Lesson overview | Lesson overview | Next part

Minimax Theorem: Part 1: Intuition to 2. Formal Definitions

1. Intuition

Intuition develops the part of minimax theorem specified by the approved Chapter 23 table of contents. The treatment is game-theoretic, not merely an optimization recipe.

1.1 adversarial decision-making

Adversarial decision-making belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$v^- = \max_{\mathbf{p}\in\Delta_m}\min_{\mathbf{q}\in\Delta_n}\mathbf{p}^\top A\mathbf{q}.$$

The formula gives the mathematical handle for adversarial decision-making. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

Strategic interaction begins when another decision maker can react to the policy being studied. Stability means the modeled behavior remains defensible after that reaction is allowed.

Worked reading.

Start with one proposed joint behavior. Freeze everyone except one player, compute that player's best alternative, then repeat for every player. If a profitable switch exists, the behavior is not strategically stable.

Three examples of adversarial decision-making:

1. A guardrail remains effective after attackers see examples of blocked prompts.
2. A model-routing policy remains attractive after providers update prices.
3. A self-play policy cannot be easily exploited by a newly trained opponent.

Two non-examples clarify the boundary:

1. A high average score on a fixed dataset.
2. A local minimum of one model's loss with no opponent.

Proof or verification habit for adversarial decision-making:

The verification habit is adversarial: search for profitable deviations rather than only confirming the proposed behavior works in the original scenario.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, adversarial decision-making is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

This is the mathematical shift from offline ML to deployed AI systems where users, competitors, and automated attacks learn from the model.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using adversarial decision-making responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: State the adaptation channel: what can the other side observe, change, and optimize?

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Adversarial decision-making gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.

1.2 worst-case loss

Worst-case loss belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$v^+ = \min_{\mathbf{q}\in\Delta_n}\max_{\mathbf{p}\in\Delta_m}\mathbf{p}^\top A\mathbf{q}.$$

The formula gives the mathematical handle for worst-case loss. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

Minimax reasoning chooses a strategy by its guaranteed performance against the strongest opponent response in a zero-sum game.

Worked reading.

The row player computes $\max_p\min_q p^\top A q$ while the column player computes $\min_q\max_p p^\top A q$. The minimax theorem says these values agree for finite zero-sum games.

Three examples of worst-case loss:

1. Robust classification against bounded perturbations.
2. A discriminator maximizing the generator's loss in a simplified GAN objective.
3. Worst-case evaluation where the tester chooses the hardest valid case.

Two non-examples clarify the boundary:

1. Average validation loss over a fixed dataset.
2. General-sum bargaining where both players can gain together.

Proof or verification habit for worst-case loss:

The LP-duality proof writes each player's guarantee as a linear program; strong duality equates the two optimal values.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, worst-case loss is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

Minimax is the mathematical backbone of adversarial robustness, but only relative to the stated threat model.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using worst-case loss responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: Check zero-sum structure before importing minimax conclusions.

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Worst-case loss gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.

1.3 maximin vs minimax

Maximin vs minimax belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$\max_{\mathbf{p}\in\Delta_m}\min_j (A^\top\mathbf{p})_j = \min_{\mathbf{q}\in\Delta_n}\max_i (A\mathbf{q})_i.$$

The formula gives the mathematical handle for maximin vs minimax. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

Minimax reasoning chooses a strategy by its guaranteed performance against the strongest opponent response in a zero-sum game.

Worked reading.

The row player computes $\max_p\min_q p^\top A q$ while the column player computes $\min_q\max_p p^\top A q$. The minimax theorem says these values agree for finite zero-sum games.

Three examples of maximin vs minimax:

1. Robust classification against bounded perturbations.
2. A discriminator maximizing the generator's loss in a simplified GAN objective.
3. Worst-case evaluation where the tester chooses the hardest valid case.

