Direct answer
Responsible backtesting of a mobile trading setup built around MT5 means you treat it as an educational simulation, not a forecast. You define what data and assumptions are used, you include realistic trading frictions, you control for common biases, and you verify results on data that was not used to build or tune the test.
Mechanics and definitions
Backtesting means replaying past market data through the same decision rules you would use in practice, producing hypothetical trade outcomes. “MT5 Mobile” typically refers to using the MT5 application on a mobile device, so the responsible part is usually not the phone itself, but the agreement between (1) how orders are executed in the simulation and (2) how execution would actually occur.
To do that, start by separating stable mechanics from variable conditions:
- Stable mechanics: the logic of your entry/exit rules, indicator calculations (if any), time handling, and the order generation model.
- Variable factors: spread, commissions, slippage, margin constraints, and any differences between simulated fills and real fills.
When you test, state assumptions for every calculation or example, such as the timeframe, the exact data source used for candles or ticks, and how orders are filled (for example, whether they assume you always get the displayed price or whether you apply a conservative fill model).
Evidence and example structure (without assuming outcomes)
A responsible backtest workflow uses checkpoints that you can explain and independently reproduce:
- Data definition: Specify whether you use bar data or tick data. If only bar data exists, note that intra-bar movement is unknown, which affects how “stop” and “limit” style logic can be approximated.
- Cost modeling: Include at least commissions and a model for spread and slippage. Even simple models are better than “zero costs,” but you must document them clearly.
- Bias controls: Avoid testing many variations until you find a favorable one. A basic control is to restrict changes after the first test, or to set aside a final segment of data for confirmation.
- Out-of-sample checks: Run the same logic on a later period that was not used to develop parameters. If results disappear, that is evidence of overfitting or sensitivity to regime.
Material limitation: historical relationships do not establish future results. A backtest can describe what would have happened under its specific assumptions, but it cannot guarantee that the same assumptions will hold later.
Limitations and failure modes
Even a careful backtest can fail due to missing realism or biased evaluation. Common failure modes include:
- Execution mismatch: The simulated fills may differ from real execution because of slippage, partial fills, or liquidity effects.
- Spread and commission drift: Costs can vary over time; assuming a constant cost can distort results.
- Data granularity limits: With fewer data points (for example, only bar opens/highs/lows), the timing of order triggers becomes uncertain.
- Overfitting: Results may reflect tailoring to one market period rather than a stable behavior.
- Non-stationarity: Markets change structure over time, so patterns can weaken or vanish.
Verification and next question
To verify responsibly, you should be able to answer four questions from your own documentation:
- What exact data was used, and what does it omit?
- What exact costs and fill assumptions were applied?
- Which parts of the logic were chosen or tuned using the same data as the evaluation?
- How did results behave when tested on a separate out-of-sample period or with altered (but documented) cost assumptions?
If you want, tell me what kind of historical data you plan to use (bar vs tick), and what order rules you intend to evaluate (for example, market vs limit behavior). I can help you translate that into a checklist of assumptions and robustness checks—without implying any expected trading performance.