Tuesday, January 17, 2023
Title: ADJOINT ALGORITHMIC DIFFERENTIATION (AAD): How to better hedge financial risks, crack some of the puzzles of condensed matter and much more with upside-down derivatives
Speaker: Luca Capriotti, Credit Suisse and Columbia University
Luca works in the Quantitative Analysis and Technology (QAT) department in New York where he is the Global Head of Quantitative Strategies Credit, and he is responsible for both front office and capital models. Previous to this role, he was the global head of Quantitative Strategies for Credit and Structured Notes; he was the EMEA head and the US head of Quantitative Strategies Global Credit Products; he worked in Commodities in New York and London, and he was part of the cross-asset modeling R&D group of QS in the London office.
Luca is also visiting professor at the Department of Mathematics at University College London, and Adjunct Professor at Columbia University, at the Departments of Mathematics and Industrial Engineering and Operations Research. His current research interests are in Credit Models, Computational Finance, and Machine Learning, with a focus on efficient numerical techniques for Derivatives Pricing and Risk Management, and applications of Adjoint Algorithmic Differentiation (AAD), which he has helped introduce to Finance and Physics, and for which he holds a US Patent. Luca has published over 70 scientific papers, with the top 3 papers collecting to date over 1000 citations (h factor 26, i10 factor 48).
Prior to working in Finance, Luca was a researcher at the Kavli Institute for Theoretical Physics, Santa Barbara, California, working in High-Temperature Superconductivity and Quantum Monte Carlo methods for Condensed Matter systems. He has been awarded the Director’s fellowship at Los Alamos National Laboratory, and the Wigner Fellowship at Oak Ridge National Laboratory.
Luca holds an M.S. cum laude in Physics from the University of Florence, and an M.Phil. and a Ph.D. cum laude in Condensed Matter Theory, from the International School for Advanced Studies, Trieste.
Adjoint Algorithmic Differentiation (AAD) is a computational technique that, despite being known in its modern form since at least the 1960’s, has become mainstream only in the last decade or so when it was “re-discovered” in Finance about 15 years ago. In this talk, I will describe what makes AAD one of the most important innovations in financial risk management and how the same ideas can be applied in other fields whenever computing accurately and efficiently a large number of derivatives is beneficial.
Thursday, January 19, 2023
Title: Optimal Turnover, Liquidity, and Autocorrelation
Speaker: Gordon Ritter, Ritter Alpha LP and Columbia University
Gordon Ritter is an Adjunct Professor in the MAFN program and founder and CIO of Ritter Alpha LP, a registered investment adviser running systematic absolute-return trading strategies across multiple asset classes and geographical regions. Before Ritter Alpha, he was a senior portfolio manager at GSA Capital and a Vice President in the Statistical Arbitrage Group at Highbridge Capital Management (HCM). Gordon completed his PhD in mathematical physics at Harvard University and his Bachelors’ degree with honors in Mathematics from the University of Chicago. While at Harvard, he published several papers in the areas of quantum field theory, differential geometry, quantum computation and abstract algebra. His current research is on portfolio optimization and statistical machine learning. Notable publications include “Optimal turnover, liquidity, and autocorrelation,” with @Bastien Baldacci of @ Université Paris Dauphine – PSL and @Jerome Benveniste of @New York University, Risk, 2022, and “Machine learning for trading,” Risk, 2017. In recognition of the latter publication, Professor Ritter was named Buy-Side Quant of the Year in 2019.
The steady-state turnover of a trading strategy is of clear interest to practitioners and portfolio managers, as is the steady-state Sharpe ratio. In this article, we show that in a convenient Gaussian process model, the steady-state turnover can be computed explicitly, and obeys a clear relation to the liquidity of the asset and to the autocorrelation of the alpha forecast signals. Indeed, we find that steady-state optimal turnover is given by gamma * sqrt(n+1) where gamma is a liquidity-adjusted notion of risk-aversion, and n is the ratio of mean-reversion speed to gamma. The steady-state portfolio size and information ratio can also be given in closed form, and thereare also explicit formulas available in the multi-asset case with nontrivial correlation structure.
Tuesday, January 24, 2023
Title: Semi-Systematic Macro Investment
Speaker: Andres Jaime Martinez
Andrés Jaime is an Associate Portfolio Manager at Capstone, based in New York. His main responsibility is to assess macro trends in Emerging Markets by relying heavily on a systematic and quantitative approach in order to come up with actionable investment ideas.
Previously, Andres was the head of Global Macro Quant & FXEM Volatility Strategy for Morgan Stanley. In addition, he led the LatAm Macro Strategy team for over four years, and was ranked Analyst of the year (1st place) by Institutional Investor in LatAm FX and Rates strategy in 2020, the last year he led the team.
Andrés joined Morgan Stanley in 2017 from Barclays, where he focused on G10 and LatAm local markets research. Prior to that, he worked for Bank of Mexico (Banxico), where he held several managerial positions in strategic and tactical asset allocation, and FX and commodities trading.
He is a regular contributor to the Reforma and El Financiero newspaper in Mexico and has lectured in finance and econometrics in graduate and undergraduate courses at ITAM.
Andrés holds an M.A. in mathematics of finance from Columbia University in New York, a Graduate Certificate in Machine Learning from Cornell University, and a B.A. in economics from ITAM University in Mexico City.
Quantitative models are not only useful as inputs into systematic strategies, but can heavily improve the performance of a discretionary portfolio. I will go through a few examples on the applications of machine learning and more classical econometric techniques into macro trading.
Thursday, January 26, 2023
Title: VOLATILITY AND ARBITRAGE
Speaker: Ioannis Karatzas, Columbia University
Ioannis Karatzas is the Eugene Higgins Professor of Applied Probability at Columbia. He was the driving force in the founding of the Mathematics of Finance MA (MAFN) program back in 1996/97 and has been a strong supporter of the program ever since. He earned his PhD from Columbia University in the City of New York in 1980, did his post-doc at Brown University, then returned to Columbia as a faculty member. He works, publishes, and advises Ph.D. students in probability, stochastic control, mathematical economics and finance. His book “Brownian Motion and Stochastic Calculus”, co-authored with Steven Shreve of Carnegie Mellon University and first published in 1987, is the standard reference in the field of Stochastic Analysis and has helped educate several generations of students and researchers. His recent book “Portfolio Theory and Arbitrage: A Course in Mathematical Finance”, published in 2021 and co-authored with Constantinos Kardaras of The London School of Economics and Political Science (LSE), may represent the ultimate formulation of the theoretical underpinnings of arbitrage pricing. Complementing his academic work, Professor Karatzas has had a long-standing association with INTECH Investment Management in Princeton, NJ
Can the market portfolio be outperformed? If not, why? If yes, under what structural conditions? over which time-horizons? by what portfolios? Can said conditions and portfolios be described in terms only of observables, without resorting to any model assumptions? Questions such as these lead to some pretty interesting mathematics — both in probability theory and in differential geometry, the flow of curves by their curvature (as in the seminal work of Richard Hamilton). We shall discuss some results in this vein, and suggest open problems. (Joint work with E.R. Fernholz and J. Ruf.)