Evolutionary Game Theory: Why Populations Settle, Not Individuals

Strategies as populations, not players

Classical game theory asks what a single rational player should do. Evolutionary game theory, developed chiefly by John Maynard Smith and George Price in the early 1970s, asks a different question: if a whole population of individuals each carries a fixed, inherited strategy, and better-scoring strategies simply reproduce faster, which strategies survive? There is no rationality, no lookahead, no learning required -- only differential reproduction, exactly as in ordinary natural selection. A strategy that scores well against whatever else is currently common leaves more copies of itself in the next generation, and the population's mix of strategies shifts accordingly.

Hawk-Dove: why aggression doesn't take over everything

The classic example is the Hawk-Dove game, modelling an animal contest over a resource. A Hawk always escalates to a fight; a Dove always displays and retreats if the opponent escalates. If two Doves meet, they split the resource peacefully. If a Hawk meets a Dove, the Hawk takes everything. If two Hawks meet, they fight, and the winner takes the resource but the loser pays a real injury cost.

payoff to row player, resource value V, injury cost C  (C > V):
              vs Hawk           vs Dove
Hawk:      (V-C)/2               V
Dove:         0                 V/2

If C > V, an all-Hawk population is unstable: every contest between two Hawks costs more (in expected injury) than the resource is worth, so a rare Dove mutant that avoids fights entirely does better on average than the surrounding Hawks and can invade. But an all-Dove population is also unstable, since a single Hawk mutant wins every contest uncontested. The population instead settles at a mixed equilibrium -- some fraction playing Hawk, the rest Dove -- where both strategies earn exactly the same average payoff, so neither can profitably increase or decrease further. This is the evolutionarily stable strategy (ESS): a strategy mix that no rare mutant strategy can invade.

Replicator dynamics: the equation behind the animation

The mathematical engine driving these populations toward their ESS is the replicator equation. If x_i is the fraction of the population playing strategy i, and f_i(x) is that strategy's payoff given the current population mix, then strategy i's share grows exactly in proportion to how much better it does than the population average:

dx_i/dt = x_i * (f_i(x) - f_avg(x))

x_i      current frequency of strategy i
f_i(x)   payoff to strategy i against the current population mix
f_avg    average payoff across the whole population

A strategy that beats the current average grows; one that underperforms shrinks; and a strategy exactly at the average payoff stays flat -- which is precisely the condition satisfied at an ESS, the fixed point where every strategy present earns the same payoff and the population mix stops changing.

The Prisoner's Dilemma: why defection can still lose locally

In the one-shot Prisoner's Dilemma, defecting is the individually dominant strategy no matter what the opponent does, so a well-mixed evolutionary population converges entirely to defection -- cooperation cannot survive when every pairing is against a random stranger. Placing the same game on a spatial grid where individuals only interact with their immediate neighbours changes the outcome: cooperators that happen to cluster together mostly interact with other cooperators, sharing the mutual-cooperation payoff and out-reproducing an isolated defector sitting at the edge of the cluster, who quickly runs out of naive cooperators to exploit. This spatial reciprocity is one of several well-studied mechanisms (alongside repeated interaction and reputation) that can sustain cooperation despite the individual incentive to defect.

Rock-Paper-Scissors: when nothing stays still

Not every game has a stable point. In Rock-Paper-Scissors payoffs, each strategy beats one and loses to another in a full cycle, so no single frequency mix ever satisfies the ESS condition of equal payoffs simultaneously for all three. Running the replicator equation on RPS payoffs instead produces a perpetual cycle -- Rock rises while Scissors is common, Paper rises to counter the rising Rock, Scissors rises to counter Paper, and the cycle repeats indefinitely, either as a closed orbit or (with added noise or spatial structure) as travelling spiral waves of dominant strategy across the grid. Real biological systems with cyclic dominance, such as the three male mating-strategy morphs of the side-blotched lizard, show exactly this kind of perpetual non-equilibrium cycling.

Frequently asked questions

Try it live

Everything above runs in your browser — open Evolutionary Game Theory and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.