The MT5 Strategy Tester report summarises a single run: how many trades the robot made, the ratio between what the positive trades added and what the negative ones took away (profit factor), the average net result per trade (expected payoff), how far the account fell from its peak (drawdown) and a Sharpe the tester computes itself. Those five numbers take a minute to read. What the report leaves out matters almost more: it does not say how many configurations you tried before this one, which costs it assumed, whether the data had holes, or whether any part of the period was kept out of the optimisation. This article explains each number, the ones that get misread and what to add before trusting the result.
What the report contains and where it comes from
The report is saved from the Backtest tab: right-click, Report, HTML or Open XML (older builds: Save as Report). At the top is the header with expert, symbol, period, inputs, initial deposit and leverage; below it, the results and the balance and equity curve.
Everything describes one run, with one configuration, over one date range. The guide on exporting the MT5 report shows the exact file step by step.
The five numbers to look at first
Total Trades: how many times the robot entered and exited. With few, the other numbers hang on two or three large results. Note the years the period covers too.
Profit Factor: the sum of the positive trades divided by the sum of the negative ones. Expected Payoff: the net result divided by the number of trades, in the account currency; it grows with the lot even though the strategy is the same.
Drawdown: there is balance drawdown and equity drawdown, each absolute, maximal and relative. For a real account or a challenge look at the maximal equity one, which includes floating losses. The fifth, the Sharpe, gets the next section.
The numbers that are easy to misread
The percentage of winning trades misleads on its own. A grid or a martingale closes almost everything in the positive and stores the loss in one large trade: compare the largest loss trade with the average profit trade. If one negative trade wipes out dozens of positive ones, the percentage is useless.
The report's Sharpe does not say which returns it used or whether it is annualised, so it is not comparable with a Sharpe computed on daily returns, like the one in Rigor's calculator. It is only useful for comparing runs in the same tester.
Tick mode and modelling quality
MT5 has four tick modes: every tick, every tick based on real ticks, 1 minute OHLC and open prices only. Without real ticks, the tester fabricates the price path inside each minute from the bars, and the stops are filled on that invented path. With real ticks it fills in generated ones where they are missing; the report shows the total, not which were real.
1 minute OHLC and open prices only are for exploring, not for the final report: they do not see the price inside the bar or know whether the stop or the take profit was touched first. For sharing, use real ticks and note the mode.
What the report never says
How many configurations you tried. The report describes the last run; the earlier ones stay in the optimisation XML. With three years of daily returns and no real edge, the best of 10 configurations shows a Sharpe of about 0.9 by luck alone; the best of 100, 1.5; the best of 1,000, 1.9.
Which costs it assumed, or on what data. There is no cost line: the spread was the setting, the commission the broker server's and the slippage only the execution delay you configured. It does not check duplicate bars, frozen prices, gaps or spikes either.
Whether there was an out-of-sample. Nothing shows that the period was split: if you chose the parameters while looking at these same dates, the result is in-sample, however good it looks.
What to add before trusting it
First, write down what the report omits: optimisation passes, tick mode, spread and commission, and the dates optimised. Export the optimisation XML together with the report; the guide on the MT5 optimisation explains how.
Second, repeat the final run with real ticks and with double and triple the cost per trade; note at which cost the net result reaches zero. Third, run it on dates you did not use to optimise and report that result separately, even if it is worse.
Fourth, compare the Sharpe with what luck would give with your years and your tries: Rigor's calculator is free and needs no signup. Rigor reads both files and measures, among other things, the significance of the Sharpe, the configurations tried, the result at 1x, 2x and 3x costs and 35 data red flags, each tagged Measured, Declared or Not measured.
FAQ
Is a Sharpe of 2 in the tester report good?
It depends on how many configurations you tried and how many years the period covers: with two years and 500 tries, the best configuration with no real edge shows about 2.2 by luck alone; with one year and 200 tries, about 2.8.
What does History Quality 100% mean?
That no one-minute bars were missing and none had a volume of 1 with differing OHLC, the only thing MT5 counts as incorrect data. It does not say whether those bars carry frozen prices, false spikes or duplicates, or whether the ticks in each minute were real or generated.
How many trades does the report need to be reliable?
The report does not answer that: it depends on the years covered and the tries before it, not only on the count. With 1,000 tries, about 3.3 years of data are needed for luck alone to fall under a Sharpe of 1.8. Put your years and your tries into the calculator.