Support Conditions in Mixed Strategies 🎲

students, when players randomize in game theory, they are not being careless. They are making a strategic choice. In this lesson, you will learn how to tell which pure actions can appear with positive probability in a mixed strategy equilibrium, and which ones cannot. This idea is called support conditions. It helps you test whether a proposed mixed equilibrium is possible before you spend time solving the whole game.

What You Will Learn

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

Define the support of a mixed strategy.
Check support conditions in equilibrium.
Eliminate impossible mixed-strategy candidates.

Support conditions are one of the most useful tools in mixed-strategy game theory because they connect a player’s probabilities to their payoffs. The big idea is simple: if a pure action gets positive probability in equilibrium, then it must give the player the same payoff as every other action in that player’s support. If one supported action gave a lower payoff, the player would stop using it.

Support: Which Actions Are Actually Played?

In a mixed strategy, a player assigns probabilities to pure actions. For example, if a player has two actions, $A$ and $B$, they might choose $A$ with probability $0.7$ and $B$ with probability $0.3$.

The support of a mixed strategy is the set of pure actions that are given positive probability. So in that example, the support is $\{A, B\}$ because both probabilities are greater than $0$.

If a player chooses $A$ with probability $1$ and $B$ with probability $0$, then the support is just $\{A\}$.

Why support matters

students, the support tells you which actions are being “kept alive” by the equilibrium. Any action outside the support is being ignored. Any action inside the support must be worth using, given what the other players are doing.

Here is the key rule:

Every action in the support must give the same expected payoff.
Any action outside the support must give a payoff that is no higher than the supported actions’ payoff.

This is because a rational player never mixes between actions that are not equally attractive. If one action were better, the player would put all probability on it instead of randomizing.

The Support Condition in Equilibrium

Suppose Player 1 is mixing among several actions. Let the expected payoff from action $a$ be $u_1(a)$. If action $a$ is in Player 1’s support, then it must satisfy

$$u_1(a) = U_1^*$$

where $U_1^*$ is Player 1’s equilibrium expected payoff.

If an action is not in the support, then it must satisfy

$$u_1(a) \le U_1^*$$

because otherwise Player 1 would want to switch to that action.

This is the foundation of support conditions.

The logic behind the rule

Imagine a student deciding whether to study by reading notes or doing practice problems. If both choices lead to the same expected result, the student might split time between them. But if practice problems clearly work better, the student will not keep spending time on notes. In the same way, in equilibrium, only equally good actions can receive positive probability.

A Simple Example: Two-Action Mix

Consider a player who can choose $A$ or $B$. Suppose the opponent mixes in a way that makes the expected payoff from choosing $A$ equal to the expected payoff from choosing $B$.

$$u(A) = u(B),$$

then the player may mix between $A$ and $B$. Both actions can be in the support.

But if

$$u(A) > u(B),$$

then $B$ cannot be in the support in equilibrium, because the player would prefer $A$ every time.

Example with numbers

Suppose Player 1’s expected payoffs are:

$u(A) = 5$
$u(B) = 5$
$u(C) = 3$

Then $A$ and $B$ can be in the support, but $C$ cannot. Why? Because $C$ gives a lower payoff than the supported actions.

If the player were mixing among $A$, $B$, and $C$, the probabilities on $C$ would have to be $0$ in equilibrium.

How to Check Support Conditions

Support conditions are usually used in two steps.

Step 1: Guess the support

You start by guessing which pure actions might be used with positive probability.

For example, if a player has three actions, you might guess that only $A$ and $B$ are in the support.

Step 2: Make supported actions indifferent

Then you solve for the opponent’s mixed strategy so that the player gets equal payoff from every action in the support.

If the guessed support is $\{A, B\}$, you set

$$u(A) = u(B).$$

This gives an equation for the opponent’s probability.

Step 3: Check excluded actions

Next, you verify that every action outside the support gives no more than the equilibrium payoff.

If $C$ is excluded, you must check

$$u(C) \le u(A) = u(B).$$

If this fails, your support guess is impossible.

This is why support conditions are powerful: they help eliminate wrong candidates quickly.

