Wes McKinney - backtesting with pandas

Wes' slides from the Boston Algorithmic Trading Meetup.

Transcript

1. pandas for algo trading research / data analysis Wes McKinney @wesmckinn Boston Algo Trading Meetup @ Quantopian 2013-10-10 Tuesday, January 22,
2. Background • Research and portfolio management at AQR • pandas library for Python • Working on a new project (stay tuned) Tuesday, January 22,
3. Book • In print now! • ~470 pages • IPython • NumPy • pandas • matplotlib • Case studies Tuesday, January 22,
4. pandas • http://pandas.pydata.org • Rich data tool built on NumPy • Used for general purpose data manipulation and analysis • Reduces data munging headache • Fast, convenient time series capabilities Tuesday, January 22,
5. pandas • In heavy production use in many places: finance, web analytics, ... • Generally much better performance than other open source alternatives (e.g. R/xts) • Still much more work to do Tuesday, January 22,
6. Rationale • At minimum: close usability gap with R (but don’t stop there) • Build more integrated, productivity-focused tools • Make Python one of most desirable language for {finance, statistics, data analysis} Tuesday, January 22,
7. Why pandas for algo research? • Broad time series support • Very fast vector operations tailored for large time series • Robust resampling • Multidimensional data wrangling Tuesday, January 22,
8. Thanks! • Follow me on Twitter: @wesmckinn • GitHub: wesm • Blog: http://blog.wesmckinney.com • Exciting things planned for 2013 Tuesday, January 22,