Optimal Pure Strategy — Definition, Formula & Example
An optimal pure strategy is one fixed move a player can use every time in a two-person zero-sum game and still get the best result possible against a smart opponent. Both players have optimal pure strategies exactly when the game has a saddle point (a strictly determined game).
Let be the payoff matrix of a two-person zero-sum game, with entries giving the row player's payoff. Row and column are optimal pure strategies when is both the minimum of row and the maximum of column . Equivalently, . This common number is the value of the game, and is a saddle point.
Key Formula
Where:
- = Payoff to the row player when row plays and column plays
- = Maximin: the largest of the row minimums (row player's guaranteed floor)
- = Minimax: the smallest of the column maximums (column player's guaranteed ceiling)
- = Value of the game, the payoff at the saddle point
Worked Example
Problem: In a zero-sum game, the column player pays the row player the amount (in dollars) shown. Rows are ; columns are . Find each player's optimal pure strategy, if one exists.
Step 1: Find the minimum of each row. The row player picks the row with the largest minimum, so this is the maximin.
Step 2: Find the maximum of each column. The column player picks the column with the smallest maximum, so this is the minimax.
Step 3: The maximin equals the minimax, so a saddle point exists at . The entry is the smallest number in its row and the largest in its column.
Answer: The row player's optimal pure strategy is and the column player's is . The value of the game is $3, paid to the row player each round.
Why It Matters
In finite math and operations research courses, checking for a saddle point is a standard early step in solving a two-person zero-sum game. If one exists, you are done with simple arithmetic. If not, you must move on to mixed strategies, which require probabilities or linear programming.
Common Mistakes
Mistake: Assuming every zero-sum game has an optimal pure strategy.
Correction: Many games have no saddle point. In Matching Pennies, with matrix , the maximin is but the minimax is . Since they differ, there is no saddle point, so the players cannot both use optimal pure strategies; at least one of them must mix moves at random to play optimally.
