Notation & Glossary

\[ \newcommand{\E}{\mathbb{E}} \newcommand{\R}{\mathbb{R}} \newcommand{\Prob}{\mathbb{P}} \newcommand{\BR}{\operatorname{BR}} \newcommand{\eps}{\varepsilon} \newcommand{\given}{\,\vert\,} \newcommand{\argmax}{\operatorname*{arg\,max}} \newcommand{\argmin}{\operatorname*{arg\,min}} \newcommand{\sm}{\setminus} \newcommand{\defeq}{\equiv} \]

A quick reference for the symbols and named concepts used throughout the book.

Symbols

Symbol Meaning
\(N=\{1,\dots,n\}\) set of players
\(i,\ -i\) a generic player; “everyone except \(i\)
\(S_i,\ s_i\) player \(i\)’s strategy space; a strategy
\(S_{-i},\ s_{-i}\) opponents’ strategy space; a profile of opponents’ strategies
\(s=(s_i,s_{-i})\) a strategy profile
\(v_i(\cdot)\) player \(i\)’s payoff function
\(\Gamma=(N,\{S_i\},\{v_i\})\) a normal-form game
\(\Delta S_i,\ \sigma_i\) the simplex of mixed strategies over \(S_i\); a mixed strategy
\(\operatorname{supp}\sigma_i\) the support of \(\sigma_i\)
\(\BR_i(s_{-i})\) player \(i\)’s best-response correspondence
\(\E[\cdot]\) expectation (expected payoff)
\(\Prob\) a probability distribution / common prior
\(\succ\) the precedence relation on nodes of a game tree
\(H_i,\ h_i\) player \(i\)’s collection of information sets; one information set
\(A_i(x)\) actions available to the mover at node \(x\)
\(z(s)\) the terminal node (outcome) induced by profile \(s\)
\(\delta\in(0,1)\) discount factor in repeated/dynamic games
\(\Theta_i,\ \theta_i\) player \(i\)’s type space; a type
\(\phi_i(\theta_{-i}\given\theta_i)\) \(i\)’s posterior belief over others’ types
\(\mu\) a system of beliefs over nodes within information sets
\(F,\ f\) a type/value distribution and its density (auctions)

Solution concepts at a glance

Complete information Incomplete information
Static Nash equilibrium (2  Static Games with Complete Information) Bayesian Nash equilibrium (4  Static Games with Incomplete Information)
Dynamic Subgame-perfect equilibrium (3  Dynamic Games with Complete Information) Perfect Bayesian equilibrium (5  Dynamic Games with Incomplete Information)

Each concept down or across the table is a refinement of the one before it, introduced to rule out predictions that the cruder concept cannot: incredible threats (handled by subgame perfection) and arbitrary off-path beliefs (handled by perfect Bayesian equilibrium and its refinements).

Glossary

Static games, complete information (2  Static Games with Complete Information)

  • Normal-form game — players, strategy spaces, payoff functions.
  • Strictly dominated / dominant strategy — a strategy that does strictly worse / strictly better than an alternative against every opponent profile.
  • Strictly dominant strategy equilibrium — every player plays a dominant strategy (e.g. \((C,C)\) in the Prisoner’s Dilemma).
  • Pareto optimal — no other profile makes someone better off without making someone worse off.
  • IESDS — iterated elimination of strictly dominated strategies; relies on common knowledge of rationality; order-independent for finite games.
  • Best response — a payoff-maximising reply to a fixed opponent profile.
  • Nash equilibrium (Definition 2.8) — a profile in which everyone best-responds to everyone else; equivalently, no player has a profitable deviation.
  • Mixed strategy — a probability distribution over pure strategies; the indifference principle (Lemma 2.2) characterises mixed best responses. Every finite game has a (possibly mixed) equilibrium (Theorem 2.1).
  • Cournot / Bertrand / Hotelling — quantity competition / price competition / spatial (electoral) competition; the last yields the median voter theorem.

Dynamic games, complete information (3  Dynamic Games with Complete Information)

  • Extensive form / game tree — moves, information, and payoffs in sequence.
  • Information set — decision nodes a player cannot tell apart; singletons mean perfect information.
  • Perfect recall — no player forgets what they once knew.
  • Pure / mixed / behavioral strategy — a contingent plan over information sets; a randomisation over plans; independent randomisation at each information set (Kuhn’s theorem links the last two).
  • Backward induction — solving a finite perfect-information game from its end.
  • Subgame; subgame-perfect equilibrium (Definition 3.11) — a profile that is a Nash equilibrium in every subgame; rules out non-credible threats.
  • One-shot deviation principle (Theorem 3.6) — to check subgame perfection it suffices to rule out single-stage profitable deviations.
  • Repeated game — a stage game played repeatedly; the discount factor \(\delta\) measures patience. Grim-trigger strategies sustain cooperation when players are patient enough (folk-theorem logic).
  • Rubinstein bargaining — alternating offers over an infinite horizon; a unique stationary subgame-perfect equilibrium with shares \(\tfrac{1}{1+\delta}\).

Static games, incomplete information (4  Static Games with Incomplete Information)

  • Bayesian game (Definition 4.1) — players have private types drawn from a common prior; strategies are type-contingent.
  • Bayesian Nash equilibrium (Definition 4.3) — every type best-responds given its posterior beliefs; equivalently, a Nash equilibrium of the Harsanyi extensive-form game with a move by Nature.
  • Adverse selection — private information about quality unravels trade (the market for lemons).
  • Pivotal voter — in committee voting, a rational juror conditions on being decisive, so voting one’s own signal need not be an equilibrium.
  • Purification — a mixed equilibrium reinterpreted as the limit of pure-strategy Bayesian equilibria of nearby games with small private shocks.
  • Auctions — second-price/English (truthful bidding weakly dominant); first-price/Dutch (shade your bid); revenue equivalence (standard formats raise the same expected revenue); winner’s curse (winning is bad news about a common value).

Dynamic games, incomplete information (5  Dynamic Games with Incomplete Information)

  • System of beliefs (Definition 5.2) — probabilities over the nodes within each information set.
  • Perfect Bayesian equilibrium (Definition 5.3) — strategies sequentially rational given beliefs, with on-path beliefs derived by Bayes’ rule.
  • Signaling game — an informed player takes a costly action that may credibly reveal type; separating (types choose distinct signals) vs pooling (types choose the same signal) equilibria. Spence’s education model: schooling signals ability even when unproductive, because imitation is too costly for the low type.
  • Intuitive criterion — a refinement (Cho–Kreps) that disciplines off-path beliefs and kills implausible pooling equilibria.
  • Cheap talk — costless, non-binding messages (Crawford–Sobel); informative communication is possible only when the sender’s and receiver’s interests are close enough (bias \(b<\tfrac14\) for a two-message equilibrium in the canonical uniform-quadratic model).