DOC PREVIEW
GT ISYE 6230 - ISYE 6230 Bayesian Games

This preview shows page 1-2-3-4-5 out of 15 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 15 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 15 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 15 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 15 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 15 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 15 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

11Recap{ Games in normal formz NBS vs. ABC, Prisoner’s dilemma, Tragedy of the commonsz Dominant or dominated actions, outcomesz Best response, Nash equilibrium (pure vs. mixed strategy){ Duopoly modelsz Cournot, Stackelberg, Bertrand; comparison with monopoly{ Multi-stage games with observed actionsz Stackelberg, Strategic investmentz Extensive form of a game, information setsz Subgame-perfect Nash equilibrium{ Repeated games (finitely or infinitely)z Prisoner’s dilemma, Cournot competitionz Trigger strategies, Friedman’s theorem{ Supply Chain Applications of Theory2Example: One-card poker{ A player is dealt an ace or king with equal probability23Example: One-card poker{ Two players{ One deck of cards, half aces, half kings{ Pay $a to play{ Each player is dealt a card face down{ After seeing his/her card, each player (simultaneously)z Action B: bets b, orz Action P: passes{ Payoffsz (B,P) or (P,B) → betting player gets the potz (B,B) or (P,P) → higher card gets the pot; in case of a tie, the pot is split4Record the resultsPlayer 1 Player 2 Payoffs---------------------------------------------------------------Game 1 Ace-Bet King-Bet (a+b, –(a+b))Game 2 King-Pass King-Bet (-a, a)Game 3 King-Pass King-Pass (0, 0)etc35Example: One-card pokerBetPassBP(K,K)(K,A)(A,K)(A,A)BetBetBetPassPassPassBBBBBBBPPPPPPP(0,0)(0,0)(0,0)(0,0)(a,-a)(a,-a)(a,-a)(a,-a)(a,-a)(-a,a)(-a,a)(-a,a)(-a,a)(-a,a)(a+b,-a-b)(-a-b,a+b)Player 1 OptionsPlayer 2 Options0.250.250.250.256Example: One-card poker (cont.){ The payoffs depend on players’actions andon card combinations{ Need to compute expected payoffs for each outcome47Example: One-card poker (cont.){ Payoffs for outcome (B,B) under different card combinations(0,0)0.25(K,K)(-a-b,a+b)0.25(K,A)(a+b, -a-b)0.25(A,K)(0,0)0.25(A,A)PayoffsProbabilityCard combination{ Expected payoff for player 1:(0.25)(0)+(0.25)(a+b)+(0.25)(-a-b)+(0.25)(0)=0{ Similarly, expected payoff for player 2 is 0{ Expected payoffs for (B,B) = (0,0)8One-card poker in normal form0, 0-a, aPassa, -a0, 0BetPassBetPlayer 1Player 2{ The unique NE is (B,B)59ISYE6230 one-card poker strategies10Cournot competition with incomplete information{ The market price, P is determined by (inverse) market demand: z P=a-Q if a>Q, P=0 otherwise. { Each firm decides on the quantity to sell (market share): q1and q2 {Q= q1+q2 total market demand{ Both firms seek to maximize profits{ The marginal cost of producing each unit of the good: c1and c2¾c1is common knowledge, however c2is known only by firm 2¾ Firm 1 believes that c2is “high” cHwith probability p and “low” cLwith probability (1-p) ¾ Firm 1’s belief about firm 2’s cost is common knowledge611Cournot Competition with Incomplete Information: Best response of Firm 2{ Suppose firm 1 produces q1{ Firm 2’s profits, if it produces q2are:π2= (P-c2)q2= [a-(q1+q2)]q2 – c2q2 = (Residual) revenue – Cost{ First order conditions:d π2/dq2= a - 2q2 – q1– c2= = RMR – MC = 0 →q2=(a-c2-q1)/2If Firm 2’s type is “high”: qH2=(a-cH-q1)/2If Firm 2’s type is “low”: qL2=(a-cL-q1)/212Cournot Competition with incomplete information: Best response of Firm 