Dominated Strategy¶

Rule out an action that yields a lower payoff than some alternative for every possible profile of opponents' actions — a belief-free test requiring no model of the opponent, licensing iterated elimination and the dominant-strategy robustness that mechanism design targets.

Core Idea¶

A dominated strategy is an action available to a player in a strategic-form game that yields a weakly (or strictly) lower payoff than at least one other action the player could choose, for every possible combination of actions by the other players. Because a dominated strategy never performs better than its dominating alternative — regardless of what opponents do — a rational player has no reason to play it, and common knowledge of rationality allows each player to assume their opponents will not play dominated strategies either.

This assumption licenses iterated elimination of dominated strategies: once dominated strategies are removed from all players' action sets, other strategies may become dominated in the reduced game, and the process can repeat. In the best cases the iterated process collapses the game to a single outcome without any equilibrium computation. In the Prisoner's Dilemma, Defect strictly dominates Cooperate for each player — a one-round elimination leaves (Defect, Defect) as the unique surviving outcome. In the Vickrey second-price sealed-bid auction, bidding one's true valuation weakly dominates every other bid: bidding below true value risks losing an auction one would have won at a profit, while bidding above risks winning at a price above one's value, and neither risk is compensated by any gain regardless of the other bidders' behavior. The second result — truth-telling as a dominant strategy — is the foundation of mechanism design's gold standard, dominant-strategy incentive compatibility (DSIC): a mechanism satisfying DSIC makes truthful reporting optimal for each participant without any assumptions about the beliefs, types, or strategies of others, making it robust in a way that equilibrium-dependent mechanisms are not.

The concept draws a sharp distinction between two solution ideas that are easy to conflate. Nash equilibrium requires that each player's strategy be a best response to what opponents actually play — it is a fixed-point condition on beliefs and strategies simultaneously. Dominated-strategy elimination requires only that each player avoid strategies that are worse than something else regardless of opponents' behavior — it is a dominance condition that can be applied without modeling opponents at all. The two concepts are related: every Nash equilibrium survives iterated elimination of strictly dominated strategies, but not conversely. The practical implication is that dominance analysis is the prior step — clear out what no rational player would ever do, then look for equilibria in what remains.

Structural Signature¶

Sig role-phrases:

the strategic-form game — players each with an explicit action set and payoffs that depend on the joint action profile
the dominance comparison — the defining test: one action yields a lower payoff than some alternative for every possible profile of opponents' actions
the strict-vs-weak grade — worse for every opponent profile (strict) versus worse-or-equal everywhere and strictly worse somewhere (weak), governing whether elimination is safe to iterate and whether the survivor is unique
the belief-free guarantee — the engineered property: a dominated action can be discarded with no probability model of opponents, because it never wins regardless of what they do
the common-knowledge-of-rationality lever — rational opponents will not play their dominated actions either, licensing their deletion from the game
the iterated elimination procedure — deleting dominated actions can make further actions dominated, repeating to monotonically shrink the game and sometimes collapse it to a single outcome
the DSIC robustness target — the design payoff: engineer the wanted action (truthful reporting) to dominate, so it is optimal regardless of anyone's beliefs, types, or strategies — robust where an equilibrium-only guarantee is fragile
the every-contingency quantifier limit — the boundary: an action merely worse in expectation is not dominated; dropping the for-every-profile quantifier smuggles a belief model back into the belief-free move

