MT5 Strategy Tester Modes: Why the Same EA Gives Different Results

The Strategy Tester (MT5’s backtesting tool) has several modes that set how finely it simulates price inside each candle. Same EA, same period, only the mode changed, and the result can flip from profit to loss. In our backtests, one system made 3.78 times more profit on simulated ticks than on real ticks, while another differed by less than one dollar. This article explains what each mode sees, and which mode suits which kind of EA.

The four main modes in MT5

The Modelling field in the Strategy Tester has these main options:

Mode	Prices the EA sees	Speed
Every tick based on real ticks	Every tick the broker actually recorded, with each tick’s spread	Slowest
Every tick	Ticks MT5 generates itself from 1-minute bar data	Slow
1 minute OHLC	4 prices per 1-minute bar: Open, High, Low, Close	Fast
Open prices only	The open price of each bar on the tested timeframe	Fastest

A tick is one price change. When the market is busy, the gold price can change many times in a single second.

• Every tick based on real ticks uses the broker’s real tick history and is the closest to reality. Worth knowing: if the chosen period has no real tick files, MT5 generates simulated ticks instead without stopping to ask. Our lab only has real gold ticks from 2026 onward.
• Every tick builds a price path that passes through the Open, High, Low and Close of each 1-minute bar, then fills in the prices in between by itself. The order and timing of those in-between prices are not real.
• 1 minute OHLC has only 4 points per minute. The tester has to guess whether the High or the Low came first, and it uses the spread recorded in the bar, which is the lowest spread of that minute.
• Open prices only sees one price per bar. It suits EAs that act only when a new bar opens. Use it for rough screening.

Why the same EA gives different results

The difference appears when an EA makes decisions inside the bar, such as moving a Trailing stop (a Stop loss that follows profit) on every tick, waiting for a pending order (an order set to trigger at a certain price) to be hit, or adding trades when price moves against it. A guessed or generated price path makes these EAs “see” moments that never really happened.

An EA that decides at bar close and places its Stop loss several dollars away barely cares about the order of prices inside the minute. Every mode gives similar results.

Real test results: from under a dollar apart to profit turned into a blown account

All of these are gold (XAUUSD) in our backtests: the same EA, the same period in 2026, only the mode changed. Backtest results do not predict the future, and these numbers are not a return you should expect. We use them only to see how much the mode distorts results.

System	Comparison	Coarser mode	Real ticks
Bar-close decisions, wide Stop loss (Donchian M15, 0.02 lot, Jan–Jul)	Simulated vs real ticks	+3,222.01	+3,221.74
8-bar box breakout on M1 with a tight Trailing stop (0.01 lot)	Simulated vs real ticks	+3,460.68	+915.30
ICT Silver Bullet style, RR 1:3 (M1, 0.10 lot, 10,000 balance, Jan–Sep)	OHLC vs real	+5,219	+5,156
Grid buying at Fibonacci 50 (M1, 3,000 balance)	Simulated vs real ticks	+27,038	−3,001
Two-sided grid adding trades every tick, no per-trade Stop loss (M1, 3,000 balance)	OHLC vs real	+155,804	about −2,900

The top and bottom rows are the two extremes. The bar-close system differs by just 27 cents. For the grid that adds trades every tick, the OHLC mode says the profit is in six figures, but on real ticks it lost almost the whole balance. For why grids break, read Grid, Martingale and averaging down.

Worked example: how much of the simulated-tick profit disappears

Using the 8-bar box breakout system in 2026, with numbers from the table above:

1. Profit on simulated ticks = $3,460.68; profit on real ticks = $915.30
2. Ratio = 3,460.68 ÷ 915.30 = 3.78 times
3. The part that doesn’t really exist = 3,460.68 − 915.30 = $2,545.38
4. That’s 2,545.38 ÷ 3,460.68 = 73.6% of the simulated-tick profit

Check back: 915.30 × 3.78 = 3,459.83, close to 3,460.68. Correct (the gap comes from rounding 3.78).

