MT5 real-tick backtests: what to check

Start with the question

A fast screening run helps narrow a large candidate set. It cannot answer every execution question. If stops, pending orders or entries depend on price movement inside a bar, inspect that movement in the final verification. Keep screening and verification reports separate and use the same risk basis when comparing candidates.

Read the history, not just the label

MT5 can substitute generated ticks when minute bars lack tick records. Record the selected mode, inspect coverage and retain the journal. A test on broker history also does not reproduce every future fill. Keep an unexplained gap marked as unavailable instead of treating a missing interval as a month with zero profit.

Keep a reproducible evidence bundle

Store the original report, inputs and executable hash together. Include start capital, currency, leverage, fees, dates and server timezone. A screenshot is a convenient summary, but the underlying files are needed to trace a figure back to a completed run. If the executable changes, verify the new one before reusing the old figures.

Checks to make before deciding

• Inventory monthly bars and ticks for every required symbol. Distinguish market closures from unexpected missing intervals, using the broker session calendar rather than calendar-day counts alone.
• Check tick timestamps, duplicate entries, bid/ask validity and impossible prices. Investigate spikes against an independent reference before treating them as either genuine volatility or removable errors.
• Review tester logs for downloaded history, synchronization problems and generated tick substitutes. MT5 can generate ticks where minute bars lack ticks; real-tick mode is not a completeness certificate.
• Check bid/ask spread distributions by hour and date. Highlight rollovers and outliers; an average spread alone can conceal the periods that determine fills and stop execution.
• Use the fast test as a screening tool only if its modeling limitations fit the strategy. Intrabar stops, pending orders or tick-sensitive entries require more detailed verification.
• Select Every tick based on real ticks for final historical verification where suitable broker history exists. Preserve the modeling label and inspect coverage instead of describing all runs simply as real ticks.
• Record broker, symbol, timeframe and strategy input timeframe separately. Chart timeframe may not be the timeframe used by an EA's internal signals or indicators.
• Check start-date inclusion and end-date exclusion in MT5. Record the effective server-time cutoff; a UTC date boundary is not automatically the broker's date boundary.