The MT5 Strategy Tester Explained: Tick Modes, Spread and Trust

The MT5 Strategy Tester is the best backtesting tool retail traders get for free. It is also very good at producing confident, detailed, thoroughly misleading results if you point it in the wrong direction.

MT5 Strategy Tester at a glance

What it is and why it works

The Strategy Tester takes historical price data, feeds it to an Expert Advisor bar by bar or tick by tick, and records every trade the EA would have taken. At the end it gives you an equity curve and a page of statistics: net profit, profit factor, expected payoff, maximum drawdown, recovery factor and more. On MT5 it does this using every processor core you have, across multiple symbols if you want, and it will optionally rent additional computing power from MetaQuotes’ cloud network for large optimisations.

It is a genuinely powerful piece of software, and it is a substantial upgrade on MT4’s tester, which runs one symbol on one core. But everything it produces is conditional on two things: the quality of the historical data you fed it, and the realism of the execution assumptions you chose. Change either and the same strategy can go from an attractive curve to a losing one.

This is where MT4 and MT5 differ in a way worth understanding. MT4 reports a modelling quality percentage (you will see 90%, 25% or n/a) which describes how much of the tick data was interpolated from one-minute bars rather than real. Traders learned to chase the 90% figure. MT5 does not report that number. Instead it asks you to choose a modelling mode up front, and the honesty of the test follows from that choice rather than from a score at the end.

The most important thing to internalise is what the tester is for. It is not a machine for finding out whether a strategy is profitable. It is a machine for finding out whether a strategy is obviously broken, how it behaves in different market conditions, and how large its worst historical drawdown was. Used as a filter to reject bad ideas, it is invaluable. Used as evidence that an idea is good, it will mislead you, because it is trivially easy to optimise a set of parameters into a curve that fits the past and predicts nothing. See backtesting a strategy for the wider discipline around this.

How to trade it, step by step

1. Open the tester and pick the symbol, period and dates. Press Ctrl+R, choose your Expert Advisor from the dropdown, then set the symbol and timeframe. Use the exact symbol you actually trade, including any broker suffix such as EURUSD.r, because the contract specification and history attached to it can differ from the plain symbol. Set a date range that includes at least one trending period, one ranging period and one high-volatility event.
2. Choose the modelling mode honestly. Every tick based on real ticks uses actual bid and ask history from your broker and is the most realistic option, provided your broker supplies that history. Every tick generates synthetic ticks from one-minute bars. 1 minute OHLC and Open prices only are dramatically faster and only valid for strategies that act on completed bars. If your EA reacts intrabar, anything below real ticks will produce results that do not exist.
3. Download the history first and check it is complete. The tester pulls history from your broker’s server, and the depth and quality vary. Open the symbol’s chart, scroll back to the start of your test period so the data downloads, and check the Journal tab of the tester afterwards for messages about missing history. A test over a period the broker does not have data for will run and produce numbers anyway.
4. Set the spread to something you would actually pay. The Current option takes the spread at the moment you start the test and applies it as a fixed value throughout, which flatters any strategy tested at a quiet moment. Enter a fixed value instead, and choose one at or slightly above your broker’s typical spread rather than its best. On strategies with small targets, this single setting can flip the result.
5. Add commission and check the swap figures. Commission is taken from the symbol’s contract specification, and on many symbols it is set to zero even though you are charged in reality. If your account pays commission per lot, and the tester shows none, your results are overstated by that amount on every single trade. Overnight swap is applied from the current specification, which will not match what it was years ago.
6. Set an execution delay rather than leaving it ideal. The tester offers instant execution by default, which nobody gets. Choose a random delay or a fixed number of milliseconds, and rerun. A strategy whose results collapse when you introduce a modest delay was relying on fills it would never have received.
7. Run it visually the first time. Tick Visual mode and watch the trades appear on the chart. You will spot things no statistic reveals: entries at absurd prices, positions held through weekends, stops that never moved, a hundred trades in an hour. Use the pause and step controls to examine individual entries. This is also how you test an indicator, since indicators can be selected in the tester and replayed bar by bar.
8. Use forward testing, and treat the forward half as the real result. Set the Forward option to reserve a portion of the period (a half, a third or a quarter) that the optimiser is not allowed to see. Optimise on the first part, then look at how the chosen settings performed on the reserved part. If the forward section falls apart, you fitted the past. That is the normal outcome, not an unusual one.
9. Re-run the winning settings with the parameters nudged. Take your best result and change each input by ten or twenty percent in both directions. A robust strategy degrades gently. If moving one input by a small amount destroys the result, you have found a peak in the optimisation landscape rather than an edge in the market, and it will not repeat.

Size every one of those entries with the position size calculator and check the trade is worth taking with the risk/reward calculator before you commit.

The conditions it needs

Rejecting bad ideas quickly

This is the tester’s highest-value use and it is underrated. If an idea loses money across five years of data with realistic costs, you have saved yourself months of live testing in an afternoon. Negative results are far more trustworthy than positive ones, because it is easy to accidentally build a strategy that looks good and hard to accidentally build one that looks terrible.