Compare the ICT Silver Bullet style: 5,219 − 5,156 = $63, divided by 5,156 gives 1.2%. And the Donchian bar-close system: 3,222.01 − 3,221.74 = $0.27, less than 0.01%.

Backtest results do not predict the future. These numbers only measure how much the mode distorts results. This breakout system has a tight Trailing stop, the type most sensitive to simulated ticks, so almost three quarters of its profit came from price paths MT5 made up.

The 3.78 divisor was measured on one system in one year. Don’t use it to divide the results of other systems or other years. The only way to know is to run the comparison yourself.

ICT on OHLC: how big is the gap?

In ICT tested on gold, 2022–2025 used the OHLC mode because there were no real ticks. So we ran 2026 in both modes to check. The results differed by about $50–600 over nine months (gold M1, 0.10 lot, $10,000 balance). The biggest gap was the RR 1:3 version of the yesterday’s High/Low sweep pattern: OHLC said −998, but real ticks gave −1,614, a $616 difference. The trade count was the same in every row, because with the Stop loss several dollars away, the order of prices within the minute rarely decided the outcome.

So OHLC works for this kind of system. But if a system’s profit is smaller than this margin of error, the OHLC result can’t tell you anything yet. Backtest results do not predict the future.

Two blind spots you can’t see in the report

Rollover spread spikes vanish. The OHLC mode uses the spread recorded in the bar, which is the lowest of that minute. During rollover (when the broker closes the trading day and charges swap), gold’s spread often widens briefly. On real ticks, this can hit small Stop losses, but OHLC doesn’t see it at all. Systems that enter near rollover or use tight Stop losses look better than reality in this mode. For more on spread and small timeframes, read Why small timeframes pay more in spread.

Fake hourly patterns. We split the 8-bar box breakout results by time of day: the same trades in 2026, only the mode changed. On simulated ticks, one afternoon-to-evening window looked best. On real ticks, the same window was the worst. Looking at 4.6 years of simulated-tick results, the differences between time windows looked clear enough to pass a statistical test, while on real ticks no window differed significantly from the others. A time filter built on results like these is tuning to fake data (Why the peak of a tuning run can fool you).

Hourly results, and anything that depends on price timing inside the bar, should be read from real ticks only.

Choosing a mode by EA type

EA type	Screen with	Decide with
Bar-close decisions, wide Stop loss	1 minute OHLC	Real ticks (compare one period to know the margin of error)
Acts only when a new bar opens	Open prices only	Real ticks
Tight Trailing stop, pending stop, small Stop loss	Real ticks	Real ticks
Grid, multiple trades, no per-trade Stop loss	Real ticks	Real ticks
Analysing hourly results or timing within the bar	Real ticks	Real ticks

Checklist before you trust a test result:

1. Check the Journal every time after a run. If it says which date real ticks begin from, the period before that date used simulated ticks.
2. Label the data type on every result. Which years are real ticks, which are OHLC or simulated ticks. Don’t lump it together as “5 years of real ticks.”
3. Run one comparison period. Pick a period with real ticks and run both a fast mode and real ticks. If the results are far apart, don’t use the fast mode for that EA again.
4. Check the spread when the EA enters. If it enters around rollover or uses a tight Stop loss, trust only the real-tick result.
5. Always run the final check on real ticks before deciding anything about that EA.

Even real-tick results are still only the past. Markets change character from one period to the next. Read more in When the market changes character.

Common mistakes

• Choosing Every tick and thinking it’s real ticks. The two modes have similar names, but plain Every tick means generated ticks.
• Choosing real ticks without checking the data exists. For periods with no real tick files, MT5 quietly uses simulated ticks instead.
• Testing grids or tight Trailing stops on OHLC. This mode makes these systems look better than reality more than any other.
• Adding up multi-year results that mix data types. Simulated-tick years can be more inflated than real-tick years, so the total is skewed.