Game Theory Lecture 3

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

Game Theory

Lecture 3

Leong Kaiwen

Assistant Professor in Economics


Nanyang Technological University

2021

Leong Kaiwen (NTU) Game Theory NTU 2021 1/1


Iterative Deletion of Dominated

Looking at a game, figuring out which strategies are dominated, deleting


them. Looking at the game again, looking at which strategies are now
dominated, deleting those and so on and so forth.
Key tips: try to identify all the dominated strategies of all players
before you delete. Then delete. Then look again. Try to identify all
the dominated strategies of all players again, and then delete. Doing
it this way will prevent you from getting to the wrong answer.

Leong Kaiwen (NTU) Game Theory NTU 2021 2/1


Model of Politics

2 players: 2 candidates
Each candidate has to choose a political position for an election.
Strategy: choose a position on a political spectrum
We will assume that this political spectrum has 10 positions
- - - - - - - - - -
1 2 3 4 5 6 7 8 9 10
↗ ↖
1 represents extreme left wing 10 represents extreme right wing
position: oppose all social hierarchy position: poor people cannot vote

Leong Kaiwen (NTU) Game Theory NTU 2021 3/1


Model of Politics

We are going to assume that there are 10% of the voters at each
position. So at position 1 there are 10% of the voters, at position 2
there are 10% of the voters that support that position etc. In other
words, voters are uniformly distributed.
Voters vote for the closest candidate-candidate whose position is closest
to voter will get the voter’s vote.
Leong Kaiwen (NTU) Game Theory NTU 2021 4/1
Model of Politics

Tie-breaking assumption: if there is a tie, then the voters split. If there


is a tie, half the voters at that position go for one candidate and the
other half of the voters go for the other candidate.
Payoffs: candidates aim to maximize their share of the vote. A higher
share of the vote gives the candidate a mandate to govern.
Are any of the strategies dominated?
Claim: 2 dominates 1. This means that choosing position 2 always
gives me a higher share of the votes than choosing position 1, no
matter where the other candidate position herself.

Leong Kaiwen (NTU) Game Theory NTU 2021 5/1


Model of Politics

Let’s start working out what share of the votes you would get if you
choose position 1 or position 2 against different positions the other
candidate can choose.
What is the pattern ?
Position chosen by My payoff from My position from
the other candidate choosing 1 choosing 2
1 U1 (1,1)=50% < U1 (2,1)=90%
2 U1 (1,2)=10% < U1 (2,2)=50%
3 U1 (1,3)=15% < U1 (2,3)=20%
4 U1 (1,4)=20% < U1 (2,4)=25%
... ... ... ...
10 ... ... ...

Leong Kaiwen (NTU) Game Theory NTU 2021 6/1


Model of Politics

Pattern: ignoring scenarios where the opponent chooses position 1 or 2


(meaning opponent choosing 1 or 2), for all other position the opponent
chooses, if I chooses 2, I will always get 5% more as compared to
choosing 1 of the votes.
Conclude: 2 strictly dominates 1 for player 1. Similarly, 9 strictly
dominates 10 for player 1.
Is 2 dominated by 3? (Before we delete anything)

Leong Kaiwen (NTU) Game Theory NTU 2021 7/1


Model of Politics

Pattern: ignoring scenarios where the opponent chooses position 1 or 2


(meaning opponent choosing 1 or 2), for all other position the opponent
chooses, if I chooses 2, I will always get 5% more as compared to
choosing 1 of the votes.
Conclude: 2 strictly dominates 1 for player 1. Similarly, 9 strictly
dominates 10 for player 1.
Is 2 dominated by 3? (Before we delete anything)
No, if the other candidate chooses 1,U1 (2,1)=90% > U1 (3,1)=85%
⇒ 3 does not dominate 2 for player 1

Leong Kaiwen (NTU) Game Theory NTU 2021 7/1


Model of Politics

If we do the process of iterative deletion of dominated strategies and


we delete the dominated strategies then 3 dominates 2.
Key: we are only deleting strategies. We are not deleting the voters
who prefer those positions when we do iterative deletion of dominated
strategies. Voters are still there.

Leong Kaiwen (NTU) Game Theory NTU 2021 8/1


Model of Politics

If we delete the strategies 1 and 10 then does 3 dominates 2?


Position chosen by My payoff from My position from
the other candidate choosing 1 choosing 2
2 U1 (2,2)=50% < U1 (3,2)=80%
3 U1 (2,3)=20% < U1 (3,3)=50%
4 U1 (2,4)=25% < U1 (3,4)=30%
5 U1 (2,5)=30% < U1 (3,5)=35%
... ... ... ...
10 ... ... ...
From 4 to 10, choosing 3 yields 5% more votes for player 1 as compared
to choosing 2.
Conclusion: 2 and 9 are not dominated but they are dominated once
we realize 1 and 10 won’t be played.

Leong Kaiwen (NTU) Game Theory NTU 2021 9/1


Model of Politics

Where is this going?

