Rebellion Research is New York based hedge fund with $20 million under management across 70 clients. What’s unique about the firm is how they’re using machine learning technology to evaluate stock from a value perspective, a growth perspective, a momentum perspective, and even a macro perspective. In the interview below, Rebellion Research’s Chairman Spencer Greenberg discusses how he feels his company is well-positioned for bringing machine learning and AI based asset management to investors. Rebellion Research is a Register Investment Advisor.
Read on for more important insights from Rebellion Research!
insideBIGDATA: Rumors in machine learning circles indicate that most data scientists have developed their own investment algorithms, some do quite well as individuals. What made you decide to form a hedge fund instead of keeping the technology to yourself?
Spencer Greenberg: If you’ve created an algorithm you believe will outperform its benchmarks, it may make sense to start by just using it to trade your own account. When we got started developing our strategy and forming Rebellion Research we really wanted to create a long term business. The founders have a mutual respect for each other and we each brought something unique to the table. That kind of collaborative creativity is what has allowed our hobby to become a business. Investing is difficult and there is no proven method to succeed so we want to find a better approach and we want to share it. Initially, we had just hoped to create a research product, but with the performance of our Machine Learning, it became clear we should offer it to investors.
insideBIGDATA: I understand you use 30 different factors that can affect a stock’s performance. Do you do continued feature engineering and reevaluate your algorithm to accommodate changing market forces?
Spencer Greenberg: We use a lot more than 30 factors, but for a given stock at a given moment in time having 30 available is reasonably typical.
We go through phases of development where we’ll focus on different aspects of our Machine Learning. We haven’t been in a feature engineering phase for a bit. You want to continually ask yourself: given what we know, what’s the lowest hanging fruit for improvement? That being said we do look to form new strategies with our algorithms to better suit the interest of investors. For instance we test strategies that are specific to a geographical region or are made up of one asset class, bonds being the most recent example.
insideBIGDATA: Can you comment on the technology stack used to implement your machine learning application?
Spencer Greenberg: We use python with numpy. Early on we considered which language would be best to use (considering, for example, ruby and java), and decided to go with python due to it being very high level and requiring relatively little time to implement ideas, while still having strong numerical computing support.
insideBIGDATA: What is your long-term strategy for your fund? Are there any new big data technologies you plan to experiment with?
Spencer Greenberg: Our long term strategy is to grow Rebellion Research to a household name. We want to bring machine learning powered asset management to everyday investors. So often you see the people with the most resources keeping the best technologies for themselves, we see our approach as the leading edge of the investment field and we want to take a very active role in that expansion. Technologically speaking we are currently experimenting with a variety of alternative approaches to risk optimization, using cvxpy for rapid iteration, to see how they compare to our current approach. As far as technologies, we’ve considered using pandas for some of our work, and will likely do so in the future. Currently, we are serving clients on 5 continents.
Sign up for the free insideBIGDATA newsletter.
Speak Your Mind