What It Is Not¶

Not "worse on average." The defining quantifier is worse for every possible profile of opponents' actions, not worse in expectation against their likely play. A choice that loses only on average is not dominated; treating it as such drops the for-every-profile quantifier and smuggles a belief model back into what is precisely the belief-free move.
Not a best-response condition. Dominance is not Nash equilibrium under another name: Nash requires each strategy to be a best response to what opponents actually play — a simultaneous fixed-point on beliefs and strategies — whereas dominance requires only that an action be worse than some alternative regardless of opponents, applicable with no model of them. Every Nash equilibrium survives iterated elimination of strictly dominated strategies, but not conversely, so dominance is the weaker, prior pass.
Not a claim that a dominated strategy is irrational to consider — only to play. The concept does not say the action is incoherent or that the player is foolish for having it in the action set; it says a rational player has no reason to choose it, because something else is at least as good no matter what. The elimination is a normative discard, not a psychological diagnosis.
Not interchangeable in its strict and weak forms. Strict dominance (worse for every opponent profile) is unconditionally safe to iterate and yields unique survivors; weak dominance (worse-or-equal everywhere, strictly worse somewhere) is not — the order of weak elimination can change which outcomes survive. Collapsing the two licenses unsafe iterated deletions.
Not about outcomes or mutual benefit. Dominance here ranks the strategies of one player across opponents' profiles, not the desirability of joint outcomes for all players. An action being dominated says nothing about whether the resulting outcome is good for everyone — the Prisoner's Dilemma's surviving (Defect, Defect) is the dominance-mandated play yet is jointly worse than mutual cooperation.
Not a concept that applies outside agent-with-strategy-set settings. "Dominated strategy" is a defined relation on a strategy space, so where there is no decision-making agent choosing among actions against others' choices — a thermostat, an ecosystem reaching equilibrium, a market clearing a price — it does not weaken into analogy, it simply has nothing to attach to. Absent the strategy space, the definition is vacuous rather than approximate.

Scope of Application¶

Because a dominated strategy is an analytical construct — a solution-concept test plus an elimination procedure, not a causal mechanism — it applies wherever its precondition holds: decision-making agents, explicit strategy sets, and payoffs that depend on the joint action profile. The fields below are real deployments of the identical test across strategic-decision settings, not analogues. The boundary is reach versus over-reading: absent a strategy space (a thermostat, a clearing market) the definition is simply vacuous, and the thin "discard the under-every-contingency-worse option" kernel travels via decision-theoretic dominance, not this construct.

Game-theory solution analysis — the workhorse first step, the belief-free pre-equilibrium reduction (iterated elimination of dominated strategies) run before any Nash search.
Auction theory — certifying that bidding one's true valuation weakly dominates every other bid in the Vickrey second-price sealed-bid auction.
Mechanism design — the gold-standard target, dominant-strategy incentive compatibility (DSIC), under which truthful reporting is optimal for each participant regardless of others' beliefs, types, or strategies.
Voting theory — the object of the Gibbard-Satterthwaite theorem, which characterizes when truthful voting can be dominant and proves no non-dictatorial rule makes it always so.
Bargaining and contracting analysis — the standard pre-equilibrium clearing of strategies no rational party would ever choose before equilibria are sought in what remains.
Single-agent decision theory under uncertainty — the thinner statewise form, where an act worse than an alternative under every state of the world is rejected regardless of the probability distribution (the construct's decision-theoretic edge, carried by general dominance).

Clarity¶

Naming dominated strategy isolates the one analytical move in game theory that requires no model of the opponent at all. The confusion it dissolves is the belief that every strategic judgment must rest on a guess about what the other player will do — a guess that demands probability assignments, type spaces, and equilibrium fixed points. Dominance shows that a whole class of actions can be ruled out before any of that machinery is invoked: if an action is worse than some alternative no matter what opponents do, a rational player discards it, and common knowledge of rationality lets every player discard their opponents' dominated actions too. This makes the reasoning belief-free, and that is precisely what licenses iterated elimination as the clean first pass that can sometimes solve a game outright.

The distinctions it sharpens are two. First, strict versus weak dominance — worse for every opponent profile, versus worse-or-equal everywhere and strictly worse somewhere — a difference that is not pedantic but governs whether elimination is safe to iterate and whether surviving outcomes are unique. Second, and more consequential for design, it makes legible the gap between a strategy being optimal given correct beliefs and a strategy being optimal regardless of beliefs. That gap is the whole point of dominant-strategy incentive compatibility: a mechanism in which truth-telling merely survives in equilibrium can unravel if participants misjudge each other, whereas one in which truth-telling dominates is robust to any beliefs, types, or strategies. Naming dominance lets a designer ask the sharper question — not "is honesty a best response here?" but "is honesty a dominant response, immune to what anyone believes about anyone else?"

