Normal Form Games Research Articles

Extreme punishments characterize weak Pareto optimality

In normal form games, we model the largely observed psychological phenomenon of systematic and extreme punishment after a deviation, regardless of the cost. After establishing basic properties, we show that this notion characterizes a weak form of Pareto optimality. Every Pareto optimal outcome can also be sustained by the threat of extreme punishment, which cannot be achieved in general through Nash equilibria strategies, nor with tit-for-tat strategies.

The Estimation of Network Formation Games with Positive Spillovers

I present a behavioural model of network formation with positive network externalities in which individuals have preferences for being part of a clique. The behavioural model leads to an associated supermodular (Topkis, 1979) normal form game. I show that the behavioural model converges to the greatest Nash equilibrium of the associated normal-form game. I propose an approximate Bayesian computation (ABC) framework, using original summary statistics, to make inferences about individuals' preferences, and provide an illustration using data on high school friendships.

Stability of the replicator dynamics for games in metric spaces

In this paper we study the stability of the replicator dynamics for symmetric games when the strategy set is a separable metric space. In this case the replicator dynamics evolves in a space of measures. We study stability criteria with respect to different topologies and metrics on the space of probability measures. This allows us to establish relations among Nash equilibria (of a certain normal form game) and the stability of the replicator dynamics in different metrics. Some examples illustrate our results.

Limits and limitations of no-regret learning in games

AbstractWe study the limit behavior and performance of no-regret dynamics in general game theoretic settings. We design protocols that achieve both good regret and equilibration guarantees in general games. We also establish a strong equivalence between them and coarse correlated equilibria (CCE). We examine structured game settings where stronger properties can be established for no-regret dynamics and CCE. In congestion games with non-atomic agents (each contributing a fraction of the flow), as we decrease the individual flow of agents, CCE become closely concentrated around the unique equilibrium flow of the non-atomic game. Moreover, we compare best/worst case no-regret learning behavior to best/worst case Nash equilibrium (NE) in small games. We prove analytical bounds on these inefficiency ratios for 2×2 games and unboundedness for larger games. Experimentally, we sample normal form games and compute their measures of inefficiency. We show that the ratio distribution has sharp decay, in the sense that most generated games have small ratios. They also exhibit strong anti-correlation between each other, that is games with large improvements from the best NE to the best CCE present small degradation from the worst NE to the worst CCE.

Empirical Distribution of Equilibrium Play and Its Testing Application

We show that in any n-player, m-action normal-form game, we can obtain an approximate equilibrium by sampling any mixed-action equilibrium a small number of times. We study three types of equilibria: Nash, correlated, and coarse-correlated. For each one we obtain upper and lower bounds on the number of samples required for the empirical distribution over the sampled action profiles to form an approximate equilibrium with probability close to 1.These bounds imply that using a small number of samples we can test whether or not players are playing according to an approximate equilibrium, even in games where n and m are large. In addition, our results substantially improve previously known upper bounds on the support size of approximate equilibria in games with many players. In particular, for the three types of equilibria we show the existence of approximate equilibrium with support-size polylogarithmic in n and m, whereas the previously best-known upper bounds were polynomial in n (Hémon et al. [Hémon S, de Rougemont M, Santha M (2008) Approximate nash equilibria for multi-player games. Algorithmic Game Theory (Springer), 267–278], Germano and Lugosi [Germano F, Lugosi G (2007) Existence of sparsely supported correlated equilibria. Econom. Theory 32(3):575–578], Hart et al. [Hart S, Mas-Colell A, Babichenko Y (2013) Simple Adaptive Strategies: From Regret-Matching to Uncoupled Dynamics, Vol. 4 (World Scientific Publishing Company)]).

Efficiently detecting switches against non-stationary opponents

Interactions in multiagent systems are generally more complicated than single agent ones. Game theory provides solutions on how to act in multiagent scenarios; however, it assumes that all agents will act rationally. Moreover, some works also assume the opponent will use a stationary strategy. These assumptions usually do not hold in real world scenarios where agents have limited capacities and may deviate from a perfect rational response. Our goal is still to act optimally in these cases by learning the appropriate response and without any prior policies on how to act. Thus, we focus on the problem when another agent in the environment uses different stationary strategies over time. This will turn the problem into learning in a non-stationary environment, posing a problem for most learning algorithms. This paper introduces DriftER, an algorithm that (1) learns a model of the opponent, (2) uses that to obtain an optimal policy and then (3) determines when it must re-learn due to an opponent strategy change. We provide theoretical results showing that DriftER guarantees to detect switches with high probability. Also, we provide empirical results showing that our approach outperforms state of the art algorithms, in normal form games such as prisoner's dilemma and then in a more realistic scenario, the Power TAC simulator.

