What is a good profit factor for an EA?

Above 1.3 is respectable and above 2.0 on a multi-year test is unusual enough to be worth checking carefully. But profit factor flatters grid and martingale strategies, which win small very often and lose large rarely, so read it beside maximum drawdown rather than on its own.

What modelling quality do I need for the test to be meaningful?

On MT5, use real ticks where the broker provides them. "1 minute OHLC" invents the path inside each bar, which matters enormously for any strategy with a stop or a pending order close to price. If a seller shows a test at 90 per cent modelling quality on MT4 or 1 minute OHLC on MT5, the result is an estimate, not a measurement.

Why does relative drawdown matter more than absolute drawdown?

Absolute drawdown is in currency and depends on the starting balance, so it cannot be compared between tests. Relative drawdown is a percentage of equity, which is what your account actually experiences, and it is the number to compare against your own tolerance.

How long should a backtest be?

Long enough to contain at least one period the strategy was not designed for. Five years is a reasonable minimum for a strategy meant to run continuously, and any test that avoids March 2020, the 2022 rate cycle and a central bank surprise has quietly selected its own weather.

Can a backtest be honest and still be useless?

Yes, and this is the common case. A test run with a fixed 1 pip spread, no commission and no swap can be arithmetically correct and still describe an account nobody can open. Costs decide whether a small edge survives, and the tester uses whatever you tell it.

How to Read an MT5 Strategy Tester Report

A backtest is an argument, not a proof. The MT5 report gives you about thirty numbers and a picture, and roughly six of them decide whether the argument holds. Here is what to read, in the order to read it, and the two settings at the top of the page that can make everything below them meaningless.

Last updated 18 September 2026

Start at the top of the page, not the bottom

Before any result means anything, check the header of the report. Three lines there can invalidate everything below.

Symbol and period. Is it the pair and timeframe you intend to run? A result on one pair says little about another.
Modelling. On MT5 this should say Every tick based on real ticks. Anything else is simulated price movement inside each bar.
Spread. A test run at a fixed spread that is tighter than your broker's is not your result. Commission and swap are set in the same place and are often left at zero.

If any of those three are wrong, stop reading the report and ask for a better one.

The six numbers that matter

1. Maximum relative drawdown

Read this before profit. It is the deepest fall from a peak, as a percentage of equity, and it is the closest thing in the report to "what this felt like to hold".

A strategy showing 15 per cent drawdown over five years is describing a normal bad month. One showing 60 per cent is telling you that at some point the account was worth less than half its peak, whatever the final line says.

2. Total net profit, against that drawdown

Profit alone is meaningless without the risk it took. The useful form is the ratio: net profit divided by maximum drawdown. Above 2 over several years is solid. Below 1 means you carried more risk than you earned.

3. Profit factor

Gross profit divided by gross loss. Above 1.3 is respectable; below 1.1 leaves nothing once real costs arrive.

The caveat matters for the EAs sold in this niche: a grid with a multiplier produces a high profit factor almost by construction, because it wins often and small. Read it next to drawdown, never alone.

4. Expected payoff

The average result per trade, in currency. If it is smaller than your spread plus commission for that pair, the strategy has no edge left after costs, however good the equity curve looks.

5. Number of trades

Under about 100 trades, the statistics are noise. Thirty trades can look magnificent by luck. This is also the check on a test that was run over five years but only traded in one of them.

6. The equity curve, read for shape

Look at the picture, not the endpoint. A curve that climbs in small steps with shallow dips is a strategy with an edge. A curve that is a smooth line with two vertical cliffs is a grid: it is telling you exactly what it does, and that the cliffs are the real risk.

Also watch for a long flat stretch, which usually means the strategy stopped trading for months, and a curve that only turns upward at the end, which often means parameters were fitted to the most recent data.

The two signs of a curve-fitted EA

Curve fitting means the settings were tuned until the past looked good. It produces reports that are impressive and predictions that are worthless.

The first sign is a suspiciously exact set of inputs. A stop loss of 253 pips and a grid step of 21 are odd numbers, and odd numbers usually come from an optimiser. That is not automatically wrong, but it means the test you are shown is the best of many, and the best of many is not an average outcome. Ask to see the same settings tested on a different pair or a different year.

The second is an optimisation graph with a single spike. If the EA is profitable at a 21 pip step and loses at 20 and 22, the setting has been fitted to noise. A robust parameter sits on a plateau: it works across a range, and the neighbours of the chosen value work too. In the MT5 optimiser, look at the 2D graph and prefer a broad hill to a needle.

Run your own, properly

Open the Strategy Tester (Ctrl+R), pick the EA, the symbol and the timeframe.
Set modelling to every tick based on real ticks, and the date range to at least five years.
Set the spread to your broker's real average, not to "current", and add commission in the account settings.
Use the deposit and leverage you will actually trade.
Run it, then run it again over a different five years and compare. Two windows disagreeing is more informative than one window agreeing with the seller.

Any product on the MQL5 Market can be downloaded as a free demo and tested exactly like this before you pay, with no time limit. There is no reason to take a screenshot on trust.

What a backtest can never tell you

It cannot show slippage on your account, requotes, how your broker behaves in fast markets, or whether the licence server is reachable from your VPS. It also cannot use MetaTrader's economic calendar, so any news filter reports itself as off during a test.

Those belong to a small live run, which is why demo works, live does not exists. And if the settings you are reading about are a grid, grid vs martingale shows what the lot ladder does to the numbers above.