Predictive power is how well a game theory model forecasts which outcome or equilibrium players will actually choose. In games with multiple equilibria, it measures whether the theory can pick the most likely result.
Predictive power in Game Theory is the ability of a model, rule, or equilibrium concept to tell you which outcome players are most likely to reach. If a game has only one clear equilibrium, prediction is easy. If it has several equilibria, predictive power is what separates a neat mathematical answer from a useful forecast.
This term shows up most often in multiple equilibrium problems, where rational players could all justify more than one outcome. Nash equilibrium tells you which outcomes are stable, but it does not always tell you which stable outcome people will actually choose. Predictive power asks the next question: among the options that are possible, which one is most likely to happen in real play?
That is why the idea is tied to equilibrium selection. A concept has stronger predictive power when it narrows the field better, especially if it points to the equilibrium that people can coordinate on more easily, or the one that tends to survive repeated play. In that sense, predictive power is not just about mathematical consistency. It is about practical usefulness.
Game theory often uses equilibrium refinements to improve predictive power. For example, in a coordination game like Battle of the Sexes or Stag Hunt, more than one equilibrium may exist, but players still need a reason to settle on one. Ideas like Pareto dominance, risk dominance, focal points, and criteria such as Harsanyi-Selten are all attempts to make the prediction sharper.
A good way to think about it is this: a theory with weak predictive power may tell you what cannot happen, but still leave the actual outcome unclear. A theory with stronger predictive power gives you a narrower, more believable forecast because it takes coordination, stability, and likely behavior into account. That is especially useful in games where the whole problem is not finding an equilibrium, but deciding which equilibrium people will choose.
Predictive power matters because game theory is not just about listing possible equilibria, it is about explaining real strategic behavior. In many classroom examples, several outcomes satisfy the rules of the game, but only one will actually show up if players have to act at the same time, communicate imperfectly, or rely on shared expectations.
This term helps you judge whether a model is doing real work or just producing a menu of options. A model with weak predictive power can describe the strategy structure of a game, but still leave you unsure about the likely result. That is a problem in coordination games, where people need a shared expectation before they can settle on one outcome.
Predictive power also connects to stability. Some equilibria are easier to maintain because small changes in behavior do not push the game away from them. Those equilibria usually have a better chance of being selected in practice, especially when players have some history, convention, or repeated interaction to guide them.
In the course, this term gives you a way to compare equilibrium concepts, not just memorize them. If you can explain why one refinement predicts behavior better than another in a Battle of the Sexes or Stag Hunt setup, you are showing that you understand the logic of equilibrium selection, not just the definition of equilibrium.
Keep studying Game Theory Unit 3
Visual cheatsheet
view galleryEquilibrium
Predictive power depends on whether an equilibrium concept can narrow down the likely outcome, not just show that outcomes are stable. If a game has many equilibria, the basic equilibrium idea may be mathematically correct but still leave the real result unclear. Predictive power is stronger when the equilibrium concept gives you a more specific forecast.
Nash Equilibrium
Nash equilibrium gives the set of stable outcomes, but it does not always tell you which one players will choose when there is more than one. Predictive power is the extra question of selection. In many problems, you start by finding the Nash equilibria and then ask which one is most likely to be chosen.
Harsanyi-Selten Criterion
The Harsanyi-Selten Criterion is one attempt to improve predictive power by selecting among multiple equilibria. It gives a rule for ranking equilibria when a game has more than one stable outcome. That makes it useful in exactly the situations where predictive power is otherwise weak.
Stag Hunt Game
The Stag Hunt Game is a classic example where several equilibria can exist, but players still need a reason to coordinate on one. Predictive power becomes a real issue because the game has more than one stable answer. This is the kind of setting where equilibrium selection matters most.
A quiz question or problem set item will usually give you a game with more than one equilibrium and ask which outcome is most likely or which equilibrium concept has the best forecasting value. Your job is to identify why the basic equilibrium set is not enough, then explain what makes one outcome more plausible than the others. That might mean pointing to coordination, stability, historical play, or a refinement like Pareto dominance or risk dominance.
If the prompt uses a game such as Battle of the Sexes or Stag Hunt, mention that the issue is not finding an equilibrium from scratch. The issue is equilibrium selection. A strong answer shows that you can move from “these outcomes are possible” to “this one is the better prediction” and justify that with game-theory language.
Equilibrium is the set of stable outcomes, while predictive power is about how well a theory forecasts which stable outcome will actually happen. You can have an equilibrium description without much predictive power if the game has multiple equilibria and the model cannot select between them.
Predictive power is a model’s ability to forecast which outcome players will actually reach in a game.
It matters most when a game has multiple equilibria and the theory needs to choose among them.
A concept can be mathematically correct and still have weak predictive power if it does not narrow the outcome enough.
In Game Theory, predictive power is tied to equilibrium selection, stability, and coordination.
Games like Battle of the Sexes and Stag Hunt are good places to see why predictive power matters.
Predictive power in Game Theory is how well a model forecasts the outcome players will choose, especially when more than one equilibrium is possible. It goes beyond listing stable outcomes and asks which one is most likely in practice. The term is especially useful in coordination games where players need a shared expectation.
Nash equilibrium tells you which outcomes are stable because no player wants to change strategy alone. Predictive power asks whether that equilibrium concept can actually predict which stable outcome people will pick. A game can have several Nash equilibria, so the prediction may still be unclear without a selection rule.
In a Battle of the Sexes game, both players may have more than one equilibrium option, but they still need to coordinate on one. The equilibrium with better predictive power is the one the players are more likely to pick based on shared expectations, payoff structure, or a refinement such as Pareto dominance. The point is not just that an outcome is stable, but that it is believable as the actual choice.
Multiple equilibria make prediction harder because the game no longer points to one clear result. If several outcomes are stable, you need extra information such as history, focal points, or selection criteria to guess which one will happen. That is why predictive power becomes a central issue in equilibrium selection.