Epistemically Robust Strategy Subsets

We define a concept of epistemic robustness in the context of an epistemic model of a finite normal-form game where a player type corresponds to a belief over the profiles of opponent strategies and types. A Cartesian product X of pure-strategy subsets is epistemically robust if there is a Cartesian product Y of player type subsets with X as the associated set of best reply profiles such that the set Y i contains all player types that believe with sufficient probability that the others are of types in Y − i and play best replies. This robustness concept provides epistemic foundations for set-valued generalizations of strict Nash equilibrium, applicable also to games without strict Nash equilibria. We relate our concept to closedness under rational behavior and thus to strategic stability and to the best reply property and thus to rationalizability.

Computing Dominance-Based Solution Concepts

Two common criticisms of Nash equilibrium are its dependence on very demanding epistemic assumptions and its computational intractability. We study the computational properties of less demanding set-valued solution concepts that are based on varying notions of dominance. These concepts are intuitively appealing, always exist, and admit unique minimal solutions in important subclasses of games. Examples include Shapley’s saddles, Harsanyi and Selten’s primitive formations, Basu and Weibull’s CURB sets, and Dutta and Laslier’s minimal covering set. Based on a unifying framework proposed by Duggan and Le Breton, we formulate two generic algorithms for computing these concepts and investigate for which classes of games and which properties of the underlying dominance notion the algorithms are sound and efficient. We identify two sets of conditions that are sufficient for polynomial-time computability and show that the conditions are satisfied, for instance, by saddles and primitive formations in normal-form games, minimal CURB sets in two-player games, and the minimal covering set in symmetric matrix games. Our positive algorithmic results explain regularities observed in the literature, but also apply to several solution concepts whose computational complexity was previously unknown.

When does restricting your opponent's freedom hurt you?

I examine the payoff consequences for a player when she removes a subset of her opponent's actions before playing a two-player complete information normal form game. When she faces a constraint on the maximal number of actions she can remove, she can be strictly better off by not removing any actions. I present such an example. I also establish sufficient conditions under which removing opponent's actions cannot hurt. As a corollary, I also characterize a necessary condition for a player's optimal Nash Equilibrium in games with generic payoffs when her opponent has strictly more actions than she does.

Equilibrium Notions for Agents with Cumulative Prospect Theory Preferences

Relying on the experimental findings that actual choice behavior often violates the axioms of expected utility theory (EUT), we study non-EUT preferences in a noncooperative game-theoretic framework. In particular, agents’ preferences are represented by the pair of functions suggested in cumulative prospect theory (CPT). Accordingly, three key aspects of CPT are incorporated: subjective probability weighting, loss aversion, and reference dependence. We introduce a correlated equilibrium and two mixed strategy equilibria for agents with CPT preferences. We prove the existence of equilibria for finite normal form games and investigate the sets of equilibria in some examples.

Stochastic Stability in the Large Population and Small Mutation Limits for General Normal Form Games

We consider models of stochastic evolution in normal form games with K>=2 strategies in which agents employ the best response with mutations choice rule. Forming the dynamics as a Markov chain with state space being the set of best responses, we overcome technical difficulties which arises with the large population. We study the long run behavior of the Markov chain for four limiting cases: the small noise limit, the large population limit, and the both orders of the double limit. We characterize conditions under which selection results of the both orders of the double limit coincide.

On the convergence of logit-response to (strict) Nash equilibria

We show that the logit-response dynamics converges to a subset of (strict) Nash equilibria for any weakly acyclic, N-player normal form game. The result holds independently of whether revision opportunities arise as in asynchronous learning or in other ways, as long as the dynamics is regular. Our analysis generalizes and organizes recent convergence results in the literature. Further, it provides a simple selection criterion, because convergence to strict Nash equilibria obtains even in the presence of non-strict, pure-strategy Nash equilibria.

Direct Evolutionary Search for Nash Equilibria Detection

<p>A Direct method of computing mixed form Nash equilibria of a normal form game by using a simple evolutionary algorithm is proposed. The Direct Evolutionary Search algorithm (DES) uses a generative relation for Nash equilibria with binary tournament selection and uniform mutation. Numerical experiments are used to illustrate the efficiency of the method.</p>

Translucent Players: Explaining Cooperative Behavior in Social Dilemmas

In the last few decades, numerous experiments have shown that humans do not always behave so as to maximize their material payoff. Cooperative behavior when non-cooperation is a dominant strategy (with respect to the material payoffs) is particularly puzzling. Here we propose a novel approach to explain cooperation, assuming what Halpern and Pass call translucent players. Typically, players are assumed to be opaque, in the sense that a deviation by one player in a normal-form game does not affect the strategies used by other players. But a player may believe that if he switches from one strategy to another, the fact that he chooses to switch may be visible to the other players. For example, if he chooses to defect in Prisoner's Dilemma, the other player may sense his guilt. We show that by assuming translucent players, we can recover many of the regularities observed in human behavior in well-studied games such as Prisoner's Dilemma, Traveler's Dilemma, Bertrand Competition, and the Public Goods game.

