Skip to content
Tutorials Center
My learning

Topics

Lesson 25 of 37 ยท Repeated and Evolutionary Games

Evolutionary Stable Strategies

Learn evolutionary game theory and ESS: Maynard Smith's conditions, the Hawk-Dove game, mixed ESS, and a Python checker for invasion by mutants.

Classical game theory assumes that players think hard and choose strategies rationally. Evolutionary game theory drops that assumption. Individuals in a large population simply follow inherited or learned behaviors and are paired at random. Behaviors that earn higher payoffs ("fitness") spread, and the others fade. The question changes from "what will clever players choose?" to "which behavior can resist being replaced by a new one?"

John Maynard Smith answered with the evolutionarily stable strategy (ESS): a strategy that, once almost everybody uses it, cannot be successfully invaded by a rare mutant.

The setting

A large population plays a symmetric two-player game. Let u(x, y) denote the payoff to someone using x against someone using y. Everyone starts by playing strategy x. A small fraction epsilon of mutants appear, playing strategy y. A random individual meets a resident with probability 1 - epsilon and a mutant with probability epsilon. The expected payoffs are

  • resident: (1 - epsilon) * u(x, x) + epsilon * u(x, y)
  • mutant: (1 - epsilon) * u(y, x) + epsilon * u(y, y)

Strategy x is an ESS if, for every mutant y different from x, the resident earns strictly more than the mutant when epsilon is small enough. Taking epsilon to zero gives two conditions:

  1. u(x, x) > u(y, x) , or
  2. u(x, x) = u(y, x) and u(x, y) > u(y, y).

In words: x must be a best reply to itself (so (x, x) is a Nash equilibrium), and if some mutant ties against x, then x must beat that mutant when they meet each other.

ESS in plain words

Residents Almost everyone plays x
Rare mutants A few play y instead
Fitness Payoff decides who spreads
ESS Residents always do better

Every ESS is a Nash equilibrium; not every Nash equilibrium is an ESS.

1 / 2

Hawk-Dove

Two animals contest a resource worth V. A Hawk fights until it wins or is injured. A Dove displays and retreats if the other fights. Two Hawks split the outcome: each wins half the time, gaining V, and loses half the time, suffering injury cost C, so the expected payoff is (V - C) / 2. A Hawk meeting a Dove takes the resource: V for the Hawk, 0 for the Dove. Two Doves share: V / 2 each.

Hawk Dove
Hawk (V - C) / 2 V
Dove 0 V / 2

(This table shows the payoff to the row strategy.)

If C > V, fighting is costly, and neither pure strategy is stable: in a population of Hawks, a Dove mutant earns 0 while Hawks earn (V - C) / 2 < 0, so Doves are better off and invade. In a population of Doves, a Hawk mutant earns V against V / 2, so Hawks invade. The stable point is a mixed strategy with a fraction V / C of Hawks.

An ESS checker

The function below tests the two conditions for a mixed strategy against a grid of mutants (pure ones are enough for 2 x 2 games, but a grid gives extra assurance). Payoff is computed with the strategy as the probability of playing the first action.

Python Checking ESS conditions for Hawk-Dove

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Only the mixed strategy x = V / C = 2/3 passes. At that mix, Hawk and Dove earn the same against the population, which is why all mutants tie in the first condition and the second condition does the work. Because the second condition holds (against a mutant, the resident does better than the mutant does against itself), the mix is an ESS.

Two readings of a mixed ESS

A mixed ESS can mean that each individual randomizes with probability V / C, or that the population is polymorphic: 2/3 of the animals are always Hawks and 1/3 are always Doves. Both readings produce the same average behavior, and with random pairing they give the same fitness.

Calculating the mixed equilibrium

For a 2 x 2 symmetric game with payoffs a, b, c, d (Hawk against Hawk, Hawk against Dove, Dove against Hawk, Dove against Dove), the interior equilibrium makes the two strategies equally fit: p*a + (1 - p)b = pc + (1 - p)*d, so p = (d - b) / (a - b - c + d).

Python Mixed equilibrium and fitness check

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Evolution does not maximize the average payoff: all Doves would earn 2.0 each, while the ESS population earns less (only 2/3). Stability is about resisting invaders, not about efficiency. The prisoner's dilemma illustrates this too: Defect is the ESS and the population ends up with a poor average payoff.

Games with no ESS: rock-paper-scissors

Rock-paper-scissors has a unique Nash equilibrium (play each action with probability 1/3), but it is not an ESS. Against that mix every strategy earns zero, so every mutant ties in the first condition. The second condition then asks whether the resident beats the mutant head to head, and it does not: the resident also earns zero against a pure mutant, and the mutant earns zero against itself. Without a strict win, mutants can drift in, and populations cycle instead of settling. The next lesson, replicator dynamics, makes the cycles visible.

Python No ESS in rock-paper-scissors

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

All four numbers are zero for every pure mutant, so the second condition, u(x, y) > u(y, y), fails. The mix is a Nash equilibrium but not an ESS, because mutants do exactly as well as residents and can drift in.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. An evolutionarily stable strategy is one that...

    Maynard Smith's definition: a population playing it resists small invasions.
  2. 2. What is the relationship between ESS and Nash equilibrium?

    The first ESS condition requires x to be a best reply to itself. The converse fails, e.g. in rock-paper-scissors.
  3. 3. In Hawk-Dove with V = 4 and C = 6, what fraction of Hawks is the mixed ESS?

    The ESS fraction is V / C = 4 / 6 = 2/3.
  4. 4. Why is the Hawk-Dove ESS not payoff-maximizing for the population?

    An all-Dove group has the best average payoff but a lone Hawk earns more than the Doves, so it is not stable.

Summary

  • Evolutionary game theory studies which behaviors survive in large populations, without assuming rational choice.
  • An ESS x satisfies: u(x, x) > u(y, x), or a tie with u(x, y) > u(y, y), for every mutant y.
  • Every ESS is a Nash equilibrium, but some Nash equilibria (like rock-paper-scissors) are not ESS.
  • In Hawk-Dove with C > V, the ESS is mixed, with V / C Hawks.
  • Stability does not mean efficiency: ESS can have a lower average payoff than a fragile cooperative outcome.

Next: Replicator Dynamics