Two non-examples clarify the boundary:

1. Average validation loss over a fixed dataset.
2. General-sum bargaining where both players can gain together.

Proof or verification habit for maximin vs minimax:

The LP-duality proof writes each player's guarantee as a linear program; strong duality equates the two optimal values.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, maximin vs minimax is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

Minimax is the mathematical backbone of adversarial robustness, but only relative to the stated threat model.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using maximin vs minimax responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: Check zero-sum structure before importing minimax conclusions.

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Maximin vs minimax gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.

1.4 zero-sum games

Zero-sum games belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$\min_\theta \max_{\boldsymbol{\delta}\in\mathcal{S}} \mathcal{L}(f_\theta(\mathbf{x}+\boldsymbol{\delta}),y).$$

The formula gives the mathematical handle for zero-sum games. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

Minimax reasoning chooses a strategy by its guaranteed performance against the strongest opponent response in a zero-sum game.

Worked reading.

The row player computes $\max_p\min_q p^\top A q$ while the column player computes $\min_q\max_p p^\top A q$. The minimax theorem says these values agree for finite zero-sum games.

Three examples of zero-sum games:

1. Robust classification against bounded perturbations.
2. A discriminator maximizing the generator's loss in a simplified GAN objective.
3. Worst-case evaluation where the tester chooses the hardest valid case.

Two non-examples clarify the boundary:

1. Average validation loss over a fixed dataset.
2. General-sum bargaining where both players can gain together.

Proof or verification habit for zero-sum games:

The LP-duality proof writes each player's guarantee as a linear program; strong duality equates the two optimal values.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, zero-sum games is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

Minimax is the mathematical backbone of adversarial robustness, but only relative to the stated threat model.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using zero-sum games responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: Check zero-sum structure before importing minimax conclusions.

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Zero-sum games gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.

1.5 why minimax appears in robust ML

Why minimax appears in robust ml belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$v^- = \max_{\mathbf{p}\in\Delta_m}\min_{\mathbf{q}\in\Delta_n}\mathbf{p}^\top A\mathbf{q}.$$

The formula gives the mathematical handle for why minimax appears in robust ml. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

Minimax reasoning chooses a strategy by its guaranteed performance against the strongest opponent response in a zero-sum game.

Worked reading.

The row player computes $\max_p\min_q p^\top A q$ while the column player computes $\min_q\max_p p^\top A q$. The minimax theorem says these values agree for finite zero-sum games.

Three examples of why minimax appears in robust ml:

1. Robust classification against bounded perturbations.
2. A discriminator maximizing the generator's loss in a simplified GAN objective.
3. Worst-case evaluation where the tester chooses the hardest valid case.

Two non-examples clarify the boundary:

1. Average validation loss over a fixed dataset.
2. General-sum bargaining where both players can gain together.

Proof or verification habit for why minimax appears in robust ml:

The LP-duality proof writes each player's guarantee as a linear program; strong duality equates the two optimal values.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, why minimax appears in robust ml is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

Minimax is the mathematical backbone of adversarial robustness, but only relative to the stated threat model.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using why minimax appears in robust ml responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: Check zero-sum structure before importing minimax conclusions.

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Why minimax appears in robust ml gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.

2. Formal Definitions

Formal Definitions develops the part of minimax theorem specified by the approved Chapter 23 table of contents. The treatment is game-theoretic, not merely an optimization recipe.

2.1 two-player zero-sum game

Two-player zero-sum game belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$v^+ = \min_{\mathbf{q}\in\Delta_n}\max_{\mathbf{p}\in\Delta_m}\mathbf{p}^\top A\mathbf{q}.$$

The formula gives the mathematical handle for two-player zero-sum game. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

Minimax reasoning chooses a strategy by its guaranteed performance against the strongest opponent response in a zero-sum game.

Worked reading.

The row player computes $\max_p\min_q p^\top A q$ while the column player computes $\min_q\max_p p^\top A q$. The minimax theorem says these values agree for finite zero-sum games.

Three examples of two-player zero-sum game:

1. Robust classification against bounded perturbations.
2. A discriminator maximizing the generator's loss in a simplified GAN objective.
3. Worst-case evaluation where the tester chooses the hardest valid case.