Measuring drawdown and behaviour, not profit

The most useful numbers in the report are maximum drawdown, the longest losing streak and the distribution of trade outcomes. Those tell you what you would have had to sit through, which is what actually determines whether you can trade a system. Profit figures from a fitted backtest tell you very little.

Watching a system in visual mode

Running the test visually and stepping through trades reveals logic errors that no statistic exposes. It is the fastest way to discover that your EA is entering on the wrong side, holding through the weekend, or firing repeatedly on the same bar.

Checking whether an indicator repaints

Because MT5 lets you run an indicator in the tester’s visual mode, you can watch it draw bar by bar as history unfolds. If a signal appears on a bar and later moves or disappears, the indicator repaints, and any impression of accuracy you formed from looking at a static historical chart is worthless.

Testing across multiple symbols and periods

A rule set that works on one pair over one two-year window has told you nothing. The same rules holding up, with the same parameters, across several instruments and several distinct market regimes is a much stronger signal, and MT5’s multi-symbol, multi-core tester makes running that comparison practical.

When it fails

• Slippage is not modelled the way you will experience it. The tester can add an execution delay, but it cannot reproduce the moment your stop is filled thirty points away because liquidity vanished on a data release. Every backtest is executed in a market that always had someone on the other side at your price. Live trading is not that market.
• Spread is applied as a constant, not as something that widens. Real spreads expand at the rollover, at session opens and around scheduled news: precisely when many strategies trade. A test at a fixed two points through a period containing a rate decision is describing a market that did not exist.
• Optimisation produces curve-fitted results by design. Testing thousands of parameter combinations and keeping the best one guarantees a good historical curve even from random data. That is arithmetic, not skill. Without forward testing and parameter-stability checks, an optimisation report is a measure of how hard you searched, not of whether an edge exists.
• Commission and swap are frequently understated. Commission comes from the symbol specification and is often zero in the tester even when you pay it live. Swap is applied at today’s rate across the whole historical period, though rates changed substantially over the last few years. Both errors always run in the same direction: they make results look better.
• Historical data is your broker’s, and it has gaps. History depth, tick availability and even the symbol specification vary between brokers and change over time. Contract sizes, minimum stop levels and margin requirements were different in the past. The tester applies the current ones.
• It cannot test you. The backtest never hesitated, never skipped a trade after three losses, never moved a stop, never doubled up to make back a bad morning. If your live results are worse than your backtest, the gap is often behavioural rather than technical, and the tester is silent on that entirely.

For different levels of experience

If you are brand new

If you are new, here is what the Strategy Tester is and is not. It is a replay machine: it takes old price data and shows you what an automated system would have done. It is not proof that anything will work in future.

The single most useful thing you can do with it as a beginner has nothing to do with Expert Advisors. Press Ctrl+R, choose an indicator instead of an EA, tick Visual mode, and watch a chart replay bar by bar with that indicator on it. You get to see how price actually develops, one bar at a time, without knowing what comes next. That is a completely different experience from scrolling back through a finished chart where every signal looks obvious, and it will teach you more in an hour than a week of reading.

Be very sceptical of anyone showing you a backtest equity curve as evidence. It takes about twenty minutes to produce a beautiful one from a strategy that loses money in reality. If you are ever tempted to buy something because the backtest looked good, that is precisely the moment to ask what settings produced it.

If your results are inconsistent

The intermediate mistake is running an optimisation, taking the best result, and trading it. Every parameter you optimise increases the chance that what you have found is a coincidence in the historical data rather than a property of the market. With five inputs and a few hundred combinations each, you are guaranteed to find something that looks excellent.

Three habits fix most of this. First, always use forward testing, and judge the strategy only on the reserved out-of-sample section. Second, after choosing your settings, nudge each one up and down by ten or twenty percent; a real edge degrades gently while a fitted one falls off a cliff. Third, run the same settings on a different symbol and a different date range without re-optimising.

Also, go and check your spread and commission settings right now. A very large share of intermediate backtests are run with the tester’s default spread and zero commission, and are consequently reporting a strategy that does not exist. On anything with small targets, entering a realistic cost is often the difference between a rising and a falling equity curve.

If you are experienced

The technical points that matter most: real-tick mode is only as good as the broker’s stored bid/ask history, and coverage is frequently thinner further back than the date range you are permitted to select. Check the tester Journal for history-related messages rather than assuming a clean run. Where real ticks are unavailable, understand that generated ticks are interpolated from M1 bars and that any intrabar logic is being evaluated against a fiction.

On optimisation, the genetic algorithm samples the parameter space rather than exhausting it, which means a rerun with the same inputs can produce a different best set: useful information in itself. Prefer custom criteria that penalise drawdown and trade concentration over raw net profit, and look at the optimisation surface rather than the top row of the results table. A broad plateau of acceptable results is worth far more than a single sharp maximum.

