Danish Munir (CIS) and  Divya Krishnan (ESE)



Financial Modeling

A system to test High-Frequency Trading Strategies




Project Advisors:      Professor Abraham Wyner, Statistics Department, Wharton

                                    Professor Insup Lee, Computer Information Sciences


Abstract:                    In the past few years, advancements in technology have enabled traders to be able to carry out trades at ever faster rates. Starting from trades being put physically at the stock exchange, to then being put over the wire, the telephone and finally over the internet, today trades can be executed at a split second’s notice through specialized systems. This rapid maturity of trading engines and the level sophistication of communications technology has resulted in what is known as “high-frequency trading” which represents the ability to place trades at speeds that are only feasible by automated computer-based systems.


This new era of trading brings with it its own set of challenges, and requires a new breed of strategies by investors in search for profits. However given the high volume of trades occurring at each second, the resulting volume of data from such trades is overwhelming and is extremely hard for many investors to analyze, and to test their strategies on. Thus it becomes hard to take a strategy that is born out of intuition and to put it through robust testing before actually taking the risk of investing money using such a strategy.


This project has two objectives: One to provide a platform that allows simulating trades at ultra-high frequencies and allows investors to back-test trading strategies using historical data. The other, is to prove the hypothesis that when trading at very high frequencies and at very small time intervals, the correlation between assets breaks down in a non-linear way, and to demonstrate a strategy that exploits this statistical property to make money. 



1.      Proving or disproving statistically, the hypothesis held by our advisor regarding breakdown of correlation over smaller time intervals

2.      Building a generic system that can be used to test high-frequency trading strategies

3.      Testing our system by employing a rebalancing strategy on a pair of equal-weighted portolios



Project Report


Project Poster