Rationalizable Strategies

Imagine you are choosing a move in a game, like rock-paper-scissors, a card game, or even deciding whether to price a product high or low in a business competition 🎮📈. In game theory, not every possible action is equally sensible. Some choices can be ruled out because a player would never choose them if they are thinking carefully and believe the other players are also thinking carefully.

In this lesson, students, you will learn how rationalizable strategies help us narrow down the set of strategies that could realistically be played. The big idea is common belief in rationality: each player is rational, each player knows the others are rational, each player knows that the others know this, and so on. This lets us eliminate strategies that cannot be justified as best responses to any reasonable beliefs about the opponents.

By the end of this lesson, you should be able to:

define rationalizable strategies,
explain how rationalizability is connected to best responses,
distinguish rationalizability from equilibrium.

What Does It Mean to Be Rationalizable?

A strategy is rationalizable if it can be justified as a best response to some beliefs about what the other players may do, where those beliefs are themselves consistent with the other players being rational. In simple terms, a rationalizable strategy is one that survives careful reasoning about what a player could reasonably expect others to do.

This does not mean the strategy is the best move against every possible choice by opponents. Instead, it means there is at least one believable story about the opponents’ behavior under which the strategy is optimal. That is why rationalizable strategies are broader than equilibrium strategies, but still more selective than listing every possible action.

Here is the core logic:

A player chooses a best response to beliefs about others’ strategies.
Those beliefs should only assign positive probability to strategies that are themselves rationalizable.
If a strategy cannot be a best response under any such beliefs, it is not rationalizable.

This idea is closely tied to iterated elimination of never-best responses. If a strategy is never a best response to any beliefs about the opponents’ actions, then a rational player would not use it. Removing such strategies repeatedly helps identify the rationalizable set.

For example, if a player has a strategy that is always worse than another strategy no matter what the opponent does, that strategy is strictly dominated and cannot be rationalizable. But rationalizability goes beyond strict dominance. A strategy may not be dominated and still fail to be rationalizable if it cannot be justified as a best response to rational beliefs.

Rationalizability and Best Responses

The idea of a best response is central here. A best response is the strategy that gives a player the highest payoff given what the player believes the others will do.

Suppose player $i$ has a belief over opponents’ strategies, represented by a probability distribution $\sigma_{-i}$. Then strategy $s_i$ is a best response if it gives at least as much expected payoff as any other strategy:

$$u_i(s_i, \sigma_{-i}) \geq u_i(s_i', \sigma_{-i}) \text{ for every alternative strategy } s_i'$$

This condition means the player is doing the best possible thing according to their beliefs. Rationalizability says: a strategy is acceptable if it is a best response to some beliefs that could arise when everyone is rational.

A useful way to think about this is like choosing a route to school 🚲. If you believe traffic will be heavy on one street and light on another, you pick the route that seems best under those beliefs. If your beliefs change, your best route may change too. Rationalizable strategies are the routes that could be optimal under some reasonable beliefs, not just one fixed situation.

Example: A Simple Two-Strategy Situation

Suppose player $A$ can choose $U$ or $D$, and player $B$ can choose $L$ or $R$. If $A$’s payoffs are higher from $U$ when $B$ chooses $L$, and higher from $D$ when $B$ chooses $R$, then both $U$ and $D$ might be best responses under different beliefs about $B$’s behavior.

If $A$ believes $B$ is likely to choose $L$, then $U$ may be rational. If $A$ believes $B$ is likely to choose $R$, then $D$ may be rational. In this case, both strategies could be rationalizable.

But if one strategy is never a best response to any belief about $B$, then it is not rationalizable. That strategy can be removed from consideration because a rational player would never choose it.

Why Beliefs Matter

Rationalizability is not about predicting one exact outcome. It is about identifying strategies that are consistent with rational choice under some belief system. The beliefs do not have to be correct; they only need to be possible and consistent with rationality.

This is important because in many games, players do not know exactly what others will do. They form expectations based on experience, observation, or reasoning. Rationalizability helps us understand what choices remain possible when players reason carefully but do not have perfect certainty.

