Saturday 10:50 a.m.–11:30 a.m.
Ways to avoid overfitting when developing a trading strategy
- Audience level:
- Industry Uses
This talk is a brief introduce about the main problem when developing a trading strategy. It would also show some solutions to the problem mentioned above, and would give some brief examples how we implement it by Python.
You never know how to develop a successful trading strategy that always gives you significant return, but you definitely should know what would make you doomed to be a failure. Survival bias, looking-forward bias, and backtest overfitting are some of the most common problems. In the brief talk, I would like to share some experience how to quantify backtest overfitting. The examples would be shown in Python toolkit like IPython, pandas, SciPy, NumPy, matplotlib, Seaborn, scikit-learn, and Statsmodels. ( And might be with some R codes ;P)