Manages Complexity¶

Solving a strategic-form game in full generality is combinatorially heavy: the joint action space grows multiplicatively in players and actions, and finding Nash equilibria means solving a simultaneous fixed-point condition in which every player's optimal choice depends on probabilistic beliefs about every other's, which in turn depend on theirs. Dominance compresses that problem by supplying a belief-free reduction that runs before any of the fixed-point machinery. The analyst asks of each action only the local question "is this worse than some alternative no matter what the others do?" — a comparison internal to one player's payoffs, requiring no model of opponents — and deletes every action that fails it. Iterated elimination then propagates: deletions in one player's set can make further actions dominated for others, and the process repeats, monotonically shrinking the game. In the favorable cases the whole high-dimensional game collapses to a single surviving outcome with no equilibrium computation at all (one round of strict-dominance elimination resolves the Prisoner's Dilemma to mutual defection). Even when it does not solve the game, it clears out everything no rational player would ever do, so the residual equilibrium search runs over a far smaller space. The complexity move is to replace a global fixed-point over all players' beliefs with a sequence of one-player-at-a-time payoff comparisons.

The second, sharper compression is in mechanism design, where dominance collapses an otherwise intractable robustness question to a checkable one. To certify that a mechanism elicits honest behavior in the equilibrium sense, the designer must reason about the entire space of participants' beliefs, types, and conjectures about one another — and the guarantee can unravel if any participant misjudges the others. The dominant-strategy criterion (DSIC) compresses that whole space to nothing: if truthful reporting is dominant, it is optimal for each participant regardless of anyone's beliefs, types, or strategies, so the designer verifies incentive compatibility one participant at a time, against no model of the rest, and obtains a guarantee that holds across all belief configurations at once. The analyst stops asking the unbounded question "will honesty survive every way these agents might think about each other?" and asks the single bounded one "does honesty dominate for this agent?" — reading the mechanism's full-population robustness off a per-agent dominance check. The branch the designer tracks is correspondingly crisp: a strategy optimal only given correct beliefs (fragile, equilibrium-dependent) versus one optimal regardless of beliefs (robust, dominance-backed), with the strict-versus-weak distinction flagging whether the elimination is safe to iterate and whether the surviving outcome is unique.

Abstract Reasoning¶

Within game theory and solution-concept analysis the concept licenses reasoning moves that all exploit the belief-free dominance comparison and the robustness it confers.

Diagnostic — test each action for dominance without modeling the opponent, and classify the dominance as strict or weak. The signature move asks a local, opponent-free question of every action: the analyst reasons FROM "is this action worse than some alternative the player has, for every possible combination of opponents' actions?" TO "if so it is dominated, and a rational player will never play it" — a judgment internal to one player's payoffs, requiring no probability assignment over opponents. A second diagnostic move grades the dominance because the grade governs what follows: reasoning FROM "is the alternative strictly better for every opponent profile (strict), or better-or-equal everywhere and strictly better somewhere (weak)?" TO "whether elimination is safe to iterate and whether the surviving outcome is unique." A third diagnostic move applies common knowledge of rationality: reasoning FROM "no rational opponent will play their dominated strategies either" TO "I may delete their dominated actions from the game before reasoning further." The move is FROM a per-action payoff comparison TO a dominance verdict reached without any model of the other players.

Interventionist — eliminate dominated strategies as the prior step and iterate, or design a mechanism so the desired action dominates. The analytic interventionist move is to clear the board before searching for equilibria: reasoning FROM "these actions are dominated" TO "delete them, then re-examine the reduced game, since deletions can make further actions dominated" — iterated elimination predicted to shrink the game monotonically and, in favorable cases (the Prisoner's Dilemma in one round), to collapse it to a single surviving outcome with no equilibrium computation. A second, design-side interventionist move targets dominance as a goal: reasoning FROM "make truthful reporting dominant for each participant (DSIC)" TO "honesty is optimal regardless of anyone's beliefs, types, or strategies," predicted to yield a mechanism robust where an equilibrium-dependent one is fragile — as in the Vickrey second-price auction, where bidding true value is engineered to weakly dominate every other bid. The analyst reasons FROM "build the payoffs so the wanted action dominates" TO "the guarantee holds across all belief configurations at once."

Boundary-drawing — separate the dominance condition from the Nash best-response fixed-point, and optimal-given-beliefs from optimal-regardless-of-beliefs. A first boundary move holds two solution ideas apart that are easy to conflate: Nash equilibrium requires each strategy to be a best response to what opponents actually play — a simultaneous fixed-point on beliefs and strategies — whereas dominance requires only that a player avoid an action worse than something else regardless of opponents, applicable without modeling them at all. The analyst reasons FROM "does this judgment need a model of opponents' play, or not?" TO "Nash analysis versus dominance analysis," and notes the relation: every Nash equilibrium survives iterated elimination of strictly dominated strategies, but not conversely — so dominance is the prior, weaker pass. A second boundary move, decisive for design, separates a strategy optimal given correct beliefs (fragile, equilibrium-dependent, can unravel if participants misjudge each other) from one optimal regardless of beliefs (robust, dominance-backed), so the analyst reasons FROM "is honesty merely a best response, or a dominant response?" TO "is this mechanism robust to what anyone believes about anyone else?" A third boundary move keeps strict and weak dominance distinct, since only strict elimination is unconditionally safe to iterate.

