The MT5 Strategy Tester offers several modelling modes. For scalping EAs, the choice between "Every tick" and "Every tick based on real ticks" can change results completely.
Every tick: ticks are generated from M1 bars using an algorithm. Price paths within each minute are synthetic.
Every tick based on real ticks: uses actual historical ticks from the broker, including real spreads.
Why it matters for scalping:
Synthetic ticks can create unrealistic fills at exact highs and lows.
Spread is fixed or simplified in generated modes, but varies in real ticks.
Stop and target order of execution within a bar may be wrong with generated ticks.
Real-tick history is limited by what the broker provides; some traders import tick data from external sources into a custom symbol.
Have you compared both modes for a scalping EA, and how different were the results?
FXS Prop Desk wrote: Every tick based on real ticks: uses actual historical ticks from the broker, including real spreads.
Real ticks fixes the price path, but it still fills every order instantly at the tick price, so there's no latency, no requote and no slippage unless you add them. For a scalping EA that's often the bigger gap between tester and live. MT5 has a delay option in the tester settings, including a random delay, which is a cheap way to see how sensitive the EA is to execution. Also worth checking how far back the broker's real tick history goes and whether it has gaps. If part of the test period has missing real ticks, look for notes in the tester journal, because mixed data can quietly flatter part of a backtest.
I compared both modes on a small EURUSD scalper, 3 pip target and 4 pip stop, same twelve months. Every tick gave a profit factor around 1.5, real ticks brought it to about 1.1, and most of the difference came from trades where target and stop sat inside the same minute.
Generated ticks invent the path inside the bar, so a bar touching both levels gets resolved tidily. On real ticks a good share of those flipped to losers.
On top of Fairman's delay point, check whose spread you're testing. Real ticks carry the spread of the account type the history came from, so if you test on a standard feed and trade raw plus commission, both the spread and the cost are wrong at once.
I now pull history from the same account type I trade, and check that commission shows in the tester deals. If not, I use a custom symbol and set it myself.
Tick density is the quiet one. Older months are often much thinner on some feeds, and an EA that reacts tick by tick behaves differently on sparse data, so I compare tick counts per day before trusting a long test.
Have you tried imported ticks on a custom symbol, and did they agree with the broker's real ticks?
Fairman, I tried the delay setting on the same EURUSD scalper after posting. With a fixed 100 ms delay the profit factor on real ticks slipped from about 1.1 to 1.04, and the random delay option took it to 0.98, so the edge did not survive anything worse than an ideal fill.
One thing worth knowing about how the delay works is that it fills at whatever tick exists once the delay ends. That captures price drift but not depth, so a news spike still fills your whole size at one price.
For that reason I add a crude slippage rule in the code for testing only. Any market order placed within two minutes of a high impact event gets 0.5 pips added against me, which is roughly what my live log shows.
The test I now run first is the cheap one. If an EA cannot stay above a profit factor of 1.2 on real ticks with a 100 ms delay, I do not spend time on a forward test at all.
FXS Prop Desk, does the desk ask for tester settings when someone submits an EA for review? A report without the modelling mode and delay listed tells you very little.