1{ Firm 1’s expected profits, if it produces q1are:π1= (P-c1)q1= p[a-(q1+qH2)]q1 +(1-p)[a-(q1+qL2)]q1 -c1q1 { First order conditions (FOC):d π1/dq1=p(a - 2q1 -qH2) – (1-p)( a - 2q1 -qL2)- c1= 0 →q1= p(a- c1 -qH2)/2 + (1-p)(a - c1-qL2)/2Cournot with complete information: q1=(a - c1– q2)/2713Cournot equilibrium under incomplete informationIf Firm 2’s type is “high”: qH2=(a-cH-q1)/2If Firm 2’s type is “low”: qL2=(a-cL-q1)/2q1= p(a- c1 -qH2)/2 + (1-p)(a - c1-qL2)/2Simultaneously solve to get:q1= (a - 2c1+ p cH+(1-p) cL)/3qH2= (a-2cH+c1)/3 + (1-p)(cH-cL)/6qL2= (a-2cL+c1)/3 - p(cH-cL)/614Comparison with the Cournot equilibrium under complete information{ Complete informationqC1=(a-2c1+c2)/3 qC2=(a-2c2+c1)/3 q’H= (a-2cH+c1)/3q’L= (a-2cL+c1)/3{ Incomplete informationq1= (a - 2c1+ p cH+(1-p) cL]/3qH2= (a-2cH+c1)/3 + (1-p)(cH-cL)/6qL2= (a-2cL+c1)/3 -p(cH-cL)/6qH2> q’Hand qL2< q’L.815Games of Incomplete Information (Bayesian Games){ In a game of complete information the players’ payoff functions are common knowledge{ In a game of incomplete information, at least one player is uncertain about another player’s payoff function16Normal Form Representation of Static Bayesian Games{ The normal form representation of an n-player static Bayesian game specifies the players’z action spaces A1,…,An, z type spaces T1,…,Tn, z beliefs p1,…,pn, andz payoff functions π1,…, πn.{ Player i’s type tiis privately known by player i, determines player i’s payoff function πi(a1,…,an; ti) and is a member of the set of possible types.{ Player i’s belief pi(t-i| ti) describes i’s uncertainty about the n-1 other players’ possible types t-i, given i’s own type ti. We denote this game by G=(A1,…,An; T1,…,Tn; p1,…,pn; π1,…, πn).917Example: Cournot Competition with incomplete information{ Actionsz Quantity choices q1 and q2 .{ Type spacesz Firm 1: T1={c1}z Firm 2: T2={cH, cL}{ Payoffsz Firm 1: π1(q1,qL2, qH2; c1) = p[a-(q1+qH2)]q1 +(1-p)[a-(q1+qL2)]q1 -c1q1.z Firm 2: π2(q1,qH2; cH) = (a- q1-qH2-cH) q2.π2(q1,qL2; cL) = (a- q1- qL2-cL) q2.{ Beliefsz Firm 1: p1(cH|c1) = p p1(cL|c1) = 1-pz Firm 2: p2(c1|t2) = 1 t2= cH, cL18Timing of a Static Bayesian Game{ Nature draws the type vector t=(t1,…,tn) according to probability distribution p(t){ Nature reveals tito player i but not to any other playerz Player i can compute his/her belief pi(t-i|ti) using Bayes’ rule (P(A|B)=P(A,B)/P(B)):pi(t-i| ti) = pi(t-i, ti) / pi(ti) = pi(t-i, ti) / ∑t-i∈T-ipi(t-i, ti) { The players simultaneously choose actions{ Payoffs are received1019Strategies in a Static Bayesian Game{ In the static Bayesian game G=(A1,…,An; T1,…,Tn; p1,…,pn; π1,…, πn)a strategyfor player i is a function si(ti), where for each type ti∈Ti, si(ti) specifies the action from the feasible set Aithat type i would choose if drawn by nature.{ Set of possible (pure) strategies for player i: Si: Ti→ Ai.{ In a separating strategy, each type ti∈Tichooses a different action from ai∈Ai.{ In a pooling strategy, for player i, each type ti∈Tichooses the same actions. 20Strategies in the Cournot GameFirm 1’s strategy: q1such thatq1= (a - 2c1+ p cH+(1-p) cL]/3Firm 2’s strategy: (q2(cH), q2(cL))q2(cH) = (a-2cH+c1)/3 + (1-p)(cH-cL)/6 = qH2q2(cL) =


View Full Document

GT ISYE 6230 - ISYE 6230 Bayesian Games

Documents in this Course
Recap

Recap

22 pages

Recap

Recap

11 pages

Load more
Download ISYE 6230 Bayesian Games
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view ISYE 6230 Bayesian Games and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view ISYE 6230 Bayesian Games 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?