Mixed strategies

What Are Mixed Strategies?

Mixed strategies are a central concept in game theory, representing a probabilistic approach to decision-making within strategic interactions. Instead of selecting a single, definite action, a player employing a mixed strategy chooses among their available pure strategy options with a predefined probability distribution. This randomization introduces unpredictability, making it more challenging for opponents to anticipate and exploit a player's moves. The use of mixed strategies is particularly relevant in situations where no single pure strategy offers a clear advantage, allowing market participants to achieve an optimal outcome based on the probabilities assigned to each action.

History and Origin

The foundational ideas of mixed strategies emerged in the early 20th century, with significant contributions from mathematicians such as Émile Borel, who discussed the concept in 1921. However, it was John von Neumann who provided the formal proof for the existence of mixed-strategy equilibria in two-person zero-sum games. The concept was further formalized and significantly expanded upon by John Nash in his seminal 1950 Ph.D. dissertation, "Non-Cooperative Games." Nash's work demonstrated that every finite game, whether cooperative or non-cooperative, possesses at least one Nash equilibrium, which could involve players using mixed strategies. His concise 26-page thesis revolutionized the field, paving the way for the widespread application of game theory in economics and other disciplines.
⁵

Key Takeaways

• Mixed strategies involve randomizing choices among pure strategies with specific probabilities.
• They are employed when no single pure strategy guarantees the best outcome against a rational opponent.
• The concept helps players achieve unpredictability, which can be crucial in competitive scenarios.
• Mixed strategy equilibria ensure that no player can unilaterally improve their expected value by changing their strategy.
• They are a cornerstone of modern game theory, providing solutions for games without pure strategy equilibria.

Formula and Calculation

The core of a mixed strategy calculation involves determining the probabilities that make an opponent indifferent to their own pure strategies, assuming they are rational. For a player choosing a mixed strategy, the expected payoff from each of their pure strategies must be equal, given the opponent's mixed strategy.

Consider a two-player, two-action game where Player 1 has strategies A and B, and Player 2 has strategies C and D. Let Player 1's mixed strategy be (\sigma_1 = (p, 1-p)) (playing A with probability (p), B with probability (1-p)), and Player 2's mixed strategy be (\sigma_2 = (q, 1-q)) (playing C with probability (q), D with probability (1-q)). The payoffs are represented in a payoff matrix.

To find Player 1's optimal mixed strategy (p), Player 2 must be indifferent between playing C and D. This means Player 2's expected payoff from C must equal their expected payoff from D.

Similarly, to find Player 2's optimal mixed strategy (q), Player 1 must be indifferent between playing A and B.

These equations are solved simultaneously to find the probabilities (p) and (q) that constitute a mixed strategy Nash equilibrium. The calculation ensures that each player's chosen probabilities maximize their own expected payoff, given the opponent's strategy. This framework is essential in understanding equilibrium concepts in non-zero-sum game scenarios.

Interpreting Mixed Strategies

Interpreting mixed strategies involves understanding that players are not simply guessing; rather, they are using probabilities to achieve a state of indifference for their opponents, preventing predictable exploitation. When a player adopts a mixed strategy, they are ensuring that their opponent gains no advantage by consistently choosing one specific action. This often leads to an equilibrium where each player's randomized action is the best response to the other's randomized action. In practical terms, this can mean diversifying actions over time or across various instances of a similar strategic interaction. For example, a company might use a mixed strategy in pricing by sometimes offering discounts and other times maintaining full price, based on probabilities calculated to maximize long-term profit against a competitor's responses.

Hypothetical Example

Consider a simplified market entry game between two firms, Alpha Corp and Beta Inc. Alpha Corp is considering entering a new market. Beta Inc, an incumbent, can either accommodate Alpha's entry or launch an aggressive advertising campaign.

Payoff Matrix (Alpha's Payoff, Beta's Payoff):

In this scenario, if Alpha enters and Beta accommodates, both get 5. If Alpha enters and Beta launches aggressive ads, both get 1. If Alpha doesn't enter, Beta gets 8 and Alpha gets 0.

There is no pure strategy Nash equilibrium here. Alpha would prefer to enter if Beta accommodates, but would prefer not to enter if Beta is aggressive. Beta would prefer to accommodate if Alpha enters (to avoid conflict) but has no direct preference if Alpha doesn't enter. This is a simplification, but it illustrates the need for mixed strategies.