Predictive — a dominated action is forecast never to be played, and a dominant-strategy mechanism is forecast robust where an equilibrium one fails. A forward move predicts behavior with no belief model: reasoning FROM "this action is dominated" TO "a rational player will not choose it, and common knowledge of rationality means opponents will not choose theirs," so the analyst forecasts the reduced game's play before any equilibrium is computed. A second predictive move forecasts the order of solution: reasoning FROM "clear out every dominated action first" TO "any remaining equilibrium search runs over a far smaller space, and may already be resolved." A third predictive move forecasts robustness: reasoning FROM "truth-telling is dominant under this mechanism" TO "it will be played across all belief configurations and survive participants misjudging one another," whereas FROM "truth-telling only survives in equilibrium" TO "the guarantee can unravel if any participant's beliefs are wrong" — predicting which mechanisms hold up under mis-specification and which do not.

Knowledge Transfer¶

Dominated strategy is an analytical construct — a solution-concept test plus an elimination procedure — rather than a causal mechanism, so the natural question is how far the test reaches, not where its mechanism stops becoming literal. Within game theory and its applied branches the construct transfers literally and exactly: wherever its precondition holds — decision-making agents, explicit strategy sets, and payoffs that depend on the joint action profile — the dominance comparison, iterated elimination, the strict-versus-weak distinction, and the dominant-strategy robustness criterion all apply without modification. That precondition is met across a broad sweep of strategic-decision settings, and the construct carries into each as the same operation: in auction theory it certifies that truth-telling weakly dominates in the Vickrey second-price auction; in mechanism design it is the gold-standard target, dominant-strategy incentive compatibility (DSIC), under which honest reporting is optimal regardless of others' beliefs, types, or strategies; in voting theory it is the object of the Gibbard-Satterthwaite theorem, which characterizes when truthful voting can be dominant and proves no non-dictatorial rule makes it always so; and it is the standard pre-equilibrium reduction in bargaining and contracting analysis. These are not analogical extensions but the same test deployed wherever a strategy space with interdependent payoffs exists — the home domain is wide, but it is one kind of substrate, which is exactly why dominated strategy is a domain-specific abstraction and not a substrate-spanning prime.

The boundary to mark is therefore reach versus over-reading, and the most important fact is a clean precondition failure rather than a mechanism-to-metaphor decay. Where there is no decision-making agent with an action set — a thermostat regulating temperature, an ecosystem reaching a stable state, a market clearing a price — the concept of a "dominated strategy" has nothing to attach to: there is no player choosing among strategies against others' choices, so the construct does not weaken into analogy, it simply does not apply. This is unlike a causal mechanism that can be loosely re-narrated off its home substrate; dominance is a defined relation on a strategy space, and absent the strategy space the definition is vacuous. The one over-reading to guard against is colloquial: calling an everyday option "dominated" when it is merely worse on average, or worse against the expected behavior of others, drops the defining quantifier — worse for every possible profile of others' actions. A choice that is only worse in expectation is not dominated; treating it as such smuggles a belief model back into what is supposed to be the belief-free move, and over-reads the construct past its warranted scope.

What genuinely generalizes beyond strategic games is the thinner decision-theoretic kernel the construct shares with dominance-style reasoning: an option worse than some alternative under every contingency should be discarded without needing to forecast which contingency obtains. That kernel recurs in single-agent decision theory under uncertainty (a statewise-dominated act is rejected regardless of the probability distribution over states) and in any choice problem with a "no matter what happens" comparison — and where the cross-domain lesson is that kind of contingency-proof elimination, it is carried by the general dominance pattern, not by "dominated strategy" with its opponents, common-knowledge-of-rationality, and DSIC machinery. That machinery is the home-bound cargo: it presupposes other strategic agents and the equilibrium apparatus it is defined against. The general "reject the under-every-contingency-worse option" idea travels via decision-theoretic dominance; the named game-theoretic construct, with its iterated elimination and incentive-compatibility role, stays within strategic-decision settings — the distinction Structural Core vs. Domain Accent makes precise below.