Two non-examples clarify the boundary:

1. Average validation loss over a fixed dataset.
2. General-sum bargaining where both players can gain together.

Proof or verification habit for two-player zero-sum game:

The LP-duality proof writes each player's guarantee as a linear program; strong duality equates the two optimal values.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, two-player zero-sum game is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

Minimax is the mathematical backbone of adversarial robustness, but only relative to the stated threat model.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using two-player zero-sum game responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: Check zero-sum structure before importing minimax conclusions.

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Two-player zero-sum game gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.

2.2 payoff matrix $A$

Payoff matrix $a$ belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$\max_{\mathbf{p}\in\Delta_m}\min_j (A^\top\mathbf{p})_j = \min_{\mathbf{q}\in\Delta_n}\max_i (A\mathbf{q})_i.$$

The formula gives the mathematical handle for payoff matrix $a$. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

Players, actions, and payoffs define the interface of a game. If any one of them is vague, the equilibrium claim is usually vague too.

Worked reading.

A payoff matrix is a compact table: rows are one player's actions, columns are another player's actions, and entries are utilities or losses induced by the joint action.

Three examples of payoff matrix $a$:

1. A row action chooses a defense, while a column action chooses an attack family.
2. An agent set lists every model or tool-using process that can affect reward.
3. A utility function converts accuracy, safety, latency, and cost into strategic incentives.

Two non-examples clarify the boundary:

1. A metric with no actor who optimizes it.
2. An action that is impossible in deployment but included for convenience.

Proof or verification habit for payoff matrix $a$:

Before proving anything, audit the model specification: every allowed action must map to a payoff for every player.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, payoff matrix $a$ is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

Payoff design is AI system design. The game will faithfully optimize the incentives it is given, including bad incentives.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using payoff matrix $a$ responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: Can you name each player, enumerate or parameterize its actions, and compute its payoff from a joint action?

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Payoff matrix $a$ gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.

2.3 mixed strategies $\mathbf{p},\mathbf{q}$

Mixed strategies $\mathbf{p},\mathbf{q}$ belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$\min_\theta \max_{\boldsymbol{\delta}\in\mathcal{S}} \mathcal{L}(f_\theta(\mathbf{x}+\boldsymbol{\delta}),y).$$

The formula gives the mathematical handle for mixed strategies $\mathbf{p},\mathbf{q}$. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

A mixed strategy is a probability distribution over actions. In equilibrium, actions used with positive probability must usually give the same expected payoff; otherwise probability can move to the better action.

Worked reading.

In matching pennies, the row player is indifferent only when the column player randomizes heads and tails equally. The same calculation makes the column player indifferent, giving the $1/2,1/2$ equilibrium.

Three examples of mixed strategies $\mathbf{p},\mathbf{q}$:

1. Randomized audits that make attackers uncertain.
2. Stochastic decoding policies that prevent deterministic exploitation.
3. Exploration policies in self-play where pure repetition would be exploited.

Two non-examples clarify the boundary:

1. Adding noise after choosing a deterministic losing action.
2. A distribution that assigns probability to an action with strictly lower payoff while another supported action is better.

Proof or verification habit for mixed strategies $\mathbf{p},\mathbf{q}$:

Set expected payoffs of supported actions equal, solve for probabilities, then verify unsupported actions are not better.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, mixed strategies $\mathbf{p},\mathbf{q}$ is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

Mixed strategies explain why robust systems often randomize: predictability can be a vulnerability when opponents adapt.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using mixed strategies $\mathbf{p},\mathbf{q}$ responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: Check both support equality and off-support inequalities.

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Mixed strategies $\mathbf{p},\mathbf{q}$ gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.

2.4 game value $v$

Game value $v$ belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$v^- = \max_{\mathbf{p}\in\Delta_m}\min_{\mathbf{q}\in\Delta_n}\mathbf{p}^\top A\mathbf{q}.$$

The formula gives the mathematical handle for game value $v$. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

The game value is the payoff both sides can guarantee in a finite zero-sum game when they use optimal mixed strategies.

