Mathematical Finance & Financial Data Science Seminar
Optimal Execution, Order-Placement Tactics, and Hamiltonian Dynamics
Speaker: Jerome Benveniste (joint work with Gordon Ritter), NYU Courant, M.S. Mathematics in Finance
Location: Warren Weaver Hall 1302
Date: Tuesday, October 3, 2017, 5:30 p.m.
Synopsis:
Schedule for Tuesday, October 3
5:10 PM: Registration
5:30 PM: Presentation
6:30 PM: Networking
7:00 PM: The END
Abstract
Institutional investors, such as hedge funds and banks, will frequently seek to make large trades on several correlated assets in order to obtain certain desired portfolio characteristics. These orders are then often given to an agent, either another department within the same institution or another firm, who has the task of executing them while incurring as few costs as possible, usually by taking advantage of microstructure effects that produce short-term predictability in asset returns. This agent will then break the original orders into smaller orders that can be traded more or less instantaneously in the market and will decide, at each moment, how to place the order in the market: they can, for instance, issue a market order, or a limit order at the inside market, or improve the inside market by one tick. The decisions of this agent, however, generally do not fully take into account the objectives of the principal, and this can, in some situations, lead to results that are highly suboptimal from the point of view of the principal. For example, the agent may choose to issue passive orders because they seem the cheapest option; then a sudden market move could cause all of the passive orders on one side (say, all the buy orders) to be filled while those on the other side remain unfilled. This will lead to an unhedged portfolio and could lead to considerable losses. In this talk, we will describe a method, using ideas from Hamiltonian mechanics, by which the principal may continuously communicate their preferences to the agent in the form of “virtual” short-term forecasts that can be naturally combined with the agent’s microstructure signals, leading to an improved order-placement strategy that aligns more closely with the principal’s objectives. The method will be illustrated by comparing the simulated performance of order-placement strategies on a particular day with a large market move.
Bio
Jerome Benveniste is an instructor in the Mathematics in Finance master's program at NYU. Before his retirement in 2014, he was a member of the Quantitative Trading Group at Highbridge Capital Management, LLC for twelve years, the last six as Managing Director and Portfolio Manager. He was involved in nearly every aspect of Highbridge's quantitative business, including forecast generation, risk modeling, transaction cost modeling, and optimization Jerome holds a Ph.D. from the University of Chicago and an A. B. from Harvard University, both in mathematics.
Notes:
Open to anyone. Sign in at the lecture hall after you arrive.