Why the Same EA Backtests Differently on Every Broker: Spread and Feed Quality, the Variable Nobody Discloses

Run the same EA with the same set file over the same dates on two different brokers and you will get two different equity curves. Buyers usually assume one of the tests is broken. Most of the time both are correct — for that broker. The variable that moved is one almost no vendor talks about: the price feed itself.

There is no single "gold spread". Each broker builds its quote stream from its own liquidity providers and adds its own markup, so the spread your backtest pays is whatever the historical ticks of that one broker contain. On short-hold strategies the spread is paid on every single trade, so a difference of a few points per trade compounds into a visibly different curve — same entries, same exits, different cost.

Two brokers do not just charge different spreads; they record different ticks. One feed may show a spike the other never printed, which is enough to fire or skip a stop, an entry, or a trailing step. From that moment the two tests are no longer the same sequence of trades at different costs — they are different trade lists entirely. "Every tick based on real ticks" mode is honest about one broker's history; it cannot speak for another's.

Sensitivity scales with holding time and stop distance. A scalper that targets a handful of points on gold lives or dies on feed quality, and this is where the gap between a vendor's screenshot and your own test is largest. A slower system that holds for hours or days and risks a wide stop pays the spread rarely relative to its target, so its results travel across brokers far better. Neither is "the true result" — they are measurements of the strategy on two different instruments that happen to share a name.

Before buying or going live, backtest on your own broker's real-tick data, on the account type you will actually trade, and treat any vendor screenshot as an upper bound produced on a feed you do not have. If a strategy only survives on one broker's feed, that is not an edge — it is an artifact. We run our own validation on real-tick data and consider cross-broker degradation a test the EA has to pass, not an excuse to be explained afterwards.

Because the cleanest feed produces the prettiest number, and the prettiest number sells. Disclosing that results vary by broker makes a product page more honest and less shiny. We would rather you test on your feed first and be unsurprised later.

An example of an EA we validate on real-tick data, with the methodology stated on the page: https://www.mql5.com/en/market/product/187329

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