To backtest an MT5 EA: press Ctrl+R to open Strategy Tester, select your EA, set the symbol and timeframe, choose 1 Minute OHLC as the model (for H1+ EAs), set at least 12 months of date range, match deposit and leverage to your broker, then click Start. Check the Report tab for quality % — must be 90%+ for valid results, 99%+ preferred. Need at least 30 completed trades for statistical meaning.
Table of Contents
MetaTrader 5’s Strategy Tester is one of the most powerful tools available to retail forex traders — but only if you use it correctly. Default settings produce results that look convincing but contain fundamental errors: wrong data model, incomplete history, incorrect spread, periods too short to be statistically valid.
This guide walks through every Strategy Tester setting in order, explains what each one does, and identifies the specific mistakes that generate impressive-looking backtests for strategies that will fail on live markets.
Opening the Strategy Tester
Three ways to open MT5 Strategy Tester:
- Keyboard shortcut: Press Ctrl+R anywhere in MT5
- Menu: View → Strategy Tester
- Toolbar: Click the Strategy Tester icon
The tester panel opens at the bottom of MT5. It has tabs across the top: Settings, Results, Graph, Report, Optimization Results. All configuration happens in the Settings tab.
Every Setting Explained
Select Your EA
Click the Expert dropdown and select your EA. If it does not appear, the .ex5 file is not in the correct Experts folder, or the EA has not been compiled. Open MetaEditor (F4), open the .mq5 source file, and press F7 to compile. The compiled .ex5 file then appears in the tester list.
Critical: When testing multiple EA versions, confirm the exact name in the dropdown. If you have QuantumForex_TITAN and QuantumForex_TITAN_v2, selecting the wrong one produces results for the wrong EA entirely — a mistake that has invalidated more than one trader’s session.
Match the EA’s Designed Pair and Timeframe
Set the symbol to the exact pair the EA was designed for. Set timeframe to the EA’s chart period. An H1 EA must be tested on H1. Testing an H1 EA on M15 produces 4× more bars and potentially inflated trade counts that do not reflect real operation.
If the symbol is greyed out or missing from the dropdown, right-click Market Watch and select “Show All” to load all available pairs.
Modelling Method — The Most Important Setting
Controls how MT5 simulates price movement. Three options:
- 1 Minute OHLC — uses real 1-minute candle data. Highly accurate for H1+ EAs. Fast. Achieves 100% quality. Recommended for most EAs.
- Every Tick — generates synthetic tick data from 1-minute bars. More granular but slower and prone to artefacts from synthetic tick generation. Only meaningfully better for M1/M5 EAs where sub-minute price movement matters.
- Open Prices Only — tests only at bar open. For quick optimisation checks only — not valid for final results.
Set a Meaningful Test Period
Minimum 12 months. Recommended 18–24 months. The tested period must include:
- At least one trending market period
- At least one ranging/consolidation period
- At least one high-volatility event window (rate decisions, major news)
Testing only on a favourable market phase is the most common way to produce misleadingly good results. Extend the date range until it covers multiple market conditions.
Match Your Real Account
Set deposit to the actual balance you will trade. Lot sizes calculated from percentage risk are balance-dependent — testing on $10,000 and running on $1,000 produces different lot sizes. Always match deposit to your real account or prop firm challenge size.
Set leverage to match your broker. Prop firms typically cap at 1:100 for EA accounts. Higher leverage does not change results unless the EA uses fixed lot sizes — it only affects margin requirements.
Data Quality — The Most Misunderstood Setting
After your backtest runs, the Report tab shows a “Quality” percentage. This is the fraction of the tested period for which MT5 found real downloaded historical data.
| Quality % | Meaning | Trust Level | Action |
|---|---|---|---|
| 99–100% | Near-complete real data | ✅ Trustworthy | Valid result — proceed with analysis |
| 90–98% | Minor data gaps filled | ⚠ Acceptable | Results likely valid with minor caveats |
| 50–89% | Significant synthetic data | ⚠ Unreliable | Re-download history from better broker source |
| Under 50% | Mostly generated data | ❌ Invalid | Results meaningless — fix data source first |
How to Get 100% Quality Data
Data quality depends on your broker. IC Markets, Pepperstone, and FP Markets retain 10+ years of 1-minute history. Some Exness accounts only hold from January 2026. To download missing history:
- Open the chart for the pair you want to test
- Switch to the M1 timeframe
- Scroll all the way back in chart history — MT5 downloads bars as you scroll
- Wait for loading to complete (progress shown in the status bar at the bottom)
- Re-run the backtest — quality should now be higher
1 Minute OHLC vs Every Tick
For EAs trading on H1, H4, or daily timeframes, 1 Minute OHLC is correct. The difference vs Every Tick is under 2% on most performance metrics. Every Tick is only meaningfully better for M1/M5 EAs where the sub-minute price movement between candles determines whether the stop or take profit was hit first.