Let (p) be the probability Alpha enters and (1-p) be the probability Alpha doesn't enter. Let (q) be the probability Beta accommodates and (1-q) be the probability Beta uses aggressive ads.

To find Alpha's mixed strategy (the value of (p)), Beta must be indifferent between accommodating and using aggressive ads:
Expected Payoff for Beta (Accommodate) = (p \times 5 + (1-p) \times 8 = 5p + 8 - 8p = 8 - 3p)
Expected Payoff for Beta (Aggressive Ads) = (p \times 1 + (1-p) \times 8 = p + 8 - 8p = 8 - 7p)

Setting them equal: (8 - 3p = 8 - 7p \Rightarrow -3p = -7p \Rightarrow 4p = 0 \Rightarrow p = 0). This indicates Alpha would never enter, which is not quite right as it's a simplification and the "Don't Enter" provides Beta a clear dominant strategy.

Let's adjust the example to one that typically does have a mixed strategy equilibrium, like "Matching Pennies" or a similar conflict game.

Revised Hypothetical Example: Market Pricing Battle

Two competing companies, Firm A and Firm B, are deciding on their pricing strategy for a new product. Each can choose a "High Price" or a "Low Price."

Payoff Matrix (Firm A's Profit, Firm B's Profit):

If both set a High Price, they both earn 20. If A sets High and B sets Low, A gets 5 (loss of market share) and B gets 30. If A sets Low and B sets High, A gets 30 and B gets 5. If both set Low, they both earn 10 (reduced margins). There is no pure strategy Nash equilibrium. If Firm A chooses High, Firm B wants Low. If Firm A chooses Low, Firm B wants High. This cyclical nature implies mixed strategies.

Let Firm A play High with probability (p) and Low with probability (1-p).
Let Firm B play High with probability (q) and Low with probability (1-q).

For Firm B to be indifferent between High and Low Price:
Expected Payoff for B (High Price) = Expected Payoff for B (Low Price)
(p \times 20 + (1-p) \times 5 = p \times 30 + (1-p) \times 10)
(20p + 5 - 5p = 30p + 10 - 10p)
(15p + 5 = 20p + 10)
(5 - 10 = 20p - 15p)
(-5 = 5p)
(p = -1). This still indicates an issue with the matrix example. The payoffs need to be structured so that a positive probability exists.

Let's simplify to a classic Rock-Paper-Scissors type game logic for a simple mixed strategy example, if not tied to direct financial numbers. Or adjust the numbers to yield a valid probability.

Let's use a simpler payoff matrix that is known to have a mixed strategy equilibrium:

Game: "Penalty Kick" (Soccer Analogy)

(Kicker's chance of scoring, Goalkeeper's chance of saving)

If the kicker shoots Left and the keeper dives Left, the kicker scores 0 (saved). If kicker shoots Left and keeper dives Right, kicker scores 1.

Let Kicker play Left with probability (p) and Right with probability (1-p).
Let Goalkeeper play Left with probability (q) and Right with probability (1-q).

For Goalkeeper to be indifferent:
Expected Payoff for Goalkeeper (Left) = Expected Payoff for Goalkeeper (Right)
(p \times 1 + (1-p) \times 0 = p \times 0 + (1-p) \times 1)
(p = 1 - p)
(2p = 1 \Rightarrow p = 0.5)

For Kicker to be indifferent:
Expected Payoff for Kicker (Left) = Expected Payoff for Kicker (Right)
(q \times 0 + (1-q) \times 1 = q \times 1 + (1-q) \times 0)
(1 - q = q)
(1 = 2q \Rightarrow q = 0.5)

So, the mixed strategy Nash equilibrium is for both the kicker and the goalkeeper to choose Left or Right with a 50% probability. This means they randomize their actions, making it impossible for the opponent to predict their move and gain an advantage. This concept can be applied to competitive market pricing, negotiation strategies, or even asset allocation when facing uncertain market conditions.

Practical Applications

Mixed strategies find wide application across various domains where strategic interactions occur. In finance, they can be seen in the trading strategies of sophisticated investors or institutional traders who might randomize their buy and sell orders to avoid tipping off the market, thereby influencing prices less significantly. This aligns with principles found in quantitative finance. In corporate finance, companies might use mixed strategies when deciding on research and development spending, marketing campaigns, or even in mergers and acquisitions, factoring in the probabilistic responses of rivals.