When Can Limited Randomness Be Used in Repeated Games?

The central result of classical game theory states that every finite normal form game has a Nash equilibrium, provided that players are allowed to use randomized (mixed) strategies. However, in practice, humans are known to be bad at generating sequences, and true random bits may be unavailable. Even if the players have access to enough random bits for a single instance of the game their randomness might be insufficient if the game is played many times. In this work, we ask whether randomness is necessary for equilibria to exist in finitely repeated games. We show that for a large class of games containing arbitrary two-player zero-sum games, approximate Nash equilibria of the n-stage repeated version of the game exist if and only if both players have Ω(n) random bits. In contrast, we show that there exists a class of games for which no equilibrium exists in pure strategies, yet the n-stage repeated version of the game has an exact Nash equilibrium in which each player uses only a constant number of random bits. When the players are assumed to be computationally bounded, if cryptographic pseudorandom generators (or, equivalently, one-way functions) exist, then the players can base their strategies on random-like sequences derived from only a small number of truly random bits. We show that, in contrast, in repeated two-player zero-sum games, if pseudorandom generators do not exist, then Ω(n) random bits remain necessary for equilibria to exist.

How strategic are children and adolescents? Experimental evidence from normal-form games

We examine the strategic sophistication of 196 children and adolescents, aged 10–17 years, in experimental normal-form games. Besides choices, we also elicit first- and second-order beliefs. The share of subjects playing Nash or expecting opponents to play Nash is fairly stable across all age groups. The likelihood of playing best response to own beliefs increases in math skills. Using a mixture model, about 40% of subjects are classified as a strategic type, while the others are non-strategic. The distribution of types is somewhat changing with age. The estimated error rates also show some dependency on age and gender.

Quantal response methods for equilibrium selection in normal form games

This paper describes a general framework for equilibrium selection by tracing the graph of the quantal response equilibrium (QRE) correspondence as a function of the variance of random disturbances. If a quantal response function satisfies C2 continuity, monotonicity and cumulativity, the graph of QRE correspondence generically includes a unique branch that starts at the centroid of the strategy simplex and converges to a unique Nash equilibrium as noises vanish. This equilibrium is called the limiting QRE of the game. We then investigate the limiting QRE in normal form games, and analyze the effects of payoff transformations and adding/eliminating dominated strategies on equilibrium selection. We find that in two-person symmetric games, any strict Nash equilibrium can be selected as the limiting QRE by appropriately adding a single strictly dominated strategy.

MIXED EQUILIBRIUM IN 2 × 2 NORMAL FORM GAMES: WHEN BURNING MONEY IS RATIONAL

We discuss the rationality of burning money behavior from a new perspective: the mixed Nash equilibrium in 2 × 2 normal form games. We support our argument analyzing the first-order derivatives of the players' mixed equilibrium expected utility payoffs with respect to their own utility payoffs. We establish necessary and sufficient conditions that guarantee the existence of negative derivatives. In particular, games with negative derivatives are the ones that create incentives for burning money behavior since such behavior in these games improves the player's mixed equilibrium expected utility payoff. We show that a negative derivative of some player i 's mixed equilibrium expected utility payoff occurs if, and only if, he has a strict preference for one of the strategies of the other player. Moreover, negative derivatives always occur when they are taken with respect to player i 's highest and lowest game utility payoffs. We also present sufficient conditions that ensure that such derivatives are always non-negative in finite normal form games.

Abstraction Using Analysis of Subgames

Normal form games are one of the most familiar representations for modeling interactions among multiple agent. However, modeling many realistic interactions between agents results in games that are extremely large. In these cases computing standard solutions like Nash equilibrium may be intractable. To overcome this issue the idea of abstraction has been investigated, most prominently in research on computer Poker. Solving a game using abstraction requires using some method to simplify the game before it is analyzed. We study a new variation for solving normal form games using abstraction that is based on finding and solving suitable sub games. We compare this method with several variations of a common type of abstraction based on clustering similar strategies.

Strongly rational sets for normal-form games

Curb sets [Basu and Weibull, Econ. Letters 36 (1991), 141-146] are product sets of pure strategies containing all individual best-responses against beliefs restricted to the recommendations to the remaining players. The concept of minimal curb sets is a set-theoretic coarsening of the notion of strict Nash equilibrium. We introduce the concept of minimal strong curb sets which is a set-theoretic coarsening of the notion of strong Nash equilibrium. Strong curb sets are product sets of pure strategies such that each player’s set of recommended strategies must contain all coalitional bestresponses of each coalition to whatever belief each coalition member may have that is consistent with the recommendations to the other players. Minimal strong curb sets are shown to exist and are compared with other well known solution concepts. We also provide a dynamic learning process leading the players to playing strategies from a minimal strong curb set. JEL classi…cation: C72