Select the right context
A result depends on rules and on the data used to study them. Check the symbol, dates, timeframe and displayed coverage before running a test.
Select a period
Duration shortcuts make it easier to request a week, month or another window. After loading, review the dates actually available. A requested window and the data coverage may differ.
- Select the relevant market.
- Choose a period or specific dates.
- Wait for loading and inspect its status.
- Check the retained start, end and timeframe.
- Run a test and verify the same information in its report.
Avoid repeating identical requests while loading. If it fails, read the message before changing the duration: availability rather than window size may be the issue.
Distinguish candles and ticks
| Data type | Useful information | Question to consider |
|---|---|---|
| OHLC candles | Open, high, low and close for a period | Does the exact sequence inside the candle matter? |
| Ticks | A more detailed series of price observations | Do coverage and test assumptions suit the question? |
More detailed data does not automatically improve a strategy. It can support a closer study of certain questions, subject to data quality and the displayed assumptions.
Inspect coverage and quality
Look for missing-data information, unexpected dates and warnings. A quote interruption can reflect the market calendar, whereas a gap during a session may deserve investigation. Examine the context rather than assuming either a defect or perfect continuity.
Some indicators need prior history for their initial context. If early trades appear later than expected, review the available period and settings before changing the signal.
Review cost assumptions
The test may offer fields such as spread, commission and slippage. Check each unit and whether it refers to an order, volume or trade. A zero value needs justification, especially for a strategy targeting small movements.
Comparison record:
Symbol and timeframe: as shown in the interface.
Report dates: effective dates, not only the selected shortcut.
Data: selected dataset and coverage.
Costs: available fields and values used.
Versions: identify baseline and variant separately.Use several research periods
Use one period to develop a hypothesis and another to check whether observations appear elsewhere. If you repeatedly change the strategy after looking at the second period, it also becomes research data.
Look at different contexts: directional movement, sideways markets and varying volatility. The aim is to understand where the strategy appears useful and where it becomes sensitive, rather than demand a favourable result everywhere.
Continue with Read a backtest and Compare in AI Lab.