The Kelly criterion maximizes the expected logarithmic growth rate of a portfolio. For a single binary bet, the answer is simple: bet a fraction of your bankroll equal to your edge divided by the odds. For a single event with multiple outcomes — a horse race, a multi-way election — the criterion extends to multinomial Kelly, where the optimal allocation across multiple mutually exclusive outcomes maximizes expected log growth.
The paper derives the full multinomial Kelly solution for a single event with implicit state positions. “Implicit state” means the bet payoffs define a complete set of states — each combination of outcomes maps to a specific portfolio return — and the optimization accounts for the portfolio's behavior across all states simultaneously.
Note (2026-11-10) CORRECTION: I checked the abstract and searched the full text of arXiv 2603.13581 (Long). The paper is a short note: it says the optimal formula for a single event with finitely many mutually exclusive outcomes is classical, that the support-selection step is often presented with Lagrange multipliers, and that its contribution is a shorter state-price derivation in which a cash fraction acts as an implicit position (a baseline stake c q_i) in every outcome, giving x_i = (p_i - c q_i)_+ and a greedy threshold rule; it states that nothing in it changes the classical theory. The essay says the paper derives the full solution in closed form where previous treatments used numerical optimization or special cases; that is not what the paper says. The essay's definition of implicit state positions as payoffs defining a complete set of states is also not the paper's.
The solution is closed-form. Previous treatments either used numerical optimization or handled special cases (binary bets, equal payoffs). The general multinomial solution involves a system of equations that can be solved analytically, yielding optimal fractions that depend on the full matrix of payoff-probability relationships, not just individual bet edges.
The key practical insight: in a multi-outcome event, the optimal bet sizes interact. Adding a bet on one outcome changes the optimal bet on every other outcome, because the portfolio's logarithmic growth depends on the joint distribution of returns across all states. Optimizing each bet independently — treating each as a separate binary Kelly problem — is suboptimal and can be significantly wrong when outcomes are correlated through the event structure. The multinomial solution captures the interaction; the binary approximation misses it.