× ×× ××× × × ×× ... ... × × ×× ××× ×× ×


1 2 3 4 5 6 7 8 9 10
× = dominated in round 1
×× = dominated in round 2
× × × = dominated in round 3
× × ×× = dominated in round 4
If we do the procedure of iterated dominated strategies, all that it left
is 5 and 6 (all else is deleted).

Leong Kaiwen (NTU) Game Theory NTU 2021 10 / 1


Model of Politics

This is a famous model in political science. The prediction of this


model is that the candidates are going to be squeezed to the middle.
They are going to choose positions close to the center and close to
each other. This is the median voter theorem. The votes at the center
5 and 6 actually decide not only the election but also what policies are
put in place.
How does this do as a prediction of the real world? eg., American
history and economics (product placement).

Leong Kaiwen (NTU) Game Theory NTU 2021 11 / 1


Model of Politics

Is this a good model of politics? What is missing?

Leong Kaiwen (NTU) Game Theory NTU 2021 12 / 1


Model of Politics

Is this a good model of politics? What is missing?


Voters are not evenly distributed. Not 10% on 1, 2, 3..., 10 respectively.
Many candidates (more than 2) can also choose not to vote.
Inability to commit to a position. Position is not believed by voters.
Primaries before general election.
Higher dimensions
It is tempting to say that we wrote this model but it is missing all these
items and so it is useless.
That is the wrong conclusion. The reason why we write these simple
models is to try and compare and test our intuitions. In this case,
crowding towards the center to get votes. The next step is to add
more into the model, see if the results change and try to understand
why.

Leong Kaiwen (NTU) Game Theory NTU 2021 12 / 1


Model of Politics

If we change the first reason of missing to a bell shaped curve for votes,
it does not change the result.
For the third reason of missing, eg., Tony Blair in the 90’s

Leong Kaiwen (NTU) Game Theory NTU 2021 13 / 1


Different Approach to Analysing Games

Best response
Player 2
l r
U (5,1) (0,2)
Player 1
M (1,3) (4,1)
D (4,2) (2,3)

Does either player have a dominated strategy?

Leong Kaiwen (NTU) Game Theory NTU 2021 14 / 1


Different Approach to Analysing Games

Is down dominated?
No, D does better than U when player 2 plays r. D does better than
M when player 2 plays l.
Similarly, r is not dominated because it is better than l when player 1
plays U. L is not dominated because it is better than r when player 1
plays M.
Nothing is dominated
If we are player 1, what would we do?

Leong Kaiwen (NTU) Game Theory NTU 2021 15 / 1


Different Approach to Analysing Games

If you knew player 2 was going to choose l, U is the best response to


l. U does best against l.
Similarly, M does best against r.
Is there a way of rationalize choosing D?

Leong Kaiwen (NTU) Game Theory NTU 2021 16 / 1


Different Approach to Analysing Games

Yes, if I believe 2 is equally likely to select l or r


Expected payoff of choosing U vs (1/2,1/2)=1/2*5+1/2*0=2.5
Expected payoff of choosing M vs (1/2,1/2)=1/2*1+1/2*4=2.5
Expected payoff of choosing D vs (1/2,1/2)=1/2*4+1/2*2=3
If I believe player 2 is equally likely to select l or r, my best response is
to choose D. It maximizes my expected payoff.

Leong Kaiwen (NTU) Game Theory NTU 2021 17 / 1


Different Approach to Analysing Games

I could also believe that player 2 is twice likely to choose l vs r. In that


case, I can recalculate the expected payoffs.
For example,
expected payoff of choosing U vs (2/3,1/3)=2/3*5+1/3*0=3.33
We can calculate the expected payoffs for any type of belief that I have.
The belief here refer to the probabilities that player 2 will choose l and
r respectively.

Leong Kaiwen (NTU) Game Theory NTU 2021 18 / 1


↑ ↑
Probability 0 means My belief that
2 will choose r player 2 will choose r
The orange line: EU1 (D,p(r))=(1-p(r))*4+p(r)*2
The blue line: EU1 (M,p(r))=(1-p(r))*1+p(r)*4
The gray line: EU1 (U,p(r))=(1-p(r))*5+p(r)*0

Leong Kaiwen (NTU) Game Theory NTU 2021 19 / 1


Different Approach to Analysing Games

If I think the probability that player 2 will choose r is less than X, then
the highest payoff I can get corresponds to the highest line EU1 (U,p(r)).
Hence, I will choose U. In other words, my best response is to choose
U.
If I think the probability that player 2 will choose r is greater than
Y, then the highest payoff that I can get corresponds to the line
EU1 (M,p(r)). Hence, I will choose M. My best response is M.
Finally, if x < p(r) < y, my best response is D.
Solve the simultaneous equations given by the 2 lines, X=0.33.

Leong Kaiwen (NTU) Game Theory NTU 2021 20 / 1

You might also like