Model=4 Backtest Explained: What "Every Tick Based on Real Ticks" Means in MetaTrader 5

What Model=4 actually means

When you open the MetaTrader 5 (MT5) Strategy Tester, the Modeling dropdown offers several ways to reconstruct how price moved through history. The most accurate of these is labelled "Every tick based on real ticks." In MQL5 community shorthand this is often called Model=4 — the highest-fidelity option.

A tick is a single change in price: a new bid or ask printed by the broker. A real chart is not made of tidy one-minute candles; it is made of thousands of these tiny updates, arriving irregularly, sometimes in bursts and sometimes in near-silence. Model=4 replays the actual recorded tick stream for the symbol — the real sequence of bid/ask updates the broker's server distributed — rather than a simplified or invented approximation of it.

That distinction is the whole point. In Model=4 your expert advisor (EA) is fed the same granular, irregular price flow it would have seen had it been running live during that period. Every value figure published by EudoraLab is produced in this mode, as a MetaTrader 5 Strategy Tester backtest — never live trading, never a forecast, and never a guarantee of future results.

The four MT5 modeling modes, from coarse to exact

MetaTrader 5 offers four ways to model price inside a single bar during a backtest. They differ in how much detail they reconstruct between the open and close of each candle, and that detail is exactly what determines whether an EA's entries, stops and grid fills are realistic.

From least to most demanding:

• Open prices only — the EA is evaluated once per bar, at the open. Fast, but blind to everything that happens inside the candle. Only valid for strategies that act strictly on completed bars.
• 1 minute OHLC — each bar is approximated using just its open, high, low and close. The tester knows the four corners but invents the path between them.
• Every tick (generated) — the tester synthesises a plausible tick sequence inside each minute bar using an interpolation algorithm. Far more granular than OHLC, but the ticks are manufactured, not real.
• Every tick based on real ticks (Model=4) — the tester replays the broker's genuine recorded ticks. No invented path, no interpolation.

The gap between "every tick generated" and "every tick based on real ticks" is the difference between a realistic-looking simulation of the market and a replay of the market that actually happened. For mechanism-sensitive systems, that gap can quietly flatter a backtest.

How the modes compare

The table below summarises the trade-offs. "Intrabar path" is the crucial column: it describes how the tester believes price travelled inside each bar, which is where stops are hit, grids fill and slippage lives.

Note that "slowest" is not a flaw — it is the cost of honesty. A backtest that runs in seconds has almost always thrown away the detail that decides whether a strategy survives contact with real intrabar movement.

Tick quality and what "real-tick data" really means

Selecting Model=4 is necessary but not sufficient. The mode replays real ticks, but the quality and completeness of those ticks depends on the data your terminal has downloaded from the broker. Two concepts matter here.

Modeling quality

At the end of a run, MT5 reports a modeling-quality percentage. A high figure (often quoted as ~99%) indicates the tester had dense, continuous tick coverage for the period. Low figures, or large gaps, mean the tester fell back on coarser data for stretches of history — which can hide or distort what the EA would actually have done.

Broker-specific ticks

Real ticks are broker-specific. Spread, the exact bid/ask, and the timing of updates differ from one broker to another, especially on instruments like gold and indices. A backtest on real ticks from Broker A is not automatically reproducible on Broker B. This is why a credible Model=4 result is always tied to a named broker, symbol and timeframe rather than floating free as an abstract number.

• Completeness — are there gaps in the tick history?
• Spread realism — were real spreads used, or a fixed assumption?
• Symbol match — gold on one feed is not identical to gold on another.
• Period — recent history is usually better-covered than a decade ago.

Why grid and scalping EAs care most

For some strategies the modeling mode barely matters. A daily-bar momentum system that only acts on completed candles — like EudoraAtomic on crypto D1 — is far less sensitive to intrabar fidelity, because it makes decisions once per bar rather than reacting to every tick.

For others, the modeling mode is decisive. Two families are especially exposed:

Scalping and mean-reversion entries

Strategies that enter on small, fast deviations — for instance the gold M5 mean-reversion logic in EudoraAegis, which fires when price strays from its mean under RSI/Stochastic and ATR conditions — live or die on intrabar detail. With coarse modeling, an entry trigger that only existed for a few real ticks can be missed entirely, or a fill can be assumed at a price that never traded.

