Homework 5.4 Recall the following theorem from the overconfidence - - PowerPoint PPT Presentation

homework 5 4
SMART_READER_LITE
LIVE PREVIEW

Homework 5.4 Recall the following theorem from the overconfidence - - PowerPoint PPT Presentation

Homework 5.4 Recall the following theorem from the overconfidence versus paranoia lecture: Theorem . In a zero-sum perfect-information game, both the paranoid and overconfident models will lead you to choose the same strategy s 1 *


slide-1
SLIDE 1

Nau: Game Theory 1 Updated 11/2/10

–1 101 0 2

Homework 5.4

 Recall the following theorem from the “overconfidence versus

paranoia” lecture:

 Theorem. In a zero-sum perfect-information game, both the

paranoid and overconfident models will lead you to choose the same strategy s1*

 Professor Prune claims that the following game tree is a

counterexample to the theorem. Is he right? Why or why not? Can you suggest any alternative interpretations that might resolve the problem?

slide-2
SLIDE 2

Nau: Game Theory 2 Updated 11/2/10

Homework 6.1

In repeated games, which of the following strategies are stationary?

 AllC: always cooperate  AllD (the Hawk strategy): always defect  Grim: cooperate until the other agent defects, then defect forever  Tit-for-Tat (TFT): cooperate on the first move. On the nth move, repeat

the other agent (n–1)th move

 Random: Randomly intersperse cooperation and defection  Tester: defect on move 1. If the other agent retaliates, play TFT. Otherwise,

randomly intersperse cooperation and defection

 Tit-For-Two-Tats (TFTT): Like Tit-for-Tat, except that it retaliates only if the

  • ther agent defects twice in a row

 Generous Tit-For-Tat (GTFT): Like Tit-for-Tat, but with a small probability c>0

  • f cooperation on the n’th move if the other agent defected on the (n–1)th move

 Pavlov (Win-Stay, Lose-Shift): Repeats its previous move if it earns 3 or 5 points in

the previous iteration, and reverses its previous move if it earns 0 or 1 points in the previous iteration

slide-3
SLIDE 3

Nau: Game Theory 3 Updated 11/2/10

Homework 6.2

 Recall that DBS’s opponent model is as follows:

  • A set of rules of the following form

if our last move was m and their last move was m' then P[their next move will be C]

  • Four rules: one for each of (C,C), (C,D), (D,C), and (D,D)
  • For example, TFT can be described as
  • (C,C) ⇒ 1, (C, D) ⇒ 1, (D, C ) ⇒ 0, (D, D) ⇒ 0

(a) Which of the strategies in Homework 6.1 can be represented by DBS’s

  • pponent model?

(b) Are there any stationary strategies that DBS’s opponent model cannot represent? Why or why not? (b) Are there any stationary strategies that DBS’s opponent model cannot represent? Why or why not?

slide-4
SLIDE 4

Nau: Game Theory 4 Updated 11/2/10

Homework 6.3

 Let S be the following set of interaction traces for the iterated battle of the

sexes with 6 iterations. (a) What are the compatible subsets of S? (b) For each compatible subset, what is the composite trace and its expected utility if we assume agents a2, a4, a6, and a8 are equally likely?

G T G T G T T G T G T G G T T G G G T T T T T T Trace 1 Trace 2 a1 a2 a3 a4 T T G G T G T T G T G G T G G T G G T T T G G G Trace 3 Trace 4 a5 a6 a7 a8

slide-5
SLIDE 5

Nau: Game Theory 5 Updated 11/2/10

Homework 6.4

(a) What is the expectiminimax value of the following game tree?

MIN MAX

2

CHANCE

4 7 4 6 5 −2 2 4 −2 0.5 0.5 0.5 0.5 3 −1

2 4 6 8 10 –2 –4 –6 1/4 3/4 1/3 2/3

slide-6
SLIDE 6

Nau: Game Theory 6 Updated 11/2/10

Homework 6.5

(b) Do we need to know z in order to know the expectiminimax value of the game tree? Why or why not?

MIN MAX

2

CHANCE

4 7 4 6 5 −2 2 4 −2 0.5 0.5 0.5 0.5 3 −1

2 4 6 8 10 –2 –4 z 1/4 3/4 1/3 2/3

slide-7
SLIDE 7

Nau: Game Theory 7 Updated 11/2/10

Homework 6.5

(c) Do we need to know y and z in order to know the expectiminimax value of the game tree? Why or why not?

MIN MAX

2

CHANCE

4 7 4 6 5 −2 2 4 −2 0.5 0.5 0.5 0.5 3 −1

2 4 6 8 10 –2 y z 1/4 3/4 1/3 2/3

slide-8
SLIDE 8

Nau: Game Theory 8 Updated 11/2/10

Homework 6.5

(d) Suppose we know in advance that every terminal node has a value in the range [–10, +10]. For what values of x (if any) will y and z have no effect

  • n the expectiminimax value of the tree?

MIN MAX

2

CHANCE

4 7 4 6 5 −2 2 4 −2 0.5 0.5 0.5 0.5 3 −1

2 4 6 8 10 x y z 1/4 3/4 1/3 2/3

slide-9
SLIDE 9

Nau: Game Theory 9 Updated 11/2/10

Homework 6.6

 Consider the double lottery game with the imitate-the-better dynamic.

Suppose agent a’s strategy is R and agent b’s strategy is RwS. If we compare the two agents’ payoffs in order to decide which of them reproduces, then what are their probabilities of reproducing?