Tax Genius · India
Backtesting & Strategy Validation
Mission Saptanga methodology for separating research backtests from validated strategies through bias controls, out-of-sample testing, costs, robustness and monitoring.
Research framework published and architecture-reviewed:
Research framework
A backtest is a research instrument. It is not evidence of a deployable strategy until bias, robustness, costs and later-period behaviour are examined.
Backtest versus validation
The first backtest answers whether an idea deserves further investigation. Validation asks whether the apparent relationship survives changes in sample, parameters, costs and market environment. Treating those two stages as the same creates avoidable overconfidence.
Bias controls
The research record should explicitly address look-ahead bias, survivorship bias, universe changes, data revisions and repeated experimentation.
- Timestamp integrity
- Universe reconstruction
- Out-of-sample evidence
- Multiple-testing awareness
- Parameter sensitivity
Economic realism
Turnover, liquidity, fees, slippage and implementation capacity are part of the strategy hypothesis. A frictionless backtest is not a complete portfolio claim.
Lifecycle
Mission Saptanga distinguishes Research, Validated, Paper, Production, Paused and Archived states so that a promising experiment is not presented as a live capability by default.
Methodology
- Freeze the research question and signal definition before evaluating the final test period.
- Separate training, validation and out-of-sample evidence where sample size permits.
- Model transaction costs, turnover and execution constraints rather than reporting frictionless returns.
- Record parameter sensitivity, regime dependence and failure conditions.
Evidence requirements
- Versioned dataset and feature definitions.
- Timestamp discipline that prevents look-ahead leakage.
- Documented universe and survivorship treatment.
- Cost and slippage assumptions linked to the intended market context.
Limitations
- Backtested performance is hypothetical and may not reflect live trading conditions.
- A strategy can degrade after discovery or after market structure changes.
- Public methodology does not imply that any described strategy is live, available or suitable.