Beyond finance, mixed strategies are crucial in fields like antitrust law, where regulators analyze how firms might collude or engage in anti-competitive behavior through complex strategic decision-making. Game theory, including the concept of mixed strategies, has been applied in various high-profile antitrust cases, such as the United States v. Microsoft case, to understand competitive implications and firm incentives. ⁴They are also applied in cybersecurity, military strategy, and even biology, where species evolve mixed strategies for survival.

Limitations and Criticisms

Despite their theoretical elegance, mixed strategies and game theory, in general, face several limitations and criticisms when applied to real-world financial contexts. One primary critique is the assumption of perfect rationality among players. In reality, human behavioral finance demonstrates that individuals often deviate from perfectly rational choices due to emotions, cognitive biases, or incomplete information. ³Game theory models, including those involving mixed strategies, can be too abstract and may not fully capture the complexities of actual market situations, where factors like external shocks or evolving market dynamics play a significant role.
²
Another challenge lies in accurately estimating the payoff matrix and the probabilities involved in real-world scenarios. In markets, the "game" is often played with imperfect information, and the precise preferences and strategies of all players are rarely known. Critics also point out that the predictive power of game theory, particularly concerning mixed strategies, can be limited because people do not always possess the assumed information or ability to perform complex calculations required for optimal mixed strategies. ¹Furthermore, the "randomization" aspect of mixed strategies can be difficult for human players to execute truly randomly, and the very act of trying to be unpredictable can sometimes lead to predictable patterns. This limits their application in practical risk management or complex financial modeling without significant simplification.

Mixed Strategies vs. Pure Strategy

The distinction between a mixed strategy and a pure strategy is fundamental in game theory. A pure strategy involves a player choosing a single, specific action with certainty (a probability of 1 for that action and 0 for all others). For example, a company might always choose to set a "High Price" in a given market. This is a deterministic choice.

In contrast, a mixed strategy involves a player randomizing their choice among two or more pure strategies, assigning a specific probability to each. The player does not commit to a single action but rather chooses their actions based on a calculated probability distribution. The purpose of a mixed strategy is to make a player's actions unpredictable, ensuring that no opponent can gain a consistent advantage by knowing the player's next move. While some games have equilibria in pure strategies, many do not, and in such cases, mixed strategies are necessary to achieve a Nash equilibrium. The selection between a pure or mixed strategy depends on the specific payoff structure of the game and whether a stable deterministic solution exists.

FAQs

What is the primary purpose of a mixed strategy?

The primary purpose of a mixed strategy is to introduce unpredictability into a player's actions, preventing an opponent from exploiting a fixed or deterministic pattern. By randomizing choices, a player can ensure that their opponent is indifferent to their own pure strategies, leading to a stable equilibrium where neither player can unilaterally improve their expected value.

Do real-world players actually use mixed strategies?

While perfect randomization as prescribed by game theory models may be difficult for humans to execute, the concept of mixed strategies informs real-world decision-making in various ways. Professionals in competitive fields like sports, poker, business, and military operations often employ probabilistic approaches or vary their tactics to avoid predictability, even if not explicitly calculating the precise probabilities of a Nash equilibrium.

What is a totally mixed strategy?

A totally mixed strategy is a type of mixed strategy where a player assigns a strictly positive probability to every one of their pure strategies. This means that the player never completely rules out any available action, ensuring that every option has some chance of being chosen, even if a small one.

How does a mixed strategy differ from guessing?

A mixed strategy is not mere guessing; it is a calculated randomization based on the payoffs of the game. The probabilities assigned to each pure strategy are determined such that an opponent gains no advantage by predicting any specific move. Guessing implies a lack of strategic calculation, whereas a mixed strategy is a deliberate and optimal strategic interaction.

Can all games have a mixed strategy equilibrium?

John Nash proved that every finite game (a game with a finite number of players and a finite number of pure strategies for each player) has at least one Nash equilibrium, which may involve mixed strategies. While not all games have a pure strategy Nash equilibrium, all finite games are guaranteed to have at least one mixed strategy Nash equilibrium.