Here's a game theory problem I've been struggling with since yesterday. It's been nagging at me, since I was pretty good in game theory but just can't understand the "correct" answer.
This is the classic Battle of the Sexes game with slightly modified payoffs:
This is the classic Battle of the Sexes game with slightly modified payoffs:
| Player 2 | |||
| Movies | Baseball | ||
| Player 1 | Movies | 2,3 | 0,0 |
| Baseball | 1,1 | 3,2 | |
(Apologies for the giant space before the table, dunno what's going on there)
Alright, so there are Nash Equilibria with pure strategies {M, M} and {B, B}. There's also a mixed strategy NE with {(1/4, 3/4), (3/4, 1/4)}, but this problem restricts itself to pure strategies.
Okay, so the situation is: Player 1 gets to choose their strategy first, and this is common knowledge (Player 2 knows that P1 moves first, and P1 knows that P2 knows, etc). The decision, however, is unknown.
The question: what is the Nash Equilibrium?
The (supposed) answer: B,B.
The explanation: Player 1 will choose the option that has the greatest payoff for him, so he will choose Baseball. Player 2, knowing this, will also choose Baseball.
My problem: Baseball does not have, in my opinion, the greatest payoff – that would make it a dominant strategy. It only has the greatest payoff if you assume P2 also chooses Baseball (more precisely if q < 3/4, where q = P(Movies)). With imperfect information and no communication, I can't solve this without bringing beliefs into the picture, which can't be right. Since P2 doesn't know P1's choice it shouldn't matter that P1 gets to move first – the game is exactly the same as if it were a simultaneous game (compare the decision trees for simultaneous and sequential versions of this game if in doubt). That means there are two possible NEs, neither of which can be said to be more likely than the other without further information.
The only relevant article I've been able to find on the subject says:
[E]xperiments strongly suggest that the player who moves first enjoys an advantage ... . These results have usually been ascribed to "positional order" or "presentation" effects that are seen as psychological or behavioral rather than fully rational responses to a change in the extensive form of the game.
Source: Beyond Normal Form Invariance: First Mover Advantage in Two-Stage Games with or without Predictable Cheap Talk – Peter J. Hammond, p.5-6
That article uses the classic form of the BotS game, with 2,1 / 1,2 the payoffs when collaboration occurs and 0,0 otherwise. I'm not sure that makes any difference, though.
It's driving me nuts. This is a problem for first-semesters, and they seem perfectly happy with Baseball, Baseball as the solution... but I'm not!
Any help would be appreciated.
[EDIT]:
Various textbooks and staff members confirm my view: the extensive form of the game (which is basically what the question describes; sequential moves with uncertainty about previous moves) is treated exactly the same as the normal form (with the grid), and in both cases neither Nash Equilibrium can be said to be more likely than the other based simply on the order in which the players move. Win!
----------------
Now playing: Tom Waits - Step Right Up

No comments:
Post a Comment