Data identity
Source, asset, interval, tested range, coverage snapshot and known gaps.
Research methodology
Trading strategy backtesting methodology.
A practical framework for checking data, defining assumptions and interpreting historical simulations without hiding uncertainty.
Four-stage research record
Keep the full path from question to interpretation.
Each stage records the inputs needed to reproduce the study and understand what a result can and cannot show.
- 01
Define the research question
State the market, interval, rule logic, tested period and comparison metric before looking at results.
- 02
Inspect the historical data
Check source identity, timestamps, coverage, known gaps and freshness for the requested range.
- 03
Record simulation assumptions
Attach starting capital, fees, slippage, position sizing and strategy version to the run.
- 04
Compare and interpret
Read trade-level evidence, drawdowns and neighboring parameter results before drawing a conclusion.
Reproducibility checklist
What stays attached to a research result.
A headline metric is only useful when its data and modeling context remain inspectable.
Rule identity
Saved strategy version, entry and exit logic, parameters and risk settings.
Execution model
Capital, fees, slippage, order assumptions and any omitted market effects.
Evidence trail
Run identifier, metrics, equity and drawdown paths, plus simulated trades and triggers.
Parameter interpretation
Prefer stable neighborhoods to one best cell.
Parameter analysis is a sensitivity check, not proof that an optimum will persist outside the tested sample.
Inspect neighboring values
A broad area of similar outcomes is more informative than an isolated peak surrounded by weak results.
Read more than return
Compare drawdown, risk-adjusted metrics, trade count and stability alongside the selected objective.
Separate discovery from validation
When data permits, reserve unseen periods or later observations for an independent check.
Interpretation limits
What a historical backtest does not establish.
Transparent research makes its uncertainty explicit rather than converting a simulation into a forecast.
- Historical conditions may not repeat and a tested period may omit future market regimes.
- Fees, slippage, fills, liquidity and latency are assumptions that can differ from real execution.
- Repeated strategy or parameter selection can overfit the same historical sample.
- Research outputs are not investment advice, trade execution or a promise of future performance.
Apply the methodology to an inspectable workflow.
Start with the historical data or examine how a versioned rule set becomes a backtest report.