Quantitative Investing and the Limits of (Deep) Learning from Financial Data

22 Pages Posted: 8 Mar 2018 Last revised: 7 May 2018

J.B. Heaton

University of Chicago Law School; Conjecture LLC

Date Written: March 2, 2018

Abstract

The idea of quantitative investing - using powerful computing power and algorithms to trade securities - inspires both awe and fear. Reality is less impressive. With a tiny handful of exceptions, most quant funds have been unimpressive. I explore some limits of quantitative investment, with a focus on the promise - or lack thereof - of techniques from deep learning and artificial intelligence. These limitations help explain the disappointing performance of many quant strategies and cast doubt on the promise of artificial intelligence techniques for improving returns. The main problem is that financial market data is unlike the data that machine learning works well on in computer vision, speech recognition, and natural language processing. While deep learning and artificial intelligence are changing the world in many ways, they are unlikely to generate fortunes for investors, who will continue to remain best-served by inexpensive and passive index products that themselves will be augmented by machine learning techniques to drive costs even lower.

Keywords: quantitative investment management, hedge funds, deep learning, artificial intelligence, index funds

JEL Classification: G02, G23

Suggested Citation

Heaton, J.B., Quantitative Investing and the Limits of (Deep) Learning from Financial Data (March 2, 2018). Available at SSRN: https://ssrn.com/abstract=3133110 or http://dx.doi.org/10.2139/ssrn.3133110

J.B. Heaton (Contact Author)

University of Chicago Law School ( email )

1111 East 60th Street
Room 608
Chicago, IL 60637
United States

Conjecture LLC

Chicago, IL
United States

HOME PAGE: http://conjecturellc.com

Register to save articles to
your library

Register

Paper statistics

Downloads
307
rank
90,824
Abstract Views
798
PlumX