How Von Neumann and Morgenstern Formalized Game Theory

In 1944, mathematician John von Neumann and economist Oskar Morgenstern published Theory of Games and Economic Behavior, a groundbreaking work that established game theory as a rigorous branch of mathematics. This article explores how the duo transformed the study of strategic decision-making by replacing traditional calculus with set theory, introducing formal mathematical proofs, establishing the Minimax Theorem, quantifying human preferences through utility theory, and creating frameworks for both zero-sum and cooperative games.

Moving Beyond Traditional Calculus

Before the 1940s, classical economics relied heavily on differential calculus and analogies borrowed from physics. This approach treated decision-makers as isolated actors responding to fixed environments, failing to account for how one person’s decisions directly alter the choices of another.

Morgenstern recognized this flaw in economic theory, while von Neumann possessed the mathematical tools to fix it. They abandoned traditional calculus in favor of set theory, combinatorics, and functional analysis. By doing so, they created a new mathematical language capable of modeling strategic interdependence—situations where an individual's success depends on the choices made by others.

Defining the "Game" Mathematically

To make game theory a strict mathematical discipline, von Neumann and Morgenstern stripped away the recreational context of parlor games like poker and chess, reducing them to abstract mathematical structures. They defined a game using precise, universal components:

  • Players: A finite set of rational decision-makers.
  • Strategies: A comprehensive plan of action a player can take for every possible situation that might arise during the game.
  • Payoffs: Numerically defined outcomes that players receive at the end of the game based on the combination of strategies chosen.

They also introduced two rigorous ways to represent these games mathematically: the "extensive form," which uses decision trees to map out sequential moves, and the "normal form," which uses matrices to represent simultaneous choices and their resulting payoffs.

The Minimax Theorem and Zero-Sum Games

The mathematical bedrock of their work was the Minimax Theorem, which von Neumann originally proved in 1928 and expanded upon in the 1944 book. This theorem addresses two-person, zero-sum games—scenarios where one player's gain is the exact equivalent of the other player's loss.

Previously, it was unclear if a mathematically "correct" way to play such games existed, especially when players needed to bluff or hide their intentions. Von Neumann and Morgenstern solved this by introducing "mixed strategies," where a player assigns a specific probability to each available action rather than choosing a single, predictable move.

Using Brouwer’s fixed-point theorem and convex set theory, they proved that for every finite, two-person, zero-sum game, there is a rational solution. Both players can adopt a strategy that minimizes their maximum possible loss, resulting in a stable equilibrium where neither player can improve their outcome by unilaterally changing their strategy.

Quantifying Preferences with Expected Utility Theory

For game theory to function as a branch of mathematics, payoffs had to be represented as real numbers. At the time, economists argued that human satisfaction (utility) was subjective and could not be measured numerically.

Von Neumann and Morgenstern bypassed this philosophical hurdle by developing Expected Utility Theory. They established four mathematical axioms of rational behavior: completeness, transitivity, independence, and continuity. They proved that if a person's preferences when facing risky outcomes (lotteries) satisfy these four logical axioms, their preferences can be translated into a numerical scale. This allowed mathematicians to calculate and compare payoffs rigorously, turning subjective desires into concrete mathematical variables.

Expanding to Multi-Player Cooperative Games

Real-world strategic interactions rarely involve only two people in a zero-sum scenario. Recognizing this, von Neumann and Morgenstern extended their mathematical framework to \(n\)-person games (involving three or more players) and non-zero-sum games, where mutual benefit or mutual destruction is possible.

They categorized these into cooperative games, where players can communicate and form binding coalitions to maximize their collective payoff. Using set theory, they introduced the concept of the "characteristic function," which calculates the guaranteed minimum payoff a coalition can secure on its own. They also defined "imputations"—mathematical vectors representing how a coalition's total winnings are divided among its members—and established criteria for "stable sets," providing a formal method to predict which alliances would form and endure.