Broker-Trader Partial Information Nash-Equilibria

Xuchen Wu, Sebastian Jaimungal · 2024

Read Broker-Trader Partial Information Nash-Equilibria by Xuchen Wu, Sebastian Jaimungal (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-17712.txt.

Introduction

Xuchen Wu and Sebastian Jaimungal study a game between a broker and a better-informed trader, in a 2024 paper. The trader knows a signal about future prices and trades several assets through the broker in a dealer market. The broker then offloads those assets on a public exchange, where the trading itself moves prices.

The broker sees only overall prices and cannot tell real trends from random noise. Both sides aim to grow expected wealth while penalizing risk. Using convex analysis, the authors characterize the Nash equilibrium for this partial-information contest and prove it exists and is unique. The equilibrium turns out to solve an unusual system of forward-backward stochastic equations with two different information flows. For short time spans they prove a unique solution exists, and they show the solution can be approximated by polynomials in the impact strength, with controlled error.

Catalog record

CREATORXuchen Wu, Sebastian Jaimungal
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE132 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-17712

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Broker-Trader Partial Information Nash-Equilibria Thanks: SJ acknowledges support from the Natural Sciences and Engineering Research Council of Canada (RGPIN-2024-04317, RGPIN-2018-05705) Xuchen Wu, Sebastian Jaimungal Abstract We study partial information Nash equilibrium between a broker and an informed trader. In this setting, the informed trader, who possesses knowledge of a trading signal, trades multiple assets with the broker in a dealer market. Simultaneously, the broker offloads these assets in a lit exchange where their actions impact the asset prices. The broker, however, only observes aggregate prices and cannot distinguish between underlying trends and volatility. Both the broker and the informed trader aim to maximize their penalized expected wealth. Using convex analysis, we characterize the Nash equilibrium and demonstrate its existence and uniqueness. Furthermore, we establish that this equilibrium corresponds to the solution of a nonstandard system of forward-backward stochastic differential equations (FBSDEs) that involves the two differing filtrations. For short enough time horizons, we prove that a unique solution of this system exists. Finally, under quite general assumptions, we show that the solution to the FBSDE system admits a polynomial approximation in the strength of the transient impact to arbitrary order, and prove that the error is controlled. K eywords Nash equilibrium \cdot trading signals \cdot partial information \cdot differing filtrations \cdot FBSDE 1 Introduction In this article, we study the Nash equilibrium between a broker and their client (an informed trader) in a partial information setting. The informed trader trades multiple assets with the Broker in an over-the-counter market, while the Broker also trades in a lit exchange where their trading activities have both transient and instantaneous impact on the prices in the lit market as in the diagram below 1 1 1 While the diagram shows noise traders who interact with the broker, we omit them from our analysis as they do not add any additional insights to the problem setting. . informed noise Broker \eta \upsilon Lit Market \nu Figure 1: Blue print of how informed trader and broker interact with each other and the lit market. The informed trader has privy to a private trading signal about the assets’ drifts, while the Broker does not. Our setup is similar to the one in [ 9 ] , where the authors study a setting with an informed trader and Broker trading a single asset and both have access to the market’s full information, including the asset’s stochastic drift as a private trading signal. Our setup differs from [ 9 ] in two key aspects. Firstly, instead of only trading a single asset, we consider trading multiple (correlated or even co-integrated) assets. Secondly, we assume the broker has access only to asset prices rather than the full market information. Both the broker and the informed trader aim to maximize their (penalized) expected wealth from trading activities over a fixed time horizon by choosing trading strategies based on their respective information content. This leads to a partial information setting for the Broker, but a full information setting for the informed trader. We use variational methods to characterize the optimal response of each agent. Based on this optimal response, we prove uniqueness and existence of Nash equilibrium between the broker and the informed trader in the small time setting. As a result, we obtain a filtered system of forward-backward stochastic differential equations (FBSDEs) that characterizes the Nash equilibrium. One of the earliest works on partial information equilibria in the trading setting is the seminal work of [ 22 ] , who studies how a market marker (MM) sets prices in the presence of noise traders and an informed trader who has a private signal (not available to the MM) regarding the future price of an asset. Other early works on partial information include [ 20 ] , who study optimal portfolio allocation with latent Ornstein–Uhlenbeck returns and [ 23 ] who study model uncertainty in the context of portfolio optimization and the optimal allocation of assets. More recently, and more directly related to this study, [ 19 ] studies optimal trading in a discrete-time, infinite-time horizon setting, where there is an unpredictable martingale component, and an independent stationary (visible) predictable component – the alpha component. The question of how trading signals affect market making decisions is studied in [ 11 ] , while [ 15 ] studies how differing trading signals alter trading decisions. The role that stochastic price impact plays on optimal trading with trading signal is investigated by [ 17 ] . Further, [ 16 ] proposes models for optimal execution when the drift is estimated/filtered from prices. The case of trading signals with general transient impact kernels is investigated by [ 1 ] . [ 5 ] studies the optimal execution problem when stochastic price impact of market orders and the arrival rates of limit and market orders are functions of the market liquidity process and [ 7 ] incorporates contextual bandit algorithms into trading decisions. Furthermore, [ 13 ] analyze how to optimally trade with latent factors that cause prices to jump and diffuse in a single-trader model. Motivated by [ 13 ] , [ 12 ] generalizes the model by considering a large collection of heterogeneous agents and combines Nash equilibrium and partial information setting. [ 14 ] further extends [ 12 ] by incorporating differing model beliefs among the agents. As in [ 12 ] and [ 14 ] , here we use a convex analysis approach and characterize the equilibrium as solution of FBSDE. In contrast to these earlier works, in our setting the broker and the informed trader have different filtrations on which their strategies are adapted. In addition to equilibrium and partial filtration, this paper ties into the field of optimal trading strategies with market signals. There are several studies in this line of research. The early work of [ 8 ] derives an optimal execution strategy when the order-flow from market participants influences both midprices and execution costs – order-flow in their paper induces a trading signal. [ 10 ] studies a broker-trader model where the broker uses the informed trader’s flow to learn price signals and derives optimal strategy which combines inventory management and signal-driven speculation. [ 6 ] combines filtering theory to estimate unobserved toxicity and variational methods in the unwinding of order flow. Contemporaneous to the first version of this paper, [ 3 ] study a related setting involving an informed trader (with a Gaussian signal) and a broker. They do so in Stackelberg like setting, and assume from the outset that the broker uses a Kalman-Bucy filter to filter the trading signal. In contrast, we provide precise mathematical characterization of the Nash equilibria without making such assumptions and do so in the multi-dimensional setting. The remainder of the paper is organized as follows. In section 2 , we introduce the notations, the model, and the performance criteria. In section 3 , we show that the broker’s and the informed trader’s performance criteria are strictly concave functionals in terms of their respective trading strategies, and compute the Gâteaux-derivatives of the criteria. Finally, in section 4 , we characterize the Nash equilibrium as the solution of a filtered system of FBSDEs and prove its existence and uniqueness for small

Read the complete text