🧬 biology

Emergence of polymorphism in stochastic evolutionary games

This paper resolves contradictions in previous simulation studies regarding the "trembling hand hypothesis" by theoretically analyzing the Hawk-Dove game with timescale separation, demonstrating that stochastic demographic noise in finite populations can drive the emergence and persistence of polymorphic states.

Original authors: Bill Nunn, Marcel Ortgiese, Tim Rogers

Published 2026-07-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Bill Nunn, Marcel Ortgiese, Tim Rogers

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine a bustling city square where people are constantly making choices: to cooperate or to compete, to share or to take. Scientists who study how these choices evolve over time use a field called evolutionary game theory. Think of it as a giant, invisible scoreboard where every move you make earns you points, and those points determine how many "copies" of your strategy get passed on to the next generation. For a long time, the most popular way to model this was with a "deterministic" approach. This is like watching a movie in slow motion where the outcome is perfectly predictable: if you know the rules and the starting numbers, you can calculate exactly where everyone will end up. In this smooth, ideal world, a population of animals playing a game is often treated as a single, blended average. It doesn't matter if every single animal is a "chameleon" that randomly switches between being aggressive and peaceful, or if the crowd is a mix of pure "fighters" and pure "peacekeepers" as long as the total ratio is the same. The math says they are identical.

However, real life isn't a smooth movie; it's a chaotic, noisy reality with finite numbers of people. This is where a famous idea called the "trembling hand" hypothesis comes in. It suggests that in a finite group, random accidents and noise should eventually wipe out the mix, leaving only one single type of animal dominating the square. But here's the twist: when scientists ran computer simulations of these noisy groups, the results didn't always match the "trembling hand" prediction. Sometimes, the mix of different types seemed to stick around forever, defying the idea that noise should force everyone to become the same. This created a puzzle: the smooth math said one thing, the noisy simulations said another, and nobody knew who was right.

This paper steps in to solve that puzzle by looking at a classic conflict scenario known as the "Hawk-Dove" game. Imagine a group of animals fighting over a single, juicy piece of food. A "Hawk" is an aggressive fighter who will chase away a "Dove" (a peaceful animal) to take the whole prize. If two Doves meet, they share the prize peacefully. But if two Hawks meet, they fight, and both get hurt, leaving them with less than if they had just shared. The math shows that in a perfect world, the population settles into a specific balance where some are aggressive and some are peaceful. The authors of this paper wanted to know: in a real, noisy world with a limited number of animals, does the population eventually become a single, uniform type (monomorphic), or can it get stuck in a long-lasting mix of different types (polymorphic)?

The researchers built a new mathematical model that treats the population not as a smooth average, but as a collection of individual animals, each with their own specific strategy. They used a technique called "timescale separation," which is like watching a fast-forwarded video of the population's quick adjustments, followed by a slow-motion look at how random noise slowly nudges the system over time. They found that the "trembling hand" idea isn't always the whole story. While the population does quickly settle onto a specific path (a "center manifold") where the mix of strategies looks stable, the random noise doesn't just push the system toward a single winner. Instead, under certain conditions, the noise actually traps the population in a metastable state.

In their simulations, they observed that the population could get stuck in a long-lasting "polymorphic" state, where you have a mix of pure fighters and pure peacekeepers coexisting for a very long time, rather than collapsing into a single type. They proved analytically that the probability of the population becoming a single, uniform type does not reach 100% even as the population gets huge; it stays bounded away from certainty. This means that the "all-Nash" state (where everyone plays the perfect mixed strategy) is favored, but it is not guaranteed. The noise can keep the population in a diverse, mixed state for a surprisingly long time.

The paper also explored what happens when new mutations (new strategies) appear. They found that if the population is split with some strategies on one side of the "balance point" and others on the other side, the system becomes incredibly stable against change. It takes an exponentially long time for one side to wipe out the other. This suggests that in the real world, we might see long-lasting diversity driven by the very randomness that was thought to destroy it. The authors conclude that while the classical math gives a good first guess, the messy, finite nature of real populations allows for a rich, persistent variety of strategies to survive, resolving the contradiction between the smooth theory and the noisy simulations.

