L4 Dominant Strategy

Dominant, Fixed &
Mixed Strategies
Bernt A. Bremdal
UiT Campus Narvik
2022
Repeated games
• A stage game is played multiple times
• What access is there to information about a concluded game?
• Is the information sharing assymmetric?
• Can we apply a memory to learn our opponents strategy?
• Mixed strategy?
• Stationary strategy?
• Monte Carlo strategy?
Dominant strategy
The dominant strategy in game theory refers to a
situation where one player has a superior tactic
regardless of how the other players act.
Example
• A company needs to choose between two strategies
• online or offline advertising
• they are equal in all aspects except for their expected payoff.
• Example: If online advertising will lead to a payoff of

$20,000, whereas offline advertising will lead to a
payoff of either $15,000 or $10,000, depending on
where their competitors advertise, then online
advertising is the dominant strategy, because it
will lead to a higher payoff.
Dominance
Strictly Dominant Outcome
• In some situations, one player enjoys a strict advantage over
their opponent.
• This means that, no matter how good the losing party’s tactic
is, the dominant strategy will always prevail.
• Here, there is no other possible strategy the opponent can use
to alter their odds.
Weakly Dominant Outcome

• In a weakly dominant outcome, the dominant player dominates
the game but against some strategies, only weakly dominates.
Dominant strategy
• A player applies a dominant strategy if the

strategy maximizes the player’s payoff
independently of other players.
• In other words: Regardless of what the other
player does a dominant strategy will always come
on top.
• Dominant strategies make games easy to solve:
«Nobrainers»
Firm 1 plays H  Firm 2 will play L

Firm 1 plays L  Firm 2 will still play L
=> Strategy L is dominant for Firm 2
More formally
• Given any particular strategy (either C or D) of agent i, there will be a

number of possible outcomes
• We say s1 dominates s2 if every outcome possible by i playing s1 is preferred
over every outcome possible by i playing s2
• A strategy profile (s1…….sn) in which every si is dominant for player i is a Nash equilibrium –
“equilibrium in dominant strategies”
Dominant strategies tend
to make things trivial
In politics it is a small country’s reality
In war it is a small country’s demise
Successive elimination of dominating
strategies
Fixed strategy vs mixed
strategy
When playing the same game more than once
The essence of poker!

Example 1
• At the roulette table:
• Three player can choose between bets on red or black on the table
• They make a series of bets
• Player 1 always plays red
• Player 2 assesses her chances and combines strikes of red and black
• Player 3 tosses a coin before each game and decides the colour based on the
outcome of the toss
• Player 1 plays a fixed (stationary) strategy

• Player 2 and 3 play a mixed strategy
• Player 3 applies a Monte Carlo strategy for his mixed approach
Implications for AI – non-supervised machine
learning
• Learning of all kinds imply both exploration and exploitation
• Exploitation is about taking advantage of existing insight
• Exploration discovering new possibilities i.e. research
• Fixed strategy is a kind of exploitation
• Random strategy is a kind of exploration
• Mixed strategy could be a mix between a random exploration and a
prevailing exploitation
Example 2
• A seller and a buyer meet often to strike a deal
• The seller starts asking a price of p1
• The buyer demands a price of p2
• After a while the buyer notices a pattern:

• p1,t ϵ {10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10}
• Obviously the seller is maintaining a fixed strategy for the opening price
• How can this insight help the buyer?
Mixed strategy
• Let (N,A,u) be a normal-form game, and for any set X let ∏ (X) be the set of all
probability distributions over X. Then the set of mixed strategies for player Si= ∏
(Ai)
• The set of mixed-strategy profiles is the Cartesian product of the individual

mixed-strategy sets, S1*…*Sn
• Si(ai) is the probability of an action ai under a mixed strategy
• The support of a mixed strategy si for player i is the set of all pure strategies
{ai | si(ai) < 0}
6-14
Fictitious play – a way to learn the opponent’s
strategy
•
• P(a) is the probability that the opponent plays action a

• a is the set of all the actions of the opponent
• w(a) is the frequency of action a during a set of repeated games
Example: Preparing for the soccer
finals
• You are a coach and want to prepare for a decisive shoot-out
• You study videos of the other team and their keeper
• In game after game, he tally marks what the keeper usually does
when there is a penalty kick: Does he go to the left or to the right
• This is what he found out: {L,L,L,R,L,L,R,R,L,L,R,L,L,R,R,L,L,R,L,L}
• What is the probability that the keeper goes to the left?
• Fictitious play: P(L)=13/20, P(R)=1-P(L)= 7/10
• Which side is the keeper inclined to throw himself?
• What would you advise your payers to do?
Battle of the sexes: Playing a mixed
strategy
• Assume that the wife and husband both understand that both needs
to yield over time, meaning that you cannot always insist on your
favourite type of movie, but go along with the preference of the
partner at time to time.
• Hence a mixed strategy is a good approach over time.
• But how do they strike a balance between Love movies and War
movies?
The concept of indifference
• The concept of indifference implies that a mixed strategy creates a
balance between the expected utility (payoff) of the outcomes for
the two players (e.g. in the husband and wife example) over time
Expected Utility Theorem
(von Neumann and Morgenstern 1947)
• Let X be the set of all probabilities on a finite set X.

• The preference relation % on X is complete, transitive, independent
and Archimedean if and only if there exists a function v : X R such
that:
where pk is the probability that outcome indexed by k with payoff xk is realized,

and function u expresses the utility of each respective payoff.
Example: Battle of the sexes game
Husband
LW WL
LW 2,1 0,0
Wife
WL 0,0 1,2
Example: Husband plays LW with probability p and WL with probability 1-p.

Wife also mixes her two actions. She must be indifferent to them given her
husband’s strategy.
Uwife(LW) = Uwife(WL)
2*p+0*(1-p) = 0*p + 1*(1-p)
p = 1/3
Now the wife is indifferent with regard to the decisions of her husband
6-20
Take aways
• Repeated games are a series of one-stage games
• A dominant strategy means that it is indifferent to whatever the
opponent(s) do(es).
• When a one-stage game is played several times a player may adopt a
fixed or a mixed strategy
• Both a fixed and mixed strategy may be learned by the opponent
• A mixed strategy implies the possibility of striking a balance between
different interests using an expected utility approach
• The expected utility theorem was defined by von Neumann and
Morgenstern

L4 Dominant Strategy

Uploaded by

Copyright:

Available Formats

L4 Dominant Strategy

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

L4 Dominant Strategy

Uploaded by

Copyright:

Available Formats

Dominant, Fixed &

• Example: If online advertising will lead to a payoff of

Weakly Dominant Outcome

• A player applies a dominant strategy if the

Firm 1 plays H  Firm 2 will play L

• Given any particular strategy (either C or D) of agent i, there will be a

The essence of poker!

• Player 1 plays a fixed (stationary) strategy

• After a while the buyer notices a pattern:

• The set of mixed-strategy profiles is the Cartesian product of the individual

• P(a) is the probability that the opponent plays action a

• Let X be the set of all probabilities on a finite set X.

where pk is the probability that outcome indexed by k with payoff xk is realized,

Example: Husband plays LW with probability p and WL with probability 1-p.

You might also like