Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
support for statistical sampling?
Grant Kiehne
posted

Does Quantopian support statistical sampling of its backtesting database? Using an algorithm running on Quantopian, the idea is to sample and analyze symbols (SIDs) (or combinations, such as pairs), looking for promising candidates. One approach would be to impose an overall run time limit on the algorithm. In a loop, the algorithm would pick SIDs at random and analyze them (e.g. run a backtest). The promising candidates could be logged and reported to the user. Alternatively, the loop could be exited upon finding the first potentially profitable trade, and report the result to the user.

More generally, it would seem that the real potential for Quantopian would be figuring out how to allow users to do large scale analyses cost-effectively (since presumably the industry folks who do this sort of thing for a living have substantial computing power).

3 responses
John Fawcett

@Grant, another great question, thank you.

As I mentioned in reply to your sid question, we are working on universe selection. That will let your algorithm request a broader swathe of trade history, without requiring your algorithm's logic to explicitly use a particular sid in that trade history. We see this as the first logical step in expanding Quantopian to support statistical programming as you describe above.

A second critical step is allowing your algorithm to work with a window of data across that entire universe. We are working on a design for that part of the problem. The goal is to construct a pandas DataFrame for prices, and another DataFrame for volumes of trade events. Each DataFrame would have a TimeSeries for each sid in the universe, and values equal to the price or volume. This would allow your algorithm to use pandas to do the heavy lifting on a trailing window of trade data, rather than just a single event.

In this way, an algorithm would specify its universe, perform statistical programming using pandas, and then make trading enter/exit decisions using logic on individual trade events.

We're really excited about this direction, and as always, would love your feedback.

Disclaimer

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.

Brian Granger

@John It is really different to think about time series analysis in an event driven model. It really shifts the perspective from the entire time series to the current + N trailing samples. It would be great if users could specify the number of trailing data samples that were available at each step and having Pandas DataFrames for that data would make everything very nice. Different trading strategies operate on different time scales and that determines the number of trailing samples that are relevant.

Brian Granger

Summary of my last post = I like the direction you are headed in!

About Quantopian
Careers
Community
Events
Documentation
Academia
Lectures
Status
Facebook
Twitter
LinkedIn
YouTube
Terms of Use
Privacy Policy

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian.

In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.