Why an MT5 Backtest Can Differ From Live Trading
An MT5 backtest and a live account can diverge before an order is even sent. The tester runs with selected history, price modelling, delay and account settings. The live account uses the quotes, permissions and execution available at the time. Choose one decision that differs and trace it from signal through broker response. The first mismatch tells you where to look next.
Steps to check
- Write down the EA build and inputs for both runs. Record the full broker symbol, timeframe, account type, test interval and clock used in the exports. Align the timestamps before comparing candles or events. Two symbols with a similar name may use different histories or trading specifications.
- Reproduce the tester setup. Save the modelling mode, execution delay, deposit settings and cost assumptions. Check which price data were available for the test dates. A comparison using different tick sources, spreads or commissions answers a different question from replaying the live decision.
- Check whether the strategy detected the signal and formed a request on both sides. A missing request points toward inputs, data, filters or permissions. If a request exists, compare its side, volume, requested price and submission time before inspecting the broker response. Save these as separate fields.
- Match the broker response and any deals to their orders. Record rejection, partial fill or no fill distinctly. For filled orders, compare executable quotes at submission, fill prices, sizes and timestamps. Keep price differences, commission, fees and swap in separate columns until you reconcile them.
- Change one assumption at a time and rerun the same narrow window. Preserve the original run and list which events still disagree. Use an exported sample or demo record to check mechanics before broadening the comparison. A match establishes agreement for those records under those settings.
Worked example
Hypothetical buy: the tester fills at 1.10020. A live request for the same signal is sent when the observed ask is 1.10025, and its deal fills at 1.10028. The total price gap is 0.00008. Of that, 0.00005 is already present between the tester fill and the live reference quote. The live fill adds another 0.00003 relative to that quote. Check timestamps before assigning a cause.
Frequently asked questions
Does a live rejection mean the entry logic failed?
It means the live order did not execute as requested. First verify that the strategy sent the expected request, then read the broker's return code.
Will testing on real ticks reproduce the same fills?
Real ticks can improve the historical price input. The test still uses its own execution and account assumptions, which may differ from the live broker's actual response.
Can I start with the profit difference?
Use it to identify a period to investigate. Then align actual requests and deals, because different trade counts, sizes or costs can produce similar net differences.
Related guides
Related video
Related walkthrough: trading costs in backtests