Examples¶

Canonical¶

Take the Prisoner's Dilemma with utilities: mutual cooperation pays (3, 3), mutual defection pays (1, 1), and unilateral defection pays 5 to the defector and 0 to the cooperator. Fix Player 1 and check Defect against Cooperate. If Player 2 cooperates, Defect yields 5 versus Cooperate's 3; if Player 2 defects, Defect yields 1 versus Cooperate's 0. Defect beats Cooperate in both columns, so Cooperate is strictly dominated — and the check never consulted any probability that Player 2 cooperates. By symmetry the same holds for Player 2. One round of elimination deletes Cooperate for both players, leaving (Defect, Defect) as the unique survivor, worth (1, 1) — jointly worse than the (3, 3) both forgo.

Mapped back: The 2×2 payoff table is the strategic-form game; checking Cooperate against Defect across both of Player 2's actions is the dominance comparison, and finding Cooperate worse in every column is the strict grade. That the check used no probability over Player 2's play is the belief-free guarantee. Deleting Cooperate for both and reading off the survivor is the iterated elimination procedure — and the survivor being jointly worse than (3, 3) shows dominance ranks a player's strategies, not outcomes.

Applied / In Practice¶

The Vickrey (second-price, sealed-bid) auction awards the item to the highest bidder but charges only the second-highest bid. Suppose your true value is $100. Bid truthfully: if the top rival bid is $80 you win and pay $80, netting $20; if it is $120 you lose and net $0. Now deviate. Shading to $70 only forfeits the $80 win you would have taken at a profit; overbidding to $130 wins the $120 case but forces you to pay $120 for a $100 item — a $20 loss. No rival configuration rewards misreporting, so bidding $100 weakly dominates every alternative, whatever others bid. This dominant-strategy logic underlies real second-price deployments: eBay's proxy-bidding system and single-item sealed-bid procurement approximate it precisely so that participants need not out-guess each other to bid well.

Mapped back: Engineering the payoffs so truthful bidding weakly dominates is the DSIC robustness target, and the fact that $100 is best "whatever others bid" is the belief-free guarantee — the reason the mechanism is robust where an equilibrium-only one is fragile. Comparing truthful bidding to every deviation across all rival bids is the dominance comparison, and the tie-in-some-cases, strictly-better-in-others pattern is exactly the weak grade.

Structural Tensions¶

T1: Belief-free certainty versus limited resolving power (a cheap pass that usually leaves the hard part). Dominance buys a rare thing — a verdict reached with no model of the opponent, no probabilities, no fixed point — and that cheapness is exactly why it is the right first pass. The tension is that the same belief-freedom bounds its reach: because it only discards actions worse under every opponent profile, most games have few or no dominated strategies, so iterated elimination frequently deletes nothing and leaves the full equilibrium problem intact. The Prisoner's Dilemma collapsing in one round is the advertised best case, not the typical one. The analyst faces a genuine trade: dominance is unconditionally safe and often uninformative, while Nash analysis is informative but demands the belief machinery dominance was prized for avoiding. Leaning on dominance alone risks mistaking "nothing was eliminated" for "nothing could be said." Diagnostic: Did dominance actually reduce this game, or is it the empty pass that now hands everything to equilibrium analysis?

T2: Strict versus weak dominance (safe iteration versus order-dependent reach). The two grades are not interchangeable, and each concedes what the other keeps. Strict dominance is unconditionally safe to iterate and yields unique survivors, but it catches only actions worse in every column with strict inequality, so it eliminates less. Weak dominance catches more — the Vickrey truth-telling result is weak — but iterated weak elimination is order-dependent: the sequence in which weakly dominated actions are removed can change which outcomes survive. The tension is that the more powerful eliminator is the less trustworthy one, so an analyst who wants weak dominance's extra reach must give up the clean, order-independent guarantee that makes strict elimination a solution concept at all. Choosing weak dominance to solve a game means accepting that a different deletion order might have solved it differently. Diagnostic: Is the elimination strict (safe to iterate in any order) or weak (extra reach, but the surviving set may depend on deletion order)?

