Domaine
Backtesting et validation
Ce qu’un test historique peut et ne peut pas montrer, ses biais et les protocoles qui le maintiennent rigoureux.
Rechercher et filtrer dans cette listeDébutant
What a backtest can prove, and what it can't
A backtest shows how a precisely defined rule would have behaved on past data under stated assumptions. It cannot show the rule will behave that way again.
- Débutant
- Backtesting et validation
- Explication
- 3 min
Bientôt disponible : 32
- Backtest, paper trading and forward testing
- Your first backtest, step by step
- Reading a backtest report: metrics that matter
- Equity curves and drawdowns: reading the shape
- Bar replay: rehearse without hindsight
- Costs, spreads and slippage in a backtest
- Backtest sanity checklist
- Overfitting in one picture
- In-sample vs out-of-sample
- Reading a Monte Carlo fan
- Parameter sensitivity: plateaus vs peaks
- Walk-forward analysis: testing as time passes
- Stress-testing a rule
- Why an LLM is not a backtest
- Look-ahead bias: the backtest that read tomorrow
- Survivorship bias: the symbols that vanished
- Data checks before any backtest
- Benchmarks and null strategies: beat something simple
- Fill assumptions: close, next open or intrabar?
- Regime analysis: when did the rule work?
- Backtest autopsy: five ways a great result was wrong
- Overfitting simulator: the best of N random rules
- Keep a record of every strategy you rejected
- Monte Carlo drawdown simulator
- What walk-forward efficiency does and doesn't say
- Backtest bias quiz: spot the flaw
- The deflated Sharpe ratio, made intuitive
- Many strategies, one dataset: selection bias
- Probability of backtest overfitting (PBO)
- A backtesting protocol checklist
- Purged and nested validation
- Reproducible research: logs, versions, seeds