I am a new research scientist at Facebook. In 2018 I finished my PhD in Computer Science at the University of Michigan, where I was advised by Michael Wellman. My research interests include games with endogenous information acquisition, empirical game solving, agent-based modeling applied to the finance, and general machine learning.
Empirical Mechanism Design for Optimizing Clearing Interval in Frequent Call Markets
Several recent authors have advocated for financial markets to move from continuous clearing to discrete or batched clearing, as a way to defeat the latency arms race: the never-ending quest for small advantages in time to access markets. How frequently should such a modern batch auction clear? We conduct a systematic simulation-based investigation on the relationship between clearing frequency and metrics of market quality, such as allocative efficiency, comparing the performance of discrete and continuous auction mechanisms under empirical equilibrium behavior of all participating traders. In effect we perform empirical mechanism design on frequent batch auctions. We find that in a wide array of environments, equilibrium efficiency is improved for small positive intervals but falls off dramatically when there are too few opportunities to trade. The result is a large range of batch frequencies that are near optimally efficient; this range is more pronounced in thick markets.
E. Brinkman, M. P. Wellman
Signal Structure and Strategic Information Acquisition
The ability to gather information can affect outcomes in auctions and other games of incomplete information. We investigate situations where agents have a choice about which information, or signals, to observe, and are informed about the signal choices of others. Our models cover common-value games where agents decide whether to coordinate on observed information, and games mixing private- and common-value components. We find that the dependence structure among available signals can produce qualitatively distinct behaviors in equilibrium, including some cases where strategic agents implicitly collude to acquire less than maximally informative signal combinations.
E. Brinkman, M. P. Wellman, S. E. Page
Understanding Financial Market Behavior through Empirical Game-Theoretic Analysis
Financial market activity is increasingly controlled by algorithms, interacting through electronic markets. Unprecedented information response times, autonomous operation, use of machine learning and other adaptive techniques, and ability to proliferate novel strategies at scale are all reasons to question whether algorithmic trading may produce dynamic behavior qualitatively different from what arises in trading under direct human control. Given the high level of competition between trading firms and the significant financial incentives to trading, it is desirable to understand the effect incentives have on the behavior of agents in financial markets. One natural way to analyze this effect is through the economic concept of a Nash equilibrium, a behavior profile of every agent such that no individual stands to gain by doing something different. Some of the incentives traders face arise from the complexities of modern market structure. Recent studies have turned to agent-based modeling as a way to capture behavioral response to this structure. Agent-based modeling is a simulation paradigm that allows studying the interaction of agents in a simulated environment, and it has been used to model various aspects of financial market structure. This thesis builds on recent agent-based models of financial markets by imposing agent rationality and studying the models in equilibrium. I use empirical game-theoretic analysis, a methodology for computing approximately rational behavior in agent-based models, to investigate three important aspects of market structure. First, I evaluate the impact of strategic bid shading on agent welfare. Bid shading is when agents demand better prices, lower if they are buying or higher if they are selling, and is typically associated with lower social welfare. My results indicate that in many market environments, strategic bid shading actually improves social welfare, even with some of the complexities of financial markets. Next, I investigate the optimal clearing interval for a proposed market mechanism, the frequent call market. There is significant evidence to support the idea that traders will benefit from trading in a frequent call market over standard continuous double auction markets. My results confirm this statement for a wide variety of market settings, but I also find a few circumstances, particularly when large informational advantages exist, or when markets are thin, that call markets consistently hurt welfare, independent of frequency. I conclude with an investigation on the effect of trend following on market stability. Here I find that the presence of trend followers alters a market’s response to shock. In the absence of trend followers, shocks are small but have a long recovery. When trend followers are present, they alter background trader behavior resulting in more severe shocks that recover much more quickly. I also develop a novel method to efficiently evaluate the effect of shock anticipation on equilibrium. While anticipation of shocks does make markets more stable, trend followers continue to be profitable.
A DAG layout library based around d3.
Math Genealogy Explorer
A simple SPA for displaying notable mathematical ancestors with links to their English Wikipedia pages.
A command line utility for generating plots according to the principles of Principiae.
Simple rust command line tool for computing statistics on a stream of utf-8 floats.
Great Lakes Distance
A visualization of how far you can be from the great lakes in any state, created using D3.