A Backtest Is an Argument, Not a Promise: Comparing It to Forward Results Honestly

A backtest is an argument, not a promise. It says: under these prices, with this spread model and these fills, these rules would have produced this equity curve. Every one of those conditions is different in forward trading, and the gap between the two is where most disappointment is manufactured.

Requotes and rejected orders. Variable spread widening at the exact moments a breakout system wants to enter. Execution latency between signal and fill. Broker-side stop levels. Terminal downtime, restarts and lost connections. A news filter that consults a live calendar the tester does not populate. None of these appear as a line item in a report — they appear as a slightly worse curve, spread thinly across every trade.

Expecting forward results to match backtest returns is the wrong test — the sample is far too short to say anything. What is worth checking is whether the system behaves the way it was described: does it trade at the expected frequency, hold for the expected duration, take losses of the expected size, and stay inside its stated drawdown? A system that was supposed to trade twice a week and trades twice a day has broken its own description, whatever the profit column says.

A few weeks of results is a handful of trades. With that sample, luck dominates: a strategy with a genuine edge can easily show a loss, and a strategy with none can easily show a profit. This cuts both ways, which is why we do not treat a good short run as validation any more than a bad one as refutation.

We publish backtests on real tick data with real spread, and we state the period, the starting balance and the settings used. We also run forward accounts and publish them — labelled plainly as demo accounts, because that is what they are. A demo account removes the fill and slippage realism that matters most, so we present it as evidence about behaviour and uptime, not as proof of profitability. Anyone who shows you a curve without telling you which kind of account produced it is asking you to guess.

Our three best-selling EAs: GOLD NEURON (AI) https://www.mql5.com/en/market/product/187329 · ATLAS PORTFOLIO https://www.mql5.com/en/market/product/182751