Perfect information is an informational condition used in game theory and economics. In its precise game-theoretic sense, a game has perfect information when every player, whenever required to act, knows the complete history of actions and observed chance outcomes leading to that decision. In broader economic usage, the term describes an idealized situation in which decision-makers possess all information relevant to their choices, such as prices and product characteristics. These meanings are related but distinct: observing every previous move does not necessarily mean knowing other players’ preferences, future actions, or future random outcomes.
Perfect information in game theory
Definition and information sets
An extensive-form game represents interaction as a tree. Nodes describe positions reached after particular histories, branches represent actions or chance outcomes, and terminal nodes specify payoffs.
An information set groups decision nodes that a player cannot distinguish when choosing an action. A game has perfect information if every information set belonging to a player contains exactly one decision node. Formally, if denotes the collection of information sets for player , perfect information requires
Thus, an acting player knows exactly where play stands in the game tree. A game has imperfect information if at least one player must choose without knowing which of several possible decision nodes has been reached.
Perfect information concerns what players observe, not how intelligently they use those observations. It does not itself imply rationality, unlimited computing power, agreement about strategy, or knowledge of future choices.
Examples
Chess, Go, and tic-tac-toe are standard examples of perfect-information games: the board and the history of moves are observable, and there are no privately held cards or concealed actions. Poker is an imperfect-information game because players do not observe opponents’ private cards.
Simultaneous decisions also ordinarily create imperfect information. In a simultaneous price-setting game, each firm chooses its price without observing its rival’s current choice. Even if the rules, costs, and possible actions are known, the game is not one of perfect information.
Randomness does not necessarily destroy perfect information. A game may include a publicly observed die roll before a player acts. The future roll is unknown beforehand, but its realized outcome is observed when relevant to the next decision. By contrast, a privately observed random draw introduces hidden information.
Distinction from related concepts
Complete information
Complete information concerns knowledge of the game’s underlying structure, especially players’ payoff functions and payoff-relevant characteristics, often called types. Perfect information concerns observation of the history of play.
The distinction can be illustrated by two cases:
- A simultaneous game with publicly known payoffs can have complete but imperfect information: players know the payoff structure but cannot observe one another’s current choices.
- A sequential interaction may have fully observed actions while a participant’s valuation or cost remains private. Its observable action history is perfect, but its payoff information is incomplete.
There is a technical qualification to the second case. In the standard representation of incomplete information, private types are modeled as initial moves by nature that are not observed by everyone. The resulting expanded game therefore has non-singleton information sets and is not a perfect-information extensive-form game. The terminology depends on whether the discussion concerns the observable sequence of actions or the full formal representation, including private types.
Symmetric information and common knowledge
Symmetric information means that participants possess the same information. They may nevertheless share substantial ignorance. Two traders who both lack knowledge of a product’s quality can be symmetrically informed without having perfect information.
Conversely, information asymmetry arises when some participants have relevant information that others lack. It is especially important when hidden product quality or private actions influence economic outcomes.
Common knowledge is a separate requirement: everyone knows a proposition, everyone knows that everyone knows it, and so on. Formal analyses often assume that the rules and relevant rationality conditions are common knowledge. This assumption should not be confused with observation of previous moves.
Strategic analysis
Backward induction
Finite games with perfect information can often be analyzed by backward induction. Analysis begins at the last decision nodes, identifies optimal actions there, and then works backward toward the initial move.
For example, an incumbent firm may threaten an entrant with an aggressive price response. If that response would be unprofitable once entry had occurred, backward induction can reveal that the threat is not credible. The entrant’s decision then depends on the response that would actually be optimal, rather than merely on the announced threat.
In a finite perfect-information game with specified payoffs, backward induction yields a pure-strategy subgame-perfect equilibrium. This refines Nash equilibrium by requiring equilibrium behavior in every subgame, including those not reached along the predicted path. If several actions yield identical payoffs, multiple equilibria may exist.
Perfect information alone does not guarantee a unique outcome, an outcome beneficial to everyone, or a practically computable solution. Infinite games require additional conditions for comparable existence results.
Observability versus computational difficulty
A player can observe the full history without being able to calculate the best move. Chess illustrates the difference between informational transparency and computational tractability: the position is visible, but the number of possible continuations is enormous.
This distinction connects perfect-information models with bounded rationality. A model may eliminate uncertainty about past actions while retaining substantial difficulty in reasoning about future play.
Perfect information in market economics
In microeconomics, perfect information is commonly used as a benchmark in which buyers and sellers know the relevant prices, qualities, contractual terms, and trading opportunities. The exact informational assumptions depend on the model.
For a consumer comparing identical products, full knowledge of available prices removes uncertainty about where the cheapest offer can be found. For a buyer evaluating a used asset, knowledge of quality removes uncertainty arising from the seller’s superior knowledge. These are different informational problems, even though both may be described as departures from perfect information.
Perfect information is often associated with textbook models of perfect competition, but it is not sufficient to establish competition. A fully informed buyer may still face a monopolist, and well-informed firms may still possess market power. Information assumptions must be distinguished from assumptions about entry, the number of traders, product differentiation, and strategic conduct.
Likewise, full information does not by itself establish Pareto efficiency. Externalities, missing markets, and other institutional features can prevent efficient allocation even when relevant facts are known.
Imperfect information and economic institutions
Relaxing perfect-information assumptions helps explain several important economic phenomena:
- Adverse selection: before a transaction, one party knows more about a relevant characteristic, such as product quality or an insurance applicant’s risk.
- Moral hazard: actions affecting outcomes are not fully observable or contractible, such as effort after an agreement is signed.
- Search costs: buyers and sellers expend time or resources discovering prices, qualities, and trading partners.
- Signaling and screening: informed parties communicate private characteristics, or uninformed parties design choices intended to reveal them.
These distinctions matter because different information problems require different models. A public list of prices may reduce price-search costs without revealing product quality or making subsequent effort observable.
In mechanism design and contract theory, the central question is often how rules and contracts operate when relevant information remains private. Such models do not assume that disclosure automatically eliminates strategic incentives.
Value and limits of the benchmark
Perfect information provides a reference point for separating informational obstacles from other sources of economic outcomes. Comparing models with and without hidden information can clarify why monitoring, reputation, certification, or intermediaries matter.
For a single decision-maker, additional information has nonnegative expected value when it is free, can be ignored, and does not otherwise alter the available choices or environment. This result does not imply that acquiring information is always worthwhile: acquisition and processing can be costly.
Nor does it establish that greater transparency benefits every participant in a strategic setting. Public information can change bargaining behavior, competitive responses, and incentives. Information useful to one participant may disadvantage another, and eliminating one information gap need not eliminate every source of inefficiency.
Finally, perfect information should not be equated with certainty. A decision-maker may know all current facts and the relevant probability distributions while remaining uncertain about future events. The defining issue is which information is available at the moment of choice, not whether the future is already determined or known.