Reproduce “Algorithmic Aspects of Strategic Trading” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Michael Kearns, Mirah Shi

Michael Kearns and Mirah Shi developed an algorithmic framework for strategic trading using a discrete-time, discrete-volume game-theoretic model, building on Chriss's work. They introduced efficient dynamic programming for best responses and adapted Follow The Perturbed Leader (FTPL) for Coarse Correlated Equilibria (CCE) in an exponentially large strategy space, demonstrating how market impact coefficients influence game structure and revealing that FTPL-derived CCE can approximate stronger equilibria and exhibit high correlation for high permanent impact.

Key insights

  • The strategic trading game can be theoretically decomposed into a weighted mixture of a potential game (from temporary market impact) and a constant-sum game (from a variant of permanent market impact), with the `κ` coefficient controlling the balance.
  • While games dominated by temporary impact are potential games where best-response dynamics converge to Nash Equilibria, the general game with both temporary and permanent impact is not a potential game, leading to potential cycles in best-response dynamics.
  • Efficient no-regret learning algorithms, specifically FTPL with its per-round best-response computation via dynamic programming, can practically approximate Coarse Correlated Equilibria in complex game settings with exponentially sized strategy spaces.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.