/mt5-robot-tester

MetaTrader 5 Command-Line Strategy Tester Reference

Use this reference to automate the MT5 Strategy Tester without its GUI and to understand the keys and enums used by this skill. Verify details marked (verify) against the installed MT5 build because exact report formats can vary.

Launch

terminal64.exe /config:C:\path\to\tester.ini

• ShutdownTerminal=1 runs the test and closes the terminal automatically, so the process ends when the backtest finishes. The orchestrator waits with a timeout.
• /portable uses the data folder next to the executable instead of the folder under %APPDATA%.
• The terminal needs broker tick data downloaded for Model=4 (real ticks), or the test will fail or use incomplete data.

[Tester] Section

Model

Optimization

OptimizationCriterion

[TesterInputs] Section

Specify a fixed EA parameter as follows:

StopLossCoef1=1.0

Specify an optimization range with value||start||step||stop||Y syntax:

GannHiLoPeriod1=43||43||4||129||Y

• The final Y or N field enables or disables optimization for that parameter.
• In Round 3, this skill ranges one parameter at a time and fixes every other parameter at its current best value before continuing to the next one.

Verified Report Behavior in Build 6061

The following behavior was reproduced with MetaTrader 5 build 6061 on an FTMO-Demo account:

• Report= must be a relative name without a path or extension. MT5 ignores an absolute path and writes <name>.htm plus the <name>.png chart at the terminal data-folder root. The orchestrator supplies a relative name and collects the file from that folder.
• Encodings differ by report type. Backtest .htm reports are UTF-16; optimization XML is UTF-8. The reader detects the BOM and NUL-byte density.
• Deals table columns are localized. A Spanish report uses Fecha/Hora rather than Hora for the time column and includes a Balance column. The parser accepts these labels to reconstruct monthly and yearly balance history and months-to-new-high.
• Optimization=3 does not populate the XML. The generated <name>.symbols.xml contains a header and zero rows; the 73 per-symbol results exist only in the binary Tester\cache\*.opt cache. Round 1 therefore runs one Optimization=0 backtest per symbol from common.symbols instead of relying on the XML.
• Single-parameter optimization works. Optimization=1 emits SpreadsheetML with one row per combination and is used by Round 3.
• ShutdownTerminal=1 closes the terminal after completion. The runner waits for that exit rather than treating the first report-file appearance as completion because optimization reports are written incrementally.

Caveats

• 32 ms delay: no documented INI key controls the random ping-based delay. The terminal usually inherits the last GUI setting. Configure it once in the Strategy Tester delay controls (verify).
• Encoding: INI files are written as UTF-8. Some builds may require UTF-16 when names or comments contain non-ASCII text.
• Market Watch: Optimization=3 tests exactly the visible Market Watch symbols. Add or remove symbols there to control that universe.
• Terminal discovery: --terminal-path takes precedence, followed by $MT5_TERMINAL_PATH, config.terminal_path, and common Program Files installations.