Worked reading.

The row player's guarantee is the lower value; the column player's guarantee is the upper value. The minimax theorem states these agree in finite zero-sum games.

Three examples of game value $v$:

1. The value of rock-paper-scissors is zero.
2. A robust classifier's worst-case loss is the value of a specified attack game.
3. Yao-style lower bounds swap randomized algorithms and input distributions under minimax reasoning.

Two non-examples clarify the boundary:

1. A general-sum welfare score.
2. A last-iterate training reward.

Proof or verification habit for game value $v$:

The finite proof can be read through LP duality: the column player's dual program certifies the same scalar value as the row player's primal program.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, game value $v$ is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

The value is useful because it is a guarantee, not a hope about average-case behavior.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using game value $v$ responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: Check the assumptions: finite action sets or the right compactness, convexity, and continuity conditions for extensions.

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Game value $v$ gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.

2.5 saddle point

Saddle point belongs to the canonical scope of Minimax Theorem. The central object is not a single optimizer but a system of decision makers whose objectives interact.

For this subsection, the working scope is zero-sum matrix games, maximin and minimax values, saddle points, LP duality, no-regret approximation, and robust AI objectives. We use players, action sets, strategies, payoffs, and response rules. The key question is whether a proposed behavior is stable when another agent adapts.

$$v^+ = \min_{\mathbf{q}\in\Delta_n}\max_{\mathbf{p}\in\Delta_m}\mathbf{p}^\top A\mathbf{q}.$$

The formula gives the mathematical handle for saddle point. In game theory, this expression should always be read with the opponent's decision rule in mind. A policy that is optimal in isolation may be exploitable once another player observes and responds to it.

Operational definition.

Minimax reasoning chooses a strategy by its guaranteed performance against the strongest opponent response in a zero-sum game.

Worked reading.

The row player computes $\max_p\min_q p^\top A q$ while the column player computes $\min_q\max_p p^\top A q$. The minimax theorem says these values agree for finite zero-sum games.

Three examples of saddle point:

1. Robust classification against bounded perturbations.
2. A discriminator maximizing the generator's loss in a simplified GAN objective.
3. Worst-case evaluation where the tester chooses the hardest valid case.

Two non-examples clarify the boundary:

1. Average validation loss over a fixed dataset.
2. General-sum bargaining where both players can gain together.

Proof or verification habit for saddle point:

The LP-duality proof writes each player's guarantee as a linear program; strong duality equates the two optimal values.

single-agent optimization:    choose theta to minimize L(theta)
game-theoretic optimization:  choose pi_i while others choose pi_-i
adversarial objective:        choose defense against best attack
multi-agent learning:         policies change the environment itself

In AI systems, saddle point is useful because modern models are deployed into adaptive environments: users learn prompt tricks, attackers search for failures, evaluators change rubrics, and other agents compete for resources.

Minimax is the mathematical backbone of adversarial robustness, but only relative to the stated threat model.

Notebook implementation will use small synthetic payoff matrices and learning dynamics. This keeps the mathematics executable while avoiding external datasets or heavyweight game solvers.

Checklist for using saddle point responsibly:

• State the players and their objectives.
• State the action spaces and information structure.
• Decide whether the game is zero-sum, general-sum, cooperative, or adversarial.
• Identify pure, mixed, or policy strategies.
• Compute best responses or exploitability before claiming stability.
• Separate equilibrium analysis from welfare analysis.
• Explain what changes if opponents adapt.

Local diagnostic: Check zero-sum structure before importing minimax conclusions.

This chapter follows Chapter 22 by adding strategic adaptation. Causal inference asks what happens when we intervene. Game theory asks what happens when other decision makers anticipate or respond to that intervention.

Modern AI makes the distinction practical. A deployed model can be optimized against by users, attackers, competitors, automated evaluators, and other models. Saddle point gives the language to reason about that pressure.

A final diagnostic question is whether a decision remains good after another agent learns from it. If not, the analysis needs game theory, not just prediction, causality, or optimization.