Capital Games and Growth Equilibria

Ben AbramowitzarXiv 2025 · DOI · arXiv

In the paper "Capital Games and Growth Equilibria," Ben Abramowitz addresses a fundamental gap between Game Theory and Ergodicity Economics. The research provides a formal framework to derive utility functions from raw payoffs by accounting for how those payoffs evolve over time. By moving the "black box" of utility out of the player's head and into the mathematical description of the game's dynamics, Abramowitz offers a rigorous solution to long-standing paradoxes in economic theory.

The Utility Problem

Standard Game Theory is built on the assumption that players aim to maximize an abstract concept called "utility." However, in real-world scenarios, we observe payoffs in "observables" like money or resources. The "Utility Problem" refers to the difficulty of determining exactly what a player's utility function should be.

For decades, economists have grappled with paradoxes where maximizing the expected value of money leads to ruinous outcomes in the long run. A classic example is the Peters and Gell-Mann (PGM) coin flip: a bet that increases wealth by 50% on heads but decreases it by 40% on tails. While the expected value is positive, repeated play under multiplicative dynamics leads to certain bankruptcy. This paper addresses the lack of a formal mechanism to translate these raw payoffs into utility functions based on the physical or economic dynamics of the environment.

The Capital Games Framework

Abramowitz introduces Capital Games, a generalization of standard game theory where payoffs are measured in units of "capital" rather than utility. A Capital Game is defined as a tuple $(N, W, A, x, D, f)$, incorporating several novel elements:

  • Initial Endowment ($W$): The starting resources of the players.
  • Duration ($D$): The time horizon over which the game is played.
  • Capital Dynamics ($f$): The specific rules governing how capital grows or shrinks over time.

To bridge the gap between dynamics and strategy, the paper introduces Dynamics Linearization. This mathematical technique identifies a function $v$ that transforms complex capital dynamics into a linear form. For example, if wealth grows multiplicatively, the linearization function is the logarithm ($\ln x$). This allows players to maximize their time-average growth rate ($\bar{g}_i$), leading to a new solution concept called Growth Equilibrium.

Key Theoretical Results

The core of the paper is the Equilibrium Correspondence (Theorem 3). This theorem proves that for any "positive capital game" with linearizable dynamics, there exists a corresponding standard game where the Nash Equilibria are identical to the Growth Equilibria. This means that computing a Growth Equilibrium is mathematically equivalent to computing a Nash Equilibrium in a transformed game.

Furthermore, the paper establishes the computational complexity of these games. Finding Growth Equilibria for positive capital games is PPAD-complete, placing it in the same complexity class as finding a Nash Equilibrium in a standard game.

Another significant finding is Pure Strategy Robustness (Theorem 4). Abramowitz demonstrates that Pure Growth Equilibria are invariant to specific dynamics; if a strategy is a best response for one type of monotonic dynamics in a pure-strategy setting, it remains a best response for all others.

Insights and Implications

The Capital Games framework yields several profound insights into strategic behavior:

  • Dynamics Dictate Behavior: The same set of monetary payoffs can lead to completely different equilibria depending on whether the game is played "in parallel" (additive dynamics) or "in sequence" (multiplicative dynamics).
  • Generalizing the Kelly Criterion: The paper effectively generalizes the Kelly Criterion—a formula for optimal betting size in gambling—into a multi-player, game-theoretic setting.
  • Resolving Paradoxes: Using this framework, the author provides a rigorous justification for why a bet with a positive expected monetary gain (like the PGM coin flip) is a "good" bet under additive dynamics but a "bad" bet under multiplicative dynamics.

Limitations and Future Work

While the paper provides a robust foundation, it acknowledges several limitations. The current model assumes deterministic dynamics that are known to the players, whereas real-world environments often involve stochasticity. Additionally, the correspondence theorem relies on the existence of a "linearization" function; future work is required to handle non-linearizable dynamics. Finally, the paper intentionally sets aside psychological factors, such as individual risk aversion, to focus on "ergodic" utility derived from dynamics. Integrating these psychological elements with ergodic theory remains an open and promising direction for future research.

Summary updated

Abstract

We introduce capital games, which generalize the definition of standard games to incorporate dynamics. In capital games, payoffs are in units of capital which are not assumed to be units of utility. The dynamics allow us to infer player utilities from their individual payoffs and linearizable capital dynamics under the assumption that players aim to maximize the time-average growth rate of their capital.
BibTeX
@misc{abramowitz2025capital,
  author = {Ben Abramowitz},
  title = {{Capital Games and Growth Equilibria}},
  year = {2025},
  eprint = {2510.00472},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2510.00472},
}
Figure