Pseudo-Mathematics and Financial Charlatanism
The 2014 Notices of the AMS paper by David H. Bailey, Jonathan M. Borwein, Marcos López de Prado and Qiji Jim Zhu is the foundational reference on backtest overfitting. It proves that high simulated performance is easily achievable after testing a relatively small number of strategy configurations, derives a Minimum Backtest Length below which an impressive in-sample Sharpe ratio is statistically expected even with zero true skill, and shows that under memory effects (serial dependence) overfitting produces negative expected out-of-sample returns. It appears in this vault as the canonical evidence that positive Markov-model backtests must be treated as weak unless the number of trials is disclosed and an overfitting diagnostic is reported.
Connections
- Overfitting in Quantitative Trading — defines, source: https://www.ams.org/notices/201405/rnoti-p458.pdf
- Probability of Backtest Overfitting — precedes, source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2326253
- David H. Bailey — proposes_model, source: https://www.ams.org/notices/201405/rnoti-p458.pdf
- Marcos López de Prado — proposes_model, source: https://www.ams.org/notices/201405/rnoti-p458.pdf