The replicator equation ẋᵢ = xᵢ(fᵢ - f̄) governs strategy frequencies in evolutionary game theory, population genetics, and reinforcement learning — its trajectories on the probability simplex converge to Nash equilibria (evolutionary stable strategies), and the Price equation provides a unified mathematical framework for all levels of selection simultaneously.

The replicator equation, derived independently in evolutionary biology, game theory, and learning theory, is: ẋᵢ = xᵢ (fᵢ(x) - f̄(x)) where xᵢ is the frequency of strategy i, fᵢ(x) = Σⱼ aᵢⱼ xⱼ is the payoff to strategy i (with payoff matrix A), and f̄ = Σᵢ xᵢ fᵢ is the mean payoff. The key property: the replicator vector field is tangent to the probability simplex Δⁿ = {x: xᵢ ≥ 0, Σxᵢ = 1}, which is a forward-invariant manifold. THE MAIN IDENTITIES ACROSS FIELDS: 1. EVOLUTIONARY GAME THEORY ↔ NASH EQUILIBRIA: Maynard Smith & Price (1973) introduced evolutionarily stable strategies (ESS) to model animal contests. An ESS is a strategy that, when adopted by all members of a population, cannot be invaded by a rare mutant. Taylor & Jonker (1978) proved that every ESS is an asymptotically stable rest point of the replicator equation. The converse is partial: stable rest points of the replicator equation correspond to Nash equilibria of the underlying game, making the replicator dynamics the natural "equilibrium selection" procedure for evolutionary games. 2. POPULATION GENETICS ↔ REPLICATOR: The Wright-Fisher model in population genetics describes the change in allele frequency xᵢ under selection: E[Δxᵢ] = xᵢ(wᵢ - w̄)/w̄, where wᵢ is fitness and w̄ is mean fitness. This is the replicator equation in discrete time. Fisher's Fundamental Theorem — dw̄/dt = Var(w) (rate of increase in mean fitness equals additive genetic variance in fitness) — is the replicator dynamics in the continuous limit. 3. REINFORCEMENT LEARNING ↔ REPLICATOR: The policy gradient theorem in reinforcement learning (Sutton et al. 2000) updates policy π(a) in proportion to the advantage function A(a) = Q(a) - V: Δπ(a) ∝ π(a) A(a). This is the discrete replicator equation: actions with above-average value become more probable. REINFORCE (Williams 1992), the policy gradient algorithm, is stochastic replicator dynamics. The Nash equilibrium of a game is the fixed point of policy gradient learning. 4. THE PRICE EQUATION — UNIVERSAL SELECTION: Price (1970) derived a covariance equation for evolution at any level of organisation: w̄ Δz̄ = Cov(w, z) + E(w Δz) where z is any character, w is fitness. The first term (Cov(w, z)) captures selection; the second (E(w Δz)) captures transmission bias. The Price equation unifies individual selection (z = allele frequency), group selection (z = group frequency), and cultural evolution (z = meme frequency) in a single framework. Every selection theorem (Fisher's fundamental theorem, secondary theorem of selection, multilevel selection) is a special case of the Price equation. 5. ROCK-PAPER-SCISSORS AS HETEROCLINIC CYCLE: The three-strategy game with cyclic dominance (rock beats scissors, scissors beats paper, paper beats rock) has no interior fixed point in the replicator equation. Instead, trajectories cycle along the boundary of the simplex (heteroclinic cycle) — a non-convergent attractor where no strategy dominates. This predicts that cyclic competition (found in competing bacteria, lizard morphs, neural circuits) will not reach equilibrium but will show persistent oscillations — confirmed experimentally. 6. REPLICATOR-MUTATOR EQUATION IN LANGUAGE EVOLUTION: Adding mutation to the replicator equation gives the replicator-mutator system: ẋᵢ = Σⱼ xⱼ fⱼ Qⱼᵢ - x_i f̄ where Qⱼᵢ is the mutation matrix. This describes language evolution (Nowak et al. 2001): words are strategies, language use is selection, and pronunciation/ spelling changes are mutations. The error threshold in this system (critical mutation rate above which information is not preserved) is the linguistic analogue of Eigen's quasispecies threshold in molecular evolution.

Taylor & Jonker (1978) published the replicator dynamics in Mathematical Biosciences; the connection to Nash equilibria was noted by game theorists later. Population geneticists and evolutionary game theorists develop parallel mathematical frameworks (Wright-Fisher vs. replicator dynamics) without realising they are the same equations. The connection to reinforcement learning (Sutton & Barto, Williams 1992) was made in the ML community without reference to evolutionary biology. Hofbauer & Sigmund (1998) is the definitive reference synthesising the connections, but it is not widely read outside mathematical biology. The Price equation is taught in evolutionary biology but rarely connected to RL theory in machine learning courses.