Optimal Mix Strategy Variance

This game theory question studies optimal play in a zero-sum setting where payoffs form a cyclic, rock-paper-scissors-like structure instead of a simple "bigger is better" ordering. Because local advantages wrap around the action space, the optimal quant prep focus is understanding how equilibrium behavior spreads probability mass across many choices rather than concentrating on a few obvious ones.

It trains your ability to construct and analyze mixed-strategy Nash equilibria in finite games, to detect and remove dominated actions, and to work with symmetry and stationarity conditions until you pin down a distribution over actions. It also exercises probabilistic skills by turning that equilibrium distribution into a concrete variance calculation.

This matters for quant interviews because real trading often resembles adversarial games: you must reason about opponents, equilibrium, and randomness to design robust, unexploitable strategies.

What it tests

In two-player zero-sum games with a finite set of actions and non-transitive dominance (where some choices cyclically beat others), optimal strategies often require randomization over a subset of actions such that no single choice can be exploited. The key is to identify dominated strategies—those that are always worse than some other option regardless of the opponent's play—and eliminate them. What remains is a set of actions that, when randomized over appropriately, make the opponent indifferent to their own choices, ensuring no unilateral improvement is possible. This equilibrium is enforced by the symmetry and cyclical dominance structure, which prevents any deterministic choice from being optimal. The principle holds because any predictable bias can be punished by the opponent, so only balanced randomization over the undominated set is stable.