Also account for specification drift. The tester applies today’s contract size, margin requirement, minimum stop level, commission and swap across the entire historical period. Anything sensitive to stop-level constraints or financing cost is being tested against a market structure that did not exist at the time. For carry-sensitive or scalping systems, that alone can invert the sign of the result.

Risk management for this strategy

The specific risk the Strategy Tester creates is false confidence, and false confidence expresses itself as position size. A trader who has seen a smooth backtest equity curve sizes up, because the drawdowns in the report looked survivable. Then the live drawdown arrives with real money attached and behaves nothing like the report.

Protect against it in two ways. Assume the live version of any strategy will have a drawdown considerably worse than the backtest showed, and size so that the backtest’s worst drawdown, doubled, would still be tolerable. And run any new system on a demo account or at minimum size for a meaningful number of trades before it earns full allocation, long enough to include a losing streak, not just a good week.

Keep your position sizing rules independent of the backtest entirely. Whatever the report says, risk a fixed small percentage per trade calculated from the stop distance using the position size calculator. Systems that only look good at sizes you would not otherwise use are not systems, they are leverage.

Where Market Structure Pro fits

An important distinction for anyone testing Market Structure Pro: MSP is an indicator, not an Expert Advisor. It does not place trades, so the Strategy Tester will not produce a profit and loss report for it. There is no equity curve to generate, because it never takes a position.

What you can do, and should, is select MSP in the tester, tick Visual mode, and replay a period of history bar by bar. That shows you how the verdict evolves as the market develops: where it switches to TRADE, where it flags TRANSITION, and how often it sits on NO TRADE through conditions that look tempting on a finished chart. Watching the ranging filter refuse a market in real time is considerably more instructive than reading about it.

This replay also demonstrates the non-repainting behaviour directly. MSP locks its state on the closed bar, so a verdict printed at a given bar stays where it was printed as the replay moves forward. That is the property you cannot verify from a static historical screenshot, and it is exactly what the tester’s visual mode is good for. MSP is decision support only; it is not a signal service and it guarantees nothing.

Frequently asked questions

How do I open the Strategy Tester in MT5?

Press Ctrl+R, or go to <span class="mono">View › Strategy Tester</span>. The tester panel opens at the bottom of the terminal, where you choose the Expert Advisor or indicator, the symbol, the timeframe, the date range and the modelling mode before clicking Start.

What is the difference between real ticks and every tick in MT5?

"Every tick based on real ticks" replays actual bid and ask history recorded by your broker, which is the most realistic option available. "Every tick" generates synthetic ticks by interpolating within one-minute bars, so the intrabar price path is invented. If your strategy reacts intrabar, only real ticks give you a meaningful result, and only if your broker supplies that history.

Does MT5 have a modelling quality percentage like MT4?

No. MT4 reports a modelling quality figure such as 90% or 25% describing how much tick data was interpolated. MT5 does not produce that number; instead you select a modelling mode before the test, and the realism follows from that choice. Check the tester’s Journal tab for warnings about missing history rather than looking for a quality score.

Why are my backtest results different from live trading?

Because the tester models a friendlier market than the real one. It applies a constant spread rather than one that widens around news, it cannot reproduce real slippage or requotes, and commission and swap are often understated or missing from the symbol specification. It also never accounts for the trades you skip, delay or interfere with.

What spread should I use in the MT5 Strategy Tester?

A fixed value at or slightly above your broker’s typical spread, rather than the "Current" option, which locks in whatever the spread happened to be when you pressed Start. On strategies with small targets this setting alone can turn a profitable-looking result into a losing one, so err towards pessimism.

Can you backtest an indicator in MT5?

You can run an indicator in the Strategy Tester’s visual mode and watch it replay bar by bar, which is extremely useful for seeing how it behaves and whether it repaints. What you cannot get is a profit report, because an indicator does not place trades. Testing an indicator means judging its behaviour, not its equity curve.

What is forward testing in the MT5 tester?

It reserves a portion of your date range (a half, a third or a quarter) that the optimiser is not allowed to use. You optimise on the earlier period and then check how the selected settings performed on the reserved period. If the forward section performs badly, the settings were fitted to the past rather than capturing anything real.

Why does my optimisation always find a profitable setting?

Because testing thousands of parameter combinations and keeping the best one will find a good historical curve even in random data. That is a property of the search, not evidence of an edge. The defences are forward testing, checking that nearby parameter values also perform reasonably, and validating on symbols and periods you did not optimise on.

Does the MT5 tester include commission?

Only if the symbol’s contract specification includes it, and on many broker symbols it is set to zero even when you are charged in reality. Check your account’s actual commission per lot and, if the tester shows none, treat every result as overstated by that amount on every trade.

Related reading

• Backtesting a strategy: The wider discipline: what a backtest can and cannot establish.
• Trading strategies: Strategies worth testing, and the conditions each one needs.
• Common MT5 problems: Including missing history and tester errors.
• Try it on demo: The step between a backtest and real money that most traders skip.