How Rationalizability Is Found

A common way to find rationalizable strategies is through iterated elimination. The process works like this:

Remove strategies that are never best responses to any beliefs.
After removing those, look at the reduced game.
Remove any new strategies that have become never-best responses in the reduced game.
Keep going until no more strategies can be eliminated.

This process is powerful because once some strategies are ruled out, other strategies may lose the support they had before. A strategy that seemed possible at first may stop being rationalizable after weaker strategies are removed.

Small Illustration

Imagine three players in a pricing game. Each can choose a high price or a low price. If a high price only works when others also choose high prices, but one player realizes that a certain low price is never a rational choice for the others, then the belief set changes. That may make the high price less believable as well, because the support for it disappears.

This is one reason rationalizability is stronger than simply checking whether a strategy has a good payoff in one scenario. The process takes into account how each player reasons about the reasoning of others.

Rationalizability Versus Equilibrium

students, this is one of the most important distinctions in the lesson.

A strategy is rationalizable if it can be a best response to some beliefs that are consistent with common belief in rationality. An equilibrium requires something stronger: each player’s chosen strategy must be a best response to the actual strategies chosen by the others.

In a Nash equilibrium, players’ strategies fit together like a stable puzzle piece. No player wants to deviate, given what the others are doing.

Rationalizability is less demanding because it only asks whether a strategy could be optimal under some believable expectations, not whether the strategies all fit together simultaneously as an equilibrium.

Key Differences

Rationalizability asks: “Could this strategy be a best response to some rational beliefs?”
Equilibrium asks: “Is this strategy a best response to the other players’ actual strategies?”

So every equilibrium strategy is rationalizable, but not every rationalizable strategy must be part of an equilibrium outcome. Rationalizability gives a broader set of possible actions.

Example of the Difference

Consider a coordination game where two friends choose whether to meet at the mall or the park. Each prefers to match the other person’s choice. Both choices may be rationalizable because each can be a best response to a belief that the other will choose it. But only the matched outcomes, where both choose the same location, are equilibrium outcomes.

That means a strategy can make sense under some belief and still not be part of a stable mutual choice in equilibrium.

Why Rationalizability Matters in Real Life

Rationalizability is useful because real decisions often happen under uncertainty. People in business, politics, sports, and everyday life rarely know exactly what others will do. They reason about what others are likely to do, and they also think about how others are reasoning.

For example:

A company may choose an advertising level based on what it believes competitors will do 📣.
A player in a game may choose a defensive move because it is a best response to an expected attack.
A negotiator may make an offer that is sensible if they believe the other side is also acting strategically.

Rationalizability helps filter out impossible or irrational actions without requiring a full prediction of the final outcome. It is a way of saying, “These strategies remain plausible once we assume everyone is thinking logically.”

Conclusion

Rationalizable strategies are the strategies that survive reasoning based on common belief in rationality. A strategy is rationalizable if it can be a best response to some beliefs about the other players’ actions, where those beliefs are themselves consistent with rational behavior.

The main takeaway is that rationalizability narrows the set of possible strategies by eliminating choices that no rational player would ever use, even under reasonable uncertainty. It is closely tied to best responses and iterated elimination, but it is not the same as equilibrium. Equilibrium requires mutual consistency of chosen strategies, while rationalizability only requires that a strategy could make sense under rational beliefs.

Study Notes

Rationalizable strategies are strategies that can be justified as best responses to some rational beliefs about opponents.
Common belief in rationality means each player is rational, knows others are rational, and knows that they know this, continuing indefinitely.
A best response satisfies $u_i(s_i, \sigma_{-i}) \geq u_i(s_i', \sigma_{-i})$ for every alternative strategy $s_i'$.
Rationalizability is often found by repeatedly removing strategies that are never best responses.
Strictly dominated strategies are not rationalizable.
Rationalizability is broader than equilibrium.
Nash equilibrium requires each player’s strategy to be a best response to the others’ actual strategies.
Every equilibrium strategy is rationalizable, but not every rationalizable strategy is part of an equilibrium outcome.
Rationalizability is useful for understanding which strategies remain plausible when players reason strategically under uncertainty.