Is the history complete enough?
Review the intended interval, gaps, duplicated records, discontinuities, and changes in data availability.
Data & methodology
Data before models. Method before metrics.
A backtest is conditional evidence. Its value depends on the market data, simulation rules, costs, execution model, experiment design, and the questions asked after the run.
01 / DATA
Know what the simulation is made from.
For tick-sensitive systems, bar data can hide the intrabar path that determines whether an order, stop, or state transition would have occurred. Where appropriate and available, research therefore uses real tick histories and records the source and test configuration.
Real tick data is not a guarantee of realism. Broker feed differences, gaps, symbol specifications, trading sessions, time zones, spread history, and data corrections can all change the test.
02 / DATA CHECKS
Inspect before optimising.
The exact checklist varies by instrument and strategy. These are the recurring questions.
Are timestamps interpreted correctly?
Document time zone, daylight-saving changes, session boundaries, and any conversion applied to the source.
What do Bid, Ask, and spread represent?
Confirm the available fields, spread behaviour, precision, and the price used for charts versus execution.
Do symbol settings match the test?
Record contract size, point value, margin rules, trading sessions, stops level, and other relevant specifications.
What changed between source and test?
Track filtering, normalisation, resampling, gap handling, custom-symbol imports, and dataset versions.
Can the input be identified later?
Keep enough source and configuration information to understand which dataset produced a result.
03 / TESTING PROTOCOL
Escalate the test as the idea survives.
- 01
Specification check
Define rules and expected behaviour before evaluating the output.
- 02
Implementation diagnostics
Use traces, targeted scenarios, and visual inspection to confirm that the code follows the specification.
- 03
Baseline simulation
Run a controlled reference test with documented data, platform settings, costs, and parameters.
- 04
Robustness checks
Vary parameters, periods, market conditions, and reasonable execution assumptions to look for fragile dependence.
- 05
Holdout or forward evaluation
Where the research design supports it, reserve unseen data or later observations for a separate evaluation.
- 06
Review and archive
Record the result, limitations, code and data context, and the decision to continue, revise, or stop.
A BACKTEST CAN SUPPORT
- Whether the coded rules behaved as specified in the simulation
- How results changed across the tested data and assumptions
- Where trades, risks, or losses were concentrated
- Which components merit deeper investigation
A BACKTEST CANNOT PROVE
- Future profitability or a guaranteed return
- That historical market structure will persist
- That live fills, latency, and costs will match the model
- That an untested failure mode does not exist
04 / RESEARCH NOTE
Real ticks and generated ticks are not interchangeable.
Our plain-English note explains how MetaTrader 5 test modes change the evidence a backtest contains.
Read the note