Technical Summary: Emergence of Polymorphism in Stochastic Evolutionary Games

Problem Statement
Deterministic evolutionary game theory, specifically the classical replicator equation, fails to distinguish between a monomorphic population where every individual adopts a mixed Evolutionarily Stable Strategy (ESS) and a polymorphic population where individuals play pure strategies in a ratio that reproduces the mixed strategy on average. While the "trembling hand" hypothesis suggests that demographic noise in finite populations should select for monomorphism (driving the population to a single strategy), simulation studies have yielded contradictory results, showing that polymorphic states can persist even in large populations. This paper addresses the discrepancy between classical theoretical predictions and stochastic simulation outcomes in the context of the Hawk-Dove game.

Methodology
The authors employ a multi-stage analytical approach combining individual-based modeling, stochastic differential equations (SDEs), and timescale separation:

Extended Replicator System: The study shifts from the classical two-strategy replicator to an "extended" system tracking three distinct mixed strategies (L, M, R) within a population. This system is shown to possess a globally attracting center manifold of steady states.
Individual-Based Model (IBM) and SDE Derivation: A finite population model of size $K$ is formulated where individuals interact in pairs and replace one another based on fitness differences. Using Kurtz's methodology, this discrete jump process is approximated by a multidimensional SDE. The drift term of this SDE corresponds to the extended replicator dynamics, while the diffusion term captures demographic noise scaling as $1/\sqrt{K}$ .
Timescale Separation and Reduction: Recognizing that for large $K$ , the system dynamics remain close to the center manifold, the authors derive a "reduced SDE" governing the evolution along this manifold. They exploit a specific property of replicator dynamics—where flow lines correspond to level sets of a closed-form function—to explicitly calculate the projection derivatives required for the reduction. This yields a one-dimensional SDE describing the stochastic evolution of the strategy proportions.
Fixation Probability Analysis: The reduced SDE is analyzed using scale functions to calculate the probability of fixation (reaching an endpoint where one strategy is eliminated) versus the persistence of polymorphism.

Key Contributions and Results

Analytical Resolution of Discrepancy: The paper provides a rigorous analytical explanation for why simulations show bounded probabilities of reaching the monomorphic Nash state, contrary to the "trembling hand" prediction that this probability should tend to unity as $K \to \infty$ .
Characterization of Metastability: The analysis reveals that when strategies exist on both sides of the Nash equilibrium, the system enters a metastable state. The time to fixation for such a polymorphic state scales exponentially with population size ( $\exp(K)$ ), whereas fixation between strategies on the same side of the Nash scales linearly ( $K$ ).
Explicit Fixation Probabilities: For the specific case of the Hawk-Dove-Nash system (Pure Dove, Pure Hawk, and Mixed Nash), the authors derive a closed-form expression for the fixation probability (Equation 25). They demonstrate that the probability of the Nash strategy fixing is bounded away from one, even for arbitrarily large $K$ .
Noise-Driven Polymorphism: The study identifies conditions under which demographic noise drives the emergence of long-lasting polymorphic states. Specifically, when a population contains strategies on both sides of the Nash, the stochastic dynamics prevent the system from collapsing into a single strategy, maintaining a mixture of pure strategists.
Comparison with Adaptive Dynamics: The authors contrast their findings with Adaptive Dynamics (AD). They note that standard AD assumptions (mutual exclusion of traits) fail in this context because the time to fixation for metastable states is exponential, not instantaneous. However, under specific mutation timescales (super-linear but sub-exponential in $K$ ), the stochastic model predicts evolutionary paths distinct from classical AD branching.

Significance
The paper claims to fully resolve the apparent contradiction between classical deterministic predictions and stochastic simulation results in finite evolutionary games. By utilizing timescale separation to reduce the dimensionality of the stochastic system, the authors demonstrate that demographic noise does not necessarily eliminate polymorphism; rather, it can sustain long-lasting polymorphic states driven by the geometry of the flow field relative to the Nash equilibrium. This work extends the tradition of using finite population models to uncover phenomena (such as spontaneous speciation and selection reversal) that are invisible in infinite population limits. The authors conclude that their framework allows for the calculation of fixation probabilities and evolutionary trajectories without the need for repeated full-scale simulations, offering a tractable method to study stochastic evolutionary dynamics in games with arbitrary strategy sets.

Technical Summary: Emergence of Polymorphism in Stochastic Evolutionary Games

More like this