T3: Dominant-strategy robustness versus its restrictiveness (the gold standard you often cannot have). DSIC is the strongest incentive guarantee available — truthful reporting optimal regardless of anyone's beliefs, types, or strategies, immune to participants misjudging each other — and that immunity is its whole appeal. The tension is that demanding so much rules out much: Gibbard-Satterthwaite shows no non-dictatorial voting rule can make honesty always dominant, and across mechanism design the dominant-strategy requirement often forces efficiency losses or impossibility where a weaker, belief-dependent equilibrium concept (Bayes-Nash) could achieve more. Insisting on dominance buys robustness at the price of what the mechanism can accomplish; accepting equilibrium incentive-compatibility buys reach at the price of fragility to mis-specified beliefs. The designer cannot generally have both maximal robustness and maximal performance. Diagnostic: Is dominant-strategy robustness worth the efficiency or feasibility it costs here, or does the setting justify a weaker equilibrium guarantee that achieves more?

T4: Ranking one player's strategies versus ranking joint outcomes (individually mandated, collectively bad). Dominance ranks the strategies of a single player across opponents' profiles; it says nothing about whether the resulting joint outcome is good for anyone. The tension is that the concept's clean individual-rationality verdict can drive the group to a jointly worse result: in the Prisoner's Dilemma the dominance-mandated (Defect, Defect) is what every rational player must play and is strictly worse for both than the (3, 3) they forgo. The very decisiveness of the strategy-level ranking is what produces the collectively bad outcome, so a solution concept that is impeccable per player is silent — even complicit — at the level of welfare. Reading a dominance verdict as endorsing the outcome, rather than merely the choice, misattributes to the concept a claim about joint desirability it never makes. Diagnostic: Is the dominance verdict being read as "this is the rational play for the player," or wrongly as "this is a good outcome for the players"?

T5: The for-every-profile quantifier versus the average-case temptation (where beliefs sneak back in). The construct's power rests entirely on a strong quantifier — worse for every possible profile of opponents' actions — and it is precisely this that keeps the move belief-free. The tension is that ordinary usage constantly erodes it: an option that is merely worse on average, or worse against opponents' likely play, feels dominated and gets called so, but that judgment silently reintroduces a probability model of the opponent, smuggling the belief machinery back into the one move designed to avoid it. A choice worse in expectation is not dominated, and treating it as such converts a contingency-proof elimination into a disguised best-response calculation. The discipline the concept demands is to hold the universal quantifier exactly, which is unnatural precisely because expected-value reasoning is the default everywhere else. Diagnostic: Is the action worse for every opponent profile, or only worse in expectation against their likely play — the latter is not dominance but a belief model in disguise?

T6: Autonomy versus reduction (a named game-theoretic construct or an instance of decision-theoretic dominance). "Dominated strategy" is a canonically defined solution concept with proprietary machinery — other strategic agents, common knowledge of rationality, iterated elimination, the Nash relationship, DSIC and Gibbard-Satterthwaite. Yet its portable kernel is thinner: discard an option worse than some alternative under every contingency without forecasting which contingency obtains is the substrate-neutral dominance pattern, which recurs in single-agent decision theory (a statewise-dominated act rejected regardless of the distribution over states) with no opponents at all. The tension is between a standalone strategic construct that earns its own elimination procedure and incentive-compatibility role, and the recognition that what travels to non-strategic choice problems is only the general contingency-proof-elimination idea. Crucially, absent a strategy space the named construct does not weaken into analogy — it simply has nothing to attach to, so the thin kernel is the only thing that leaves home. Diagnostic: Resolve toward the parent (decision-theoretic dominance) when the setting is any "worse under every contingency" choice; toward the named construct when there are strategic agents, interdependent payoffs, and an equilibrium apparatus to reduce.

