Home › Learn › Tutorials

How to Backtest an EA in the MT4 Strategy Tester

Updated 1 September 2026 · 8 min read · PipTax education

If you want to backtest an EA in the MT4 Strategy Tester properly, you need more than just hitting "Start" — the quality of your data, your modelling settings, and your cost assumptions all decide whether the results mean anything at all. Get those wrong and you'll build false confidence in a strategy that was never actually profitable.

This guide walks through the practical steps: setting up the tester, choosing modelling quality, feeding in realistic costs, and reading the results without fooling yourself.

Setting Up the MT4 Strategy Tester Correctly

Open the Strategy Tester in MT4 (View → Strategy Tester, or Ctrl+R). Before running anything, check these settings:

• Expert Advisor: select the compiled EA you want to test from the dropdown.
• Symbol: match the exact pair the EA was designed for — results on EURUSD won't tell you much about a GBPJPY strategy.
• Period: the chart timeframe the EA reads price from (M1, M15, H1, etc.).
• Model: this is the single most important setting — covered in detail below.
• Spread: fixed value in points, or "Current" to use the spread at test time.
• Use date: tick this and set a From/To range so you're testing a deliberate period, not whatever history happens to be loaded.

Also confirm your initial deposit and leverage match a realistic account size — testing with £100,000 when you'd actually trade £2,000 changes position sizing behaviour and can hide margin issues.

Choosing the Right Modelling Quality

MT4 offers three modelling options, and they are not interchangeable:

| Mode | How it works | Best for | |---|---|---| | Every tick | Reconstructs sub-minute price movement from the smallest data available | Final validation, final go/no-go decision | | Control points | Samples fewer price points, interpolates between them | Quick sanity checks | | Open prices only | Uses only the open price of each bar | Very fast scans of many settings |

For any serious decision — before you fund a live account or hand an EA real money — run the full test in Every tick mode. It's slower and needs more disk space for historical data, but it's the only mode that properly simulates stop-loss and take-profit hits that happen intrabar rather than exactly on the candle open.

Use "Open prices only" purely for rough, early-stage filtering when you're testing dozens of parameter combinations and just want to eliminate obviously broken ones.

Historical Data Quality Matters More Than You Think

MT4's built-in history is often incomplete or has gaps, especially for older dates or less liquid pairs. Before trusting any backtest:

• Use the History Centre (F2) to check how far back your data goes and look for suspicious gaps.
• Download extra history if your broker's default feed is short — many brokers, and some third-party providers, offer downloadable tick data.
• Test across at least 2–3 years and varied conditions — a strategy that only survives a quiet trending period isn't proven.
• Re-run the same test twice to confirm consistent results; if numbers shift, your data or modelling quality isn't stable.

Poor data is the single most common reason a backtest looks great in the tester and then falls apart live — not because the strategy logic is wrong, but because the historical price action being tested was never real to begin with.

Modelling Realistic Trading Costs

This is where most home-built backtests quietly lie to their owners. The default MT4 tester often assumes a flat spread and no commission, which flatters almost any EA.

To model costs properly:

• Set the spread field to match what you'd actually pay, not zero or an optimistic guess.
• If your broker charges commission per lot (common on ECN/raw-spread accounts), add it manually — MT4's tester doesn't always include it automatically depending on account type settings.
• Factor in swap/rollover if the EA holds positions overnight — this compounds over longer backtests.
• Remember spreads widen around news releases and thin liquidity — a fixed-spread backtest can't show that risk.

Rather than guessing at numbers, check current live spreads and commissions for your actual account type using PipTax's [cost audit tool](/audit.html) and the [broker comparison pages](/brokers/index.html) — for example, comparing a standard account vs a raw-spread account on Pepperstone, or IG's own platform pricing vs its MetaTrader offering. Plug those real figures into the tester before you trust the output.

Optimisation and Walk-Forward Testing

MT4's Optimization tab lets you test many parameter combinations at once and rank them by profit, drawdown, or other criteria. It's a useful tool — and an easy way to fool yourself.

• Optimising on one dataset and testing on the same dataset is curve-fitting — you're tuning parameters to fit noise, not finding a genuine edge.
• Split your data: optimise on an earlier period (in-sample), then run the resulting settings unchanged on a later period (out-of-sample) the EA has never "seen."
• If performance collapses out-of-sample, the strategy was likely overfitted.
• Repeat this walk-forward, rolling the window forward through history, to build more confidence than a single split gives you.
• Be wary of any optimisation result with an enormous profit factor on a tiny number of trades — it's usually a statistical fluke, not an edge.

From Backtest to Live Account

Once you've run a full "Every tick" backtest with realistic costs and confirmed it survives out-of-sample testing, the next step isn't going live with full size:

1. Forward test on a demo account for several weeks to confirm live execution matches backtest behaviour. 2. Compare live costs to backtest assumptions — pull current spread and commission data via the [cost audit tool](/audit.html) and check the [rates page](/rates.html) for swap figures, since backtested numbers can go stale. 3. Start small on a live account with minimum position size before scaling up. 4. Keep monitoring — market conditions shift, and a strategy validated on 2019–2022 data may behave differently in a new volatility regime.

For a deeper grounding in testing methodology and how PipTax verifies broker cost data, see the [methodology page](/methodology.html), and browse the [school section](/school/index.html) for related tutorials on demo testing and risk sizing.

Conclusion

Learning to backtest an EA in the MT4 Strategy Tester properly — with quality data, "Every tick" modelling, honest cost inputs, and out-of-sample validation — is the difference between a useful pre-flight check and an expensive illusion. Trading remains risky regardless of how clean a backtest looks, so treat these results as one input among several, not a guarantee, and always verify your cost assumptions against live broker numbers before committing real money.

Keep going: Audit Methodology Index Index