsewhere.

2. Prefer strategies whose logic is scale-based (ATR, channels, daily ranges) over ones needing tick-perfect fills (ultra-thin take-profits are the classic warning sign).

3. Run the free demo in your own MT5 tester on your own broker's data. If the seller's numbers collapse there, you have your answer before spending anything.

Our products and free utilities: https://www.mql5.com/en/users/app.develop.sk/seller