Structural–Framed Character¶

Dominated strategy sits toward the structural end of the spectrum but stops short of the pole — best read as mixed-structural: a genuine contingency-proof-elimination test wearing game-theoretic machinery. On four of the five criteria its structural credentials are strong. Its evaluative weight is nil — the dominance comparison ranks one player's strategies across opponents' profiles and says nothing about the desirability of joint outcomes; the entry is explicit that the dominance-mandated (Defect, Defect) is jointly worse than mutual cooperation, so "dominated" is a normative discard of an action (never play what is worse no matter what), never a verdict on the player or the outcome. It is not human-practice-bound in the constitutive sense: dominance is a defined relation on a strategy space, holding wherever agents, action sets, and interdependent payoffs are present — including the fitness matrices evolutionary game theory reads off non-deliberating organisms — so no human practice constitutes it, and absent the strategy space it does not soften into analogy but simply has nothing to attach to. Its institutional origin is none: it is a formal solution concept derived within the theory, discovered rather than instituted by any survey or agency. And within its proper range cross-domain reuse is recognition, not import: across game-theoretic solution analysis, auctions, mechanism design, voting theory, and bargaining it is the same test on the same formal machinery, and the thinner kernel it shares with single-agent decision theory is the genuinely-same statewise dominance relation, not a metaphor.

What keeps it off the structural pole is the remaining criterion, vocab-travels, which it fails. The operative vocabulary distinctive to "dominated strategy" — opponents' profiles, common knowledge of rationality, iterated elimination, dominant-strategy incentive compatibility, the Nash-survival relationship — is irreducibly game-theoretic and does not float free of the strategic-agent form the way "growing quantity" or a statewise comparison does in a pure structural prime; within strategic settings it carries full content, but the only thing that leaves home is the thin belief-free-elimination idea, stripped of every strategic component. The portable structural skeleton is discard an option that is worse than some alternative under every contingency, without forecasting which contingency obtains. That skeleton is genuinely portable, and it is exactly the part the catalogue already carries as the general decision-theoretic dominance pattern that dominated strategy instantiates by filling the contingencies with opponents' action profiles and adding common-knowledge-of-rationality to license iteration; the cross-domain reach belongs to that parent, while the iterated-elimination, DSIC, and Gibbard-Satterthwaite machinery distinctive to the named construct is exactly the part that stays home. Its character: structural in skeleton — a real, evaluatively neutral, recognized-in-formalism contingency-proof-elimination test — but stated in game-theoretic vocabulary that pins it to the strategic-agent form, leaving it mixed-structural rather than the free-floating dominance prime it specializes.

Structural Core vs. Domain Accent¶

This section pins down why dominated strategy is a domain-specific abstraction and not a prime, and it carries the case for that placement in the same stroke — isolating the thin contingency-proof-elimination test that lifts from the game-theoretic machinery that stays home.

What is skeletal (could lift toward a cross-domain prime). Strip the opponents away and a spare relational structure survives: discard an option that is worse than some alternative under every contingency, without forecasting which contingency obtains. The portable pieces are abstract — a chooser with a menu of options, a set of contingencies each option is scored against, a payoff comparison, and a belief-free verdict (this option is beaten in every column, so drop it with no probability model of the contingency). That skeleton is genuinely substrate-portable, which is exactly why it recurs in the catalog as the general decision-theoretic dominance pattern dominated strategy instantiates. This is the core the construct shares with statewise dominance in single-agent decision theory, not what makes it dominated strategy.

What is domain-bound. Almost everything that makes the construct dominated strategy in particular is game-theory furniture, and none of it survives extraction intact: the other strategic agents whose action profiles fill the contingencies; common knowledge of rationality, the lever that licenses deleting opponents' dominated actions too; iterated elimination of dominated strategies as a procedure that repeats and monotonically shrinks the game; the relationship to Nash equilibrium (every Nash equilibrium survives strict-dominance elimination, not conversely); dominant-strategy incentive compatibility (DSIC) and its mechanism-design role; and the Gibbard-Satterthwaite impossibility. These are the solution-concept apparatus the subfield actually runs, and each presupposes a strategy space with interdependent payoffs. The decisive test: remove the other strategic agents and their profiles — put a single agent facing states of the world instead — and the construct is no longer dominated strategy but plain statewise dominance, because there is no opponent to model, no common-knowledge lever, and no equilibrium apparatus to reduce. And a sharper boundary yet: absent any strategy space at all (a thermostat, a clearing market), the construct does not weaken into analogy — it simply has nothing to attach to, and the definition is vacuous.

