1. Holes in the price history
Tick-precision and M1 simulation build the price path from M1 bars. A missing minute is a stretch with no path at all, so a stop or limit inside it fills at the next price that exists and not where it would have triggered. A higher-timeframe chart rarely notices, and a tick-precision backtest does. Check how much of your M1 history is missing, fill only short holes, leave long outages visible, and test on both the original and the repaired series. GapSmith does the fill into a separate symbol.
2. An indicator that rewrites its newest bar
A backtest only sees closed bars. If an indicator recomputes the bar that is still forming, the live signal can appear and disappear while the backtest never saw it. Only closed bars should advance an indicator’s state.
3. A model or input that changed after validation
Two implementations of the same neural network, one in each platform, will not stay identical. Train once, write the SQX model and the matching MT5 model from the same run, and have each program read its own. Retraining over a live chart should be a decision, not a side effect.
4. A calendar on the wrong clock
MetaTrader reports calendar times in the broker server’s timezone, which may be a fixed offset or may shift with daylight saving. An hour of error is not rounding: a fifteen-minute blackout then covers a quiet stretch beside the release and misses the release itself. Measure the offset against releases whose New York wall-clock time never changes, and do not trust a setting.
5. External data that is current in one place and frozen in the other
The Strategy Tester reads whatever data was exported, and a live terminal extends it forward. If the two disagree about a recent value, you are trading something the backtest never saw. Re-derive a window of recent history before every write, compare it with the research, and refuse to update if they differ.
6. A contract that differs between the two
StrategyQuant X keeps its own description of each instrument: tick size, point value, minimum stop distance, swap, commission and trading session. If those differ from what your broker’s terminal reports, the same strategy costs one thing in the tester and another on the account. Export the terminal’s values, compare them with the SQX broker profile, and fix the rows that differ. SpecSmith reads the terminal and changes nothing in it; the article StrategyQuant X vs MT5 results: check the broker profile has the detail.
Six questions before going live
- Is my M1 history continuous where the test relies on it?
- Does every indicator advance only on closed bars?
- Is each model or data file the same file in both programs?
- Was every timestamp converted with a measured offset and not an assumed one?
- Does anything refuse to trade if its input goes stale?
- Do tick size, point value, stop distance, swap, commission and session match the broker’s terminal?
None of this guarantees a result. The workflow page shows where each check sits, and the parity checklist in the documentation lists what to compare.