Grid, recovery-grid and martingale-style systems

This is the sharpest case. EudoraLab is explicit that several of its EAs are controlled grid / recovery-grid / martingale-style systems. They add positions in spaced steps as price moves against the basket, and they carry floating drawdown by design — that is the trade-off these structures accept, and they can incur large losses. None of these EAs is "grid-free" or "martingale-free," and any claim to the contrary would be false.

Because a grid's entire risk profile is defined by where and when each additional level fills, the intrabar path is not a rounding detail — it is the strategy. Consider:

• The spaced grid progression in EudoraFuji on cross-yen pairs, where each added level depends on exactly how far price extended within the bar.
• The 1.5x-step recovery grid in EudoraAegis, whose drawdown depth is set by how many levels filled before mean reversion.
• The leveraged controlled-martingale recovery grid in EudoraAntaeus on the S&P 500, which adds a deeper level each time price falls another StepATR x ATR, up to 12 levels.

Run such a system on OHLC or generated ticks and the simulation can quietly fill levels at convenient prices, understate how deep a basket really went, or skip a stop-out that real ticks would have triggered. The result looks smoother than reality. Model=4 removes that flattery: it forces the grid to fill against the price path that actually occurred.

A controlled-grid backtest on real ticks is the only kind worth trusting, because the grid's risk is the intrabar path. On synthetic ticks, the most dangerous moments — the deep, fast adverse runs — are exactly the ones most likely to be smoothed away.

How EudoraLab uses Model=4 as a reproducibility standard

For "the Operators," Model=4 is not a badge to wave — it is a contract. Every published figure across the four EAs is a MetaTrader 5 Strategy Tester backtest at Model=4 ("every tick based on real ticks"), and each product ships the exact set file, symbol and timeframe so a buyer can reload the same configuration and reproduce the same run on comparable real-tick data.

That is the practical meaning of a reproducibility standard: the headline number is not the deliverable; the recipe is. A representative set of those backtested results:

Reproducibility cuts both ways, and honesty requires naming the limits of even a real-tick backtest:

• A backtest is not the future. All figures are historical simulations. Past backtested performance does not guarantee, predict or imply future results.
• Hindsight can still creep in. Fuji's 11-year edge incorporates regime-masking; unmasked, its profit factor is roughly 0.83, and a representative recent regime (2025-02 to 2026-05) reads nearer +40% at PF 1.42. Real ticks fix the execution model; they do not remove curve-fitting in the strategy.
• Drawdown is real. Antaeus concentrates its returns in a few bounce days — remove the best ~5% of days and Showcase falls to roughly break-even. Atomic's gains cluster in crypto bull regimes. These grid and recovery structures carry meaningful floating drawdown and are not suitable for prop-firm daily/total drawdown rules.
• Broker dependence remains. Your spreads, fills and tick feed will differ from the test broker, so your own results will differ too.

For more on how these tests are framed and what the set files contain, see the FAQ.

Running a Model=4 backtest yourself

If you want to reproduce or stress-test any EudoraLab result — which is exactly the point of shipping set files — the workflow in MetaTrader 5 is straightforward:

1. Open the Strategy Tester (View → Strategy Tester, or Ctrl+R).
2. Select the EA, then the matching symbol and timeframe from the product's set file (e.g. XAUUSD M5 for Aegis).
3. Set Modeling to Every tick based on real ticks.
4. Load the supplied set file so inputs match exactly.
5. Choose the documented date range and starting deposit, then run.
6. Check the reported modeling quality and tick coverage — if it is low, download fuller history before drawing conclusions.

Expect it to be slow, and expect minor differences from the published run because your broker's recorded ticks and spreads are not identical to the test broker's. Small deviations are normal and, frankly, reassuring — they confirm you are looking at a real-tick simulation rather than a frozen marketing screenshot.

None of this is financial advice. Automated trading carries real risk of loss; trade only with risk capital you can afford to lose, and size any grid or recovery-grid system conservatively against its documented drawdown.

The Operators · EudoraLab

Written by the EudoraLab desk — the operators who build, backtest and ship these MetaTrader 5 systems. Every figure we publish is a Model=4 real-tick backtest, packaged with reproducible set files so you can run it yourself. Not financial advice.