Backtesting is the practice of running a trading strategy against historical price data to see how it would have performed if you had traded it in the past. It's the single most important step between "this looks like a good idea" and actually risking real money on it — and it applies just as much to a fully mechanical EA as it does to a manual strategy you follow by hand. Every one of the free EAs on this site is a real .mq4/.mq5 file, and none of them should ever be dragged onto a live chart before being backtested first.
Running a Backtest in the Strategy Tester
MetaTrader 4 and MetaTrader 5 both include a built-in Strategy Tester, and it's the standard tool for this job — you don't need any third-party software to get started. Open it from View → Strategy Tester (or the Ctrl+R shortcut), select the EA you want to test, choose a symbol and timeframe, set a date range for the historical data, and press Start. The tester replays that history tick by tick (or bar by bar, depending on the modeling quality selected) and simulates every trade the EA would have opened and closed, exactly as its code dictates.
Since every EA on this site is downloadable source code, you can drop any of them straight into MQL4/Experts or MQL5/Experts and test it this way before ever attaching it to a live or even a demo chart. The tester outputs a full report: total net profit, drawdown, win rate, profit factor, and an equity curve chart. That report is genuinely useful — it tells you whether a strategy's logic has any historical basis at all — but it is the start of the evaluation process, not the end of it.
Why a Great Backtest Doesn't Guarantee Live Results
A backtest report showing a smooth, steadily rising equity curve is encouraging, but it rests on assumptions that rarely match live trading exactly. The Strategy Tester has to model spread and slippage somehow, and by default it often uses a fixed or simplified spread value rather than the real, constantly shifting spread your broker actually charges — see Spread and Slippage for why that gap matters. Live spread widens during news events and thin liquidity, and slippage on market orders means your actual fill price is sometimes worse than the price the tester assumed. A strategy with a thin per-trade edge can look profitable in a backtest that assumes ideal fills and turn breakeven or worse once real trading costs are subtracted.
Execution speed and requotes are another gap: a backtest fills every valid signal instantly, while a live broker's server can introduce latency, and a fast-moving market can requote or reject an order altogether. None of this means backtesting is worthless — it means a backtest result is a first filter, not a final verdict, and it should be read with those simplifying assumptions in mind rather than taken at face value.
Overfitting: When a Perfect Backtest Is a Warning Sign
The deeper danger in backtesting isn't the tester's spread model — it's what happens when a trader tunes a strategy's parameters over and over against the same historical data until the equity curve looks flawless. This is called overfitting, or curve-fitting: adjusting the Moving Average period, the Stop Loss distance, the ADX threshold, or any other input repeatedly, chasing a better result on one specific stretch of history. The parameters that emerge usually didn't capture a real, repeatable market edge — they captured the specific noise and coincidences of that one dataset, which will never occur in exactly the same way again.
The diagram above is the core of this lesson. The jagged, near-perfect curve on the left is what an overfit strategy's backtest often looks like — it climbs almost without a wobble, right up until live data begins, and then it falls apart, because the "edge" it found was really just a description of the past rather than a forecast of the future. The steadier curve on the right belongs to a more robust strategy: it didn't backtest as spectacularly, but it keeps climbing past the point where the test data ends, because its logic reflects something durable about how the market behaves rather than a coincidence in one sample. A backtest that looks too perfect deserves more suspicion, not less.
Walk-Forward Testing: Checking on Data the Strategy Never Saw
The standard defense against overfitting is walk-forward testing, also called out-of-sample testing. The idea is straightforward: split your historical data into two segments. Tune the strategy's parameters only on the first segment (the "in-sample" period), then, without adjusting anything further, run the same strategy on the second segment (the "out-of-sample" period) — data the strategy never saw during tuning. If performance holds up reasonably well out-of-sample, that's a genuinely encouraging sign. If it collapses, the strategy was very likely fit to noise in the first segment rather than to a real pattern.
A practical version of this in the Strategy Tester is simple: tune parameters using, say, 2022-2023 data, then re-run the identical settings against 2024 data you deliberately avoided looking at during tuning. Repeating this across several different out-of-sample periods, rather than checking just one, gives a far more honest read on whether a strategy — including any of the EAs on this site — has a real edge worth trading.
Demo Testing Is Still Required After a Good Backtest
Even a strategy that survives walk-forward testing convincingly still needs to prove itself in real time before real money is involved. A backtest, no matter how carefully built, is still a simulation running on historical data all at once; a demo account runs the strategy forward, tick by tick, under the platform's actual live execution conditions. See Demo vs Live Account for what specifically changes between the two — the point relevant here is that demo trading is where spread behavior, order execution, and how the strategy performs across genuinely new market conditions can finally be observed directly, rather than assumed.
Run any EA on a demo account across multiple weeks and varied market conditions — trending and ranging, calm and volatile — before ever considering live money. While demo testing, size positions the same way you would with real capital, using the Risk Management Basics approach of risking a small, fixed percentage per trade, so the habits and settings you build during testing carry over unchanged to a live account later.
Why This Matters
Skipping backtesting means trading on a hunch with no historical evidence either way. Backtesting without walk-forward testing means risking that your best-looking result is really just an overfit description of one lucky stretch of history. And backtesting well but skipping demo testing means finding out about real-world execution gaps with real money already on the line, rather than in an environment built specifically to reveal them first. Each step in this lesson exists to catch a different kind of false confidence — running all of them, in order, on any strategy or EA before it ever touches a live account is what turns "this looks good" into an honestly-tested decision.