Why your EA needs a real-tick MT5 backtest
A professional MT5 backtest is one run in the Strategy Tester with "Every tick based on real ticks", real spreads, a realistic execution delay and enough history to include bad markets. Anything less can make a losing EA look profitable — most often scalpers, grids and EAs with tight stops. This guide explains why, in plain language, and what to check before you trust any MT5 backtest report.
What an MT5 backtest actually is
A backtest replays old prices through your expert advisor and records every trade it would have taken. MetaTrader 5 does this in the Strategy Tester (View → Strategy Tester, or Ctrl+R). The result is a backtest report: net profit, drawdown, profit factor, every trade, and an equity curve. If you want to know what each number means, read how to read an MT5 backtest report.
The catch is in one setting most people never touch: Modelling. It decides how the tester invents the price movement inside each bar — and that is exactly where most EAs make or lose their money.
The four modelling modes in MT5, compared
MT5 offers four ways to rebuild price history. Only the first one uses ticks that really happened.
| Modelling mode | What the tester feeds the EA | Trust it for |
|---|---|---|
| Every tick based on real ticks | The tick history your broker actually recorded, with the real bid/ask spread at each moment | Every EA, especially scalpers, grids and tight stops |
| Every tick | Ticks generated by an algorithm from one-minute bars | Slow trend EAs with wide stops, as a first look |
| 1 minute OHLC | Only open, high, low and close of each one-minute bar | Quick rough checks and optimisation passes |
| Open prices only | Only the opening price of each bar | EAs that only act on a new bar — nothing else |
"Every tick" sounds like the real thing, but it is not: the ticks are made up to fit inside each minute's high and low. The tester does not know whether the price touched the high first or the low first, how wide the spread really was, or how fast the price moved. A real-tick test does not have to guess.
What generated ticks hide from you
- The order of the high and the low. If your stop loss and take profit both sit inside the same minute, a generated test must decide which was hit first. Choose wrong and a loser becomes a winner. EAs with small targets live in exactly this zone.
- Spread spikes. Spreads widen at the daily rollover, around news and in thin markets. Real ticks carry the real bid and ask; generated ticks smooth that away. A scalper that targets a few points can look excellent until the real spread is included.
- Gaps and fast markets. Weekend gaps and news spikes jump past stop levels. Real ticks show the price you would really have got; a smooth generated path can fill you right at your stop.
- Grid and averaging behaviour. Grids open more orders as price moves against them. A slightly different path inside a minute changes how many orders open and at what price — and therefore the drawdown that decides whether the account survives. See how grid EAs work.
Rule of thumb: the shorter the EA's trades and the tighter its stops, the bigger the difference between a generated test and a real-tick test.
History quality: what 99% and 100% really mean
Every MT5 report shows History Quality. MetaQuotes defines it as the percentage of correct versus incorrect one-minute data in the test period. A high number is necessary, but it is not proof that every minute used real ticks: where the broker has no tick data for a stretch of time, the tester fills it from minute bars without stopping the test. Two practical consequences:
- Start the test where your broker's real tick history starts. Older periods on many brokers have gaps, and a test that runs through them is partly generated.
- Different brokers, different data. The same EA on the same dates can give a different result on another broker's ticks and spreads. A report always describes one data set, not the market in general.
Spread, commission and execution delay
A backtest is only as honest as its costs. Three settings matter most:
- Spread. In real-tick mode the spread comes from the recorded ticks. In the other modes it comes from bar data, so it is an average rather than the spread at the moment of the trade.
- Commission. Raw-spread and ECN accounts charge per lot. The symbol settings of the account you test on decide whether commission is included at all.
- Delays. MT5 can add an execution delay (fixed, or random within a range) between the EA's order and the fill. Zero delay is the perfect world; a few dozen to a few hundred milliseconds is closer to a retail account, and scalpers react to it the most.
What makes a backtest report "professional"
Put together, a backtest you can rely on has all of these:
- Modelling: every tick based on real ticks, on dates where real ticks exist.
- Real costs: recorded spread, commission if your account pays it, and a realistic delay.
- Enough time and enough trades: months that include a trend, a range and a news shock, and a few hundred trades rather than a few dozen.
- Your settings: the exact .set file you will trade, not the seller's best-case inputs.
- One report per timeframe you might run, because an EA that only works on one timeframe is fragile.
- Out-of-sample dates: a period the EA was not tuned on. This is also the best defence against "history reader" EAs — see the backtest red flags.
How to get a real-tick MT5 backtest report
You can run one yourself: open the Strategy Tester, select the EA, the symbol and the dates, set Modelling to Every tick based on real ticks, choose a delay, load your .set file in the Inputs tab and press Start. Expect it to take a while — real ticks are millions of price updates.
If you would rather not tie up your computer, or you want an independent run on your exact settings, use the MT5 backtest report service. You upload the .ex5 and your .set file, choose the symbol, dates (up to 365 days), timeframes (D1, H4, H1, M15, M5, M1), delay, deposit and leverage. The test runs in an isolated MT5 with every tick based on real ticks where the history supports it, and you receive one genuine MT5 report per timeframe in Telegram. Timeframes that cannot run are reported as failed with the reason — nothing is estimated.
No EA yet? Build one in the free EA Builder, or find one in the marketplace or the free library, then backtest it before a demo account. The guide how to test a free EA on demo covers the next step.
Backtest, then demo, then small live
A real-tick backtest tells you whether a strategy could have worked on that data. It cannot promise the future, and it cannot model your broker's live slippage perfectly. The sensible order is: real-tick backtest → at least two weeks on a demo account → the smallest live size your broker allows. Skip a step and you are guessing with money.
Frequently asked questions
What is the most accurate backtest mode in MT5?
Every tick based on real ticks. It replays the tick history the broker actually recorded, including the real spread at each moment, instead of generating ticks from bars.
Is "Every tick" the same as real ticks in MT5?
No. "Every tick" generates artificial ticks from one-minute bars. Only "Every tick based on real ticks" uses recorded ticks.
Does 99% history quality mean the backtest is perfect?
No. History quality measures how much of the one-minute data is correct. Where the broker has no ticks, MT5 fills the gap from minute bars, so start tests where real tick history begins.
Why does my EA make money in the tester but lose on demo?
The usual causes are generated ticks instead of real ticks, ignored spread or commission, zero execution delay, inputs tuned to the tested dates, or a test period that missed bad market conditions.
How long should an MT5 backtest be?
Long enough to include a trend, a range and a volatile news period, and to produce a few hundred trades. For most EAs that means several months to a year of real ticks.
Can I get a real-tick MT5 backtest without running it myself?
Yes. The metatraderea.shop backtest service runs your .ex5 and .set file in an isolated MT5 with every tick based on real ticks and delivers one report per timeframe to Telegram after payment.