Practical rule: if your EA’s entry and exit logic only checks at bar close, use 1 Minute OHLC. If your EA trails stops or closes intra-bar, Every Tick is more accurate — but ensure you have real tick data, not generated ticks.
Running the Backtest Step-by-Step
Confirm all settings and click Start
Verify: EA selected and correct version, symbol correct, timeframe matches EA, model = 1 Minute OHLC, date range 12+ months, deposit and leverage entered. Click Start. Progress bar shows completion. Typical time: 5–60 seconds depending on EA complexity and date range.
Check Journal tab for errors
If the tester hangs at 0% for 30+ seconds, the EA likely has a compilation error or infinite loop. Check the Journal tab for error messages. If progress jumps from 0% to 100% with zero trades, the EA’s entry conditions did not fire — confirm you selected the correct pair and timeframe.
Check Report tab before looking at profit
Before reading any profit numbers, verify: (1) Quality % is 90%+. (2) Total trades is 20+. (3) Modelling confirms “1 Minute OHLC” and not “Open Prices Only” by mistake. Only after these three checks should you proceed to analyse the performance metrics.
Key Metrics Explained
| Metric | What It Measures | Target (Prop Firm EA) | Red Flag |
|---|---|---|---|
| Profit Factor | Gross profit ÷ gross loss | 1.5–2.5 | Over 3.0 = likely over-optimised |
| Win Rate | % trades closed in profit | 65%–80% | Over 90% = check SL/TP ratio |
| Max Drawdown | Largest equity decline from peak | Under 5% | Over 10% = not suitable for prop firms |
| Expected Payoff | Average profit per trade | Positive | Negative = losing system regardless of win rate |
| Sharpe Ratio | Return relative to volatility | Over 3.0 | Under 1.0 = too volatile for the return generated |
| Total Trades | Completed round trips | 30–150/year | Under 20 = statistically meaningless |
| Recovery Factor | Net profit ÷ max drawdown | Over 3.0 | Under 1.0 = DD too large relative to gain |
5 Mistakes That Produce Fake Results
- 1. Testing with zero spread. Default tester spread is often 0 or minimum. Real spreads on CHF pairs during off-hours can reach 5–8 pips. Always set spread to “Current spread” or manually enter the real average spread your broker charges.
- 2. Data quality under 90%. Synthetic data fills gaps with generated candles that can have impossible OHLC relationships or miss real news-spike candles. Download full history before testing.
- 3. Test period too short or too favourable. Testing only on 2020–2021 and calling the result validated is misleading. Test through a full cycle including trending and consolidating phases.
- 4. Optimise and validate on the same data. If you optimise parameters on Jan–Dec 2024, then run a “backtest” on the same period, you are measuring curve-fitting, not strategy validity. Always split: optimise on one period, validate on a separate out-of-sample period.
- 5. Ignoring the trade count. 8 trades with 100% win rate is not validated — it is luck. A statistically valid sample requires at least 30 completed trades, ideally 50+. Under 20 trades and any conclusions about win rate or profit factor are meaningless.
Visual Backtest Mode
Visual backtest replays the test in real time on an animated chart, showing each trade entry and exit as it happens. Useful for:
- Confirming the EA enters trades in the right market conditions
- Spotting entries that should not have fired based on your reading of the market
- Recording video evidence of EA performance for documentation or sales pages
- Debugging unexpected entries that the log alone cannot explain
To enable: tick the “Visualization” checkbox in Settings before clicking Start. Use the speed slider to control playback — 60–70% is ideal for recording without missing detail.
The Out-of-Sample Test — The Only Real Validation
After you are satisfied with a backtest result, run the EA on a date range it has never seen — the most recent 3–6 months. If profit factor and win rate hold within 20% of the in-sample result, the strategy is likely not curve-fitted. If performance collapses on new data, the parameters were tuned specifically to the historical period.
TITAN AutoTrader’s original 17-month backtest (Jan 2023–May 2024) showed 78.6% WR and PF 1.95 on CADCHF. When run on the unseen 2025–2026 period: 73.81% WR and PF 1.93. Results held within 7% — confirmation the strategy is not curve-fitted and holds on live market data.
TITAN AutoTrader — 100% Quality, Validated Out-of-Sample
17-month backtest at 99%+ quality, confirmed on unseen 2025–2026 data: 73.81% WR, PF 1.93, DD 1.00%. Not curve-fitted. Built and tested the right way.
See Full Results → Get TITAN