The MT5 Strategy Tester is free, fast, and built into software millions of traders already have open. It is also the tool most likely to hand you a confidently wrong answer, because a single dropdown most people never touch decides how honestly your strategy is simulated. And there is a second surprise: if you were hoping to click through history placing trades by hand, this is not that tool at all.
The Strategy Tester ships with MetaTrader 5 at no cost. Open the terminal, press Ctrl+R, and you have an engine that can run a coded strategy across years of history, sweep hundreds of parameter combinations, and do it across multiple CPU cores. For what it is, it is excellent, and nothing free matches its optimisation power.
The thing to know first, before anything else: the Strategy Tester tests Expert Advisors. It is a harness for automated, coded strategies. It does not offer discretionary backtesting — you cannot step forward one bar at a time and place trades by hand the way a manual trader practises.
This trips up an enormous number of people. Visual mode looks like it should let you trade: a chart appears, bars advance, trades open and close. But those trades are being placed by the EA, not by you. You are watching a replay of a robot's decisions. If your method lives in your judgement rather than in code, the Strategy Tester cannot test it, and no setting changes that.
This is the dropdown that decides whether your backtest means anything. MetaQuotes' own documentation, read on 17 August 2026, names five modes.
| Mode | What it simulates | Honest verdict |
|---|---|---|
| Every tick based on real ticks | Actual tick data from your broker, no simulation | The only mode to trust for short-term strategies. Requires large downloads and is broker-specific. |
| Every tick | Ticks generated by MT5's own interpolation algorithm | MetaQuotes calls it "the most accurate but the slowest". Accurate simulation — the ticks are still invented. |
| 1 minute OHLC | Four prices per minute bar | Adequate for slow strategies. Intrabar path is fiction. |
| Open prices only | The open price of each bar | Optimisation triage only. Never report a result from this mode. |
| Math calculations | No history, no symbols, no ticks at all | Not a backtest. Pure parameter maths. |
Here is the problem stated with numbers from our own archive rather than in the abstract. Across 1,911,141 EURUSD one-minute bars from 3 January 2021 to 30 April 2026, the average high-to-low range inside a single one-minute bar is roughly 1.5 pips. On an hourly chart that intrabar travel compounds enormously.
"Open prices only" throws all of it away. Your stop loss sitting two pips inside that range is never touched, because the price that would have touched it is never simulated. A strategy with tight stops can post a flawless equity curve in this mode and be structurally unprofitable in reality — the losses simply were not modelled.
Volatility is not evenly distributed, so a mode that discards intrabar movement discards wildly different amounts depending on when your strategy trades. From the same 1.9 million EURUSD bars:
| UTC hour | Avg one-minute range | Bars sampled |
|---|---|---|
| 14:00 | 2.83 pips | 80,029 |
| 13:00 | 2.71 pips | 80,036 |
| 08:00 | 2.13 pips | 80,033 |
| 21:00 | 0.73 pips | 76,385 |
A 3.9× spread between the busiest and quietest hour. A strategy trading the 14:00 UTC window has almost four times as much unmodelled intrabar movement being ignored as one trading 21:00. The same modelling setting is therefore far more dangerous for a London-session strategy than an Asian-session one — which is exactly the kind of asymmetry a single "modelling quality: 99%" figure hides.
After a test, MT5 reports a modelling quality percentage, and traders treat 99% as a seal of approval. It is not a measure of whether your results are correct. It measures how completely the tester was able to build ticks from the history data available — a data-coverage score, not an accuracy score.
You can achieve 99% modelling quality on a strategy whose logic is curve-fitted to noise, whose stops are unrealistically tight, and whose spread assumptions are fantasy. The number will still say 99%. It answers "did the tester have enough data?", never "should you trade this?".
Two further gaps worth naming plainly:
Being fair to it, because the criticism above is about misuse rather than quality:
| You are… | Verdict |
|---|---|
| Coding EAs in MQL5 | Use it. Nothing free is better. |
| Optimising parameters on a rule-based system | Use it — with forward testing enabled. |
| Testing a scalping EA | Use it, but only on real ticks. Any other mode will lie to you. |
| A discretionary trader wanting screen time | Wrong tool. It cannot do manual backtesting at all. |
| Practising price action bar by bar | Wrong tool. Use a replay tool. |
| On a Mac, or working from a phone | Wrong tool. MT5 is a desktop install. |
| Not a programmer | Wrong tool. Without MQL5 there is nothing to test. |
The dividing line is simple: the Strategy Tester tests code. If your edge lives in code, it is one of the best free tools in trading. If your edge lives in your reading of a chart, it cannot help you, however many settings you change.
If what you actually wanted was to step through history placing your own trades, FXAbsolute does that in a browser: 28,412,683 one-minute bars across 15 forex, index and crypto instruments from 2021 to 2026, bar-by-bar replay, drawing tools, limit orders, an automatic journal and live statistics. No account, no download, no payment.
The trade-off runs the other way, and it is a real one:
| MT5 Strategy Tester | FXAbsolute | |
|---|---|---|
| Price | Free | Free |
| Manual bar-by-bar trading | No | Yes |
| Automated EA testing | Yes | No |
| Parameter optimisation | Yes | No |
| Real tick data & spread | Yes | No — mid-price OHLC |
| Requires programming | Yes (MQL5) | No |
| Install | Windows desktop | None — browser |
| Works on phone | No | Yes |
They are not really competitors. One tests robots with high fidelity; the other trains a human's judgement. Traders who do both use both, and neither costs anything.
Yes. It is built into the MetaTrader 5 terminal at no cost, with no paid tier, no request cap and no upsell. You need a broker account to download history data, but demo accounts work for testing.
No. The Strategy Tester runs Expert Advisors — coded, automated strategies. Visual mode lets you watch an EA place trades on a chart, but you cannot place trades yourself bar by bar. For discretionary manual backtesting you need an add-on such as Soft4FX, or a browser tool such as FXAbsolute.
For realism, "Every tick based on real ticks" — it replays actual tick data from your broker rather than simulating it. MetaQuotes describes plain "Every tick" as the most accurate simulation but the slowest. "Open prices only" and "Math calculations" should never be used to judge a strategy, only to triage parameters quickly.
It measures how completely the tester could construct ticks from the available history data — a data-coverage score, not a correctness score. A curve-fitted strategy with unrealistic stops can still report 99% modelling quality. It answers whether the tester had enough data, not whether the result is trustworthy.
The usual causes are modelling mode discarding intrabar price movement, assumed rather than real spread, broker-specific tick data, and slippage that is not simulated. Strategies with tight stops suffer most, because the price movement that would have hit the stop was never generated.
MT5, clearly. It supports real tick data, multi-currency testing, multi-core optimisation and built-in forward testing, none of which MT4 offers. MT4 remains common only because of legacy EAs and broker support.
FXAbsolute replays five years of real one-minute data bar by bar in your browser — no account, no download, no card. The half of backtesting MT5 cannot do.
Open FXAbsolute →