Worked Example: Eliminating an Impossible Support

Suppose Player 1 has three actions $A$, $B$, and $C$, and the expected payoffs against the opponent’s mixed strategy are:

$$u(A) = 2p + 1$$

$$u(B) = p + 3$$

$$u(C) = 4 - p$$

where $p$ is the probability that the opponent chooses one of their actions.

Can $A$ and $B$ both be in the support?

For both to be in the support, they must give equal payoff:

$$2p + 1 = p + 3$$

Solving gives

$$p = 2.$$

But $p$ is a probability, so it must satisfy

$$0 \le p \le 1.$$

Since $p = 2$ is impossible, the support $\{A, B\}$ cannot happen.

Can $A$ and $C$ both be in the support?

Set their payoffs equal:

$$2p + 1 = 4 - p$$

Then

$$3p = 3$$

$$p = 1.$$

Now check the excluded action $B$:

$$u(B) = p + 3 = 4$$

and

$$u(A) = 2(1) + 1 = 3.$$

Since

$$u(B) > u(A),$$

action $B$ would be better than the supported actions. So $A$ and $C$ also cannot be the support.

This example shows the full logic: equalize the support, then verify the outside actions.

Support and Best Responses

A mixed strategy equilibrium must satisfy the best response condition. That means every action in the support must be a best response to the opponent’s strategy.

If an action is not a best response, it cannot receive positive probability.

So support conditions are really a way of saying:

the player is indifferent among all actions used with positive probability,
and all unused actions are weakly worse.

This is a direct consequence of rational behavior.

Important warning

students, do not assume that any action with a decent payoff can be in the support. It must match the payoff of the other supported actions exactly. Even a tiny difference breaks the equilibrium because a player would shift probability toward the better action.

How Support Conditions Help in Two-Player Games

In two-player games, each player’s mix affects the other player’s payoffs. Support conditions let you solve for these mixes by turning the problem into equations.

For example, if Player 2 wants Player 1 to mix between $A$ and $B$, Player 2 must choose probabilities that make

$$u_1(A) = u_1(B).$$

At the same time, Player 1’s mix must make Player 2 indifferent among the actions in Player 2’s support.

This creates a system of equations. Once you solve it, you must check that all probabilities lie between $0$ and $1$ and that excluded actions are not better than supported ones.

Common Mistakes to Avoid ⚠️

Mistake 1: Forgetting the outside action check

It is not enough to make supported actions equal. You must also confirm that excluded actions are no better.

Mistake 2: Allowing negative probabilities

If solving the indifference equations gives something like $p = -\frac{1}{2}$ or $p = 2$, the candidate support is impossible.

Mistake 3: Mixing with unequal payoffs

If one action in the support gives more payoff than another, the player would not randomize between them.

Mistake 4: Confusing support with all available actions

The support is only the set of actions used with positive probability, not the entire action set.

Real-World Intuition 🌍

Think about choosing a route to school. If Route 1 and Route 2 take about the same time, you might switch between them. Those routes are like actions in the support. But if Route 3 is always slower, you will stop using it. That route is outside the support.

In game theory, players do the same thing with strategies: they keep only the choices that are equally attractive given everyone else’s behavior.

Conclusion

Support conditions are a central tool for analyzing mixed-strategy equilibria. students, the main idea is simple but powerful: every action played with positive probability must give the same expected payoff, and every action not played must give no more than that payoff. By using this rule, you can test guessed supports, eliminate impossible equilibria, and solve mixed-strategy problems more efficiently.

Whenever you see a mixed-strategy problem, start by asking: Which actions could actually be in the support? Then check whether those actions are indifferent and whether the excluded actions are weakly worse. That process is the key to finding valid mixed equilibria.

Study Notes

The support of a mixed strategy is the set of pure actions assigned positive probability.
In equilibrium, all actions in the support must give the same expected payoff.
Any action outside the support must give a payoff that is no greater than the supported actions’ payoff.
To test a candidate support, first equalize the payoffs of supported actions.
Then check that excluded actions are not better.
If solving the indifference equations gives probabilities outside $[0,1]$, the candidate support is impossible.
Support conditions help eliminate impossible mixed-strategy candidates before completing the full equilibrium calculation.
Mixed-strategy equilibrium combines indifference and best response behavior.
A player only randomizes among actions that are equally attractive.
Support conditions are especially useful in two-player games where each player’s mix affects the other player’s payoffs.