Why this does not clear the prime bar. A prime's vocabulary travels and its transfer is recognition of the same mechanism, not analogy. Dominated strategy's transfer is bimodal, split by whether a strategy space with interdependent payoffs is present. Within game theory and its applied wings — solution analysis, auction theory, mechanism design, voting theory, bargaining — it travels intact as genuine recognition: it is the same test on the same formal machinery, deployed wherever agents, action sets, and joint-profile payoffs exist. Beyond that form the belief-free-elimination kernel still recurs as a real mechanism, but it appears under its own native apparatus — statewise dominance in single-agent decision theory under uncertainty — not as "dominated strategy," whose opponents, common-knowledge-of-rationality, and DSIC machinery have no purchase there. And when the bare structural lesson is wanted cross-domain — reject the under-every-contingency-worse option without forecasting the contingency — it is already carried, in more general form, by the parent the entry instantiates: decision-theoretic dominance. The cross-domain reach belongs to that parent; "dominated strategy," as named, carries the iterated-elimination, DSIC, and Gibbard-Satterthwaite machinery that should stay home. The one over-reading to guard against is colloquial: calling an option "dominated" when it is merely worse in expectation drops the for-every-contingency quantifier and smuggles a belief model back into the belief-free move.

Relationships to Other Abstractions¶

Current abstraction Dominated Strategy Domain-specific

Parents (1) — more general patterns this builds on

Dominated Strategy is a kind of Game-Theoretic Strategy Prime

A dominated strategy is a game-theoretic strategy for which another available strategy performs at least as well against every opponent profile.

Children (2) — more specific cases that build on this

Global Games Domain-specific is part of Dominated Strategy

Global Games contains iterated deletion of dominated strategies as the engine that moves inward from extreme-signal dominance regions.
Traveler's Dilemma Domain-specific is part of Dominated Strategy

Traveler's Dilemma contains a 98-step chain of weakly dominated claims whose iterative elimination produces its behaviorally fragile endpoint.

Hierarchy path (1) — routes to 1 parentless root

Dominated Strategy → Game-Theoretic Strategy → Function (Mapping)

Not to Be Confused With¶

Dominant strategy (the dual). The complementary construct: an action that is (weakly) best against every opponent profile, so a player should always play it. A dominated strategy is one that is (weakly) worse than some specific alternative against every profile, so a player should never play it. They are opposite ends of the same dominance relation, and a game can be riddled with dominated strategies to eliminate while granting no player an outright dominant one. Tell: is the action the one to keep no matter what (dominant) or the one to discard no matter what (dominated)?
Nash equilibrium. A best-response fixed-point: each player's strategy is optimal given what opponents actually play, a simultaneous condition on beliefs and strategies. Dominated-strategy elimination asks only whether an action is worse than some alternative regardless of opponents, needing no model of them. Every Nash equilibrium survives iterated elimination of strictly dominated strategies, but not conversely, so dominance is the weaker, prior pass. Tell: does the judgment require a model of opponents' actual play (Nash) or none at all (dominance)?
Decision-theoretic (statewise) dominance. The thin single-agent kernel: an act worse than some alternative under every state of the world is rejected regardless of the probability distribution over states — no opponents, no common-knowledge-of-rationality, no equilibrium apparatus. This is the substrate-neutral dominance parent that dominated strategy instantiates by filling the contingencies with opponents' action profiles. Tell: are the contingencies impersonal states of nature (decision-theoretic dominance, the parent) or the strategy choices of other rational agents (dominated strategy)? (Treated fully in earlier sections.)
Pareto dominance. A relation on joint outcomes for all players — one outcome Pareto-dominates another if nobody is worse off and someone is better off. It shares the word but ranks whole social outcomes, not one player's actions across opponents' profiles. The prisoner's dilemma separates them cleanly: Cooperate is the strategically dominated action, yet the surviving (Defect, Defect) is Pareto-dominated by mutual cooperation. Tell: is the comparison between two of a single player's strategies across opponents' choices (strategic dominance) or between two group outcomes ranked for everyone (Pareto dominance)?
Rationalizability (iterated-elimination survivors). The solution concept naming the strategies that survive iterated elimination of strictly dominated strategies — the residual set left after the deletion procedure, sometimes glossed via best-response reasoning under common knowledge of rationality. A dominated strategy is the per-action property that drives the elimination; rationalizability is the output set the procedure converges to. Tell: are you naming a single action that some alternative beats in every column (dominated strategy), or the whole set of actions still standing after all such actions are stripped away (rationalizability)?

Neighborhood in Abstraction Space¶

Dominated Strategy sits in a crowded region of the domain-specific corpus (0^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Strategic Interaction & Game Theory (23 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12