Mixed Strategies

Lecture 11
Feb 10, 2025
Building blocks

So far, we only discussed on pure strategy Nash equilibrium.

There are options where a player may choose to randomize
between several of his pure strategies.
You may think why anyone would wish to randomize between
actions.
There are many games for which there will be no equilibrium
predictions if we do not consider the players’ ability to choose
stochastic strategies.
For example, Matching Pennis game or Rock-Paper-Scissor
game.
 chapters we restricted players to using pure strategies and we
ssing the option that a player may choose to randomize between
Matching
strategies. Pennis
You may wonder why anyone would wish to randomize
his turns out to be an important type of behavior to consider, with
tions and interpretations. In fact, as we will now see, there are many
here will be no equilibrium predictions if we do not consider the
Consider
choose stochastic the following zero-sum game called Matching
strategies.
llowing classic zero-sum
Pennies. game called
Players 1 andMatching
2 each Pennies. 1 Players
put a penny on a table
penny on a table simultaneously.
simultaneously. If thetwo
If the twopennies
pennies come
come upupthe
the same side
or tails) then player 1 gets both; otherwise player 2 does. We can
(heads or tails) then player 1 gets both; otherwise player 2
e following matrix:
does.
Player 2
H T
H 1, −1 −1, 1
Player 1
T −1, 1 1, −1

cludes the best-response choices of

No pure strategy each player using the method
equilibrium.
ection 5.1.1 to find pure-strategy Nash equilibria. As you can see,
ot work: Given a belief that player 1 has about player 2’s choice,
match it. In contrast, given a belief that player 2 has about player
 Rock-Paper-Scissor
d Strategies

In Rock-Paper-Scissors game, rock beats scissors, scissors beats paper,

ors beats paper, and paper beats rock. If winning gives the player a
and paper beats rock. If winning gives the player a payoff of +1 and the
d the loser a payoff
loser a of −1, and
payoff if we
of −1, andassume that a tie
if we assume thatis aworth
tie is 0, then0, then we have
worth
e this game by the following matrix:
the following game matrix:

Player 2
R P S
R 0, 0 −1, 1 1, −1

Player 1 P 1, −1 0, 0 −1, 1

S −1, 1 1, −1 0, 0

ghtforward to write down the best-response correspondence for player

P
eves that player 2 will play one of his pure
 when s2 = R
strategies as follows:
s1 (s2 ) = S when s2 = P

R when s2 = S

⎧
⎨P when s2 = R
Mixed Strategy

{Definition} Let Si = {si1 , si2 , ..., sim } be player i’s finite set of pure
strategies. Define ∆Si as the simplex of Si , which is the set of all
probability distributions over Si . A mixed strategy for player i is an
element σi ∈ ∆Si , so that σi = {σi (si1 ), σi (si2 ), ..., σi (sim )} is a probability
distribution over Si , where σi (si ) is the probability that player i plays si .
A mixed strategy for player i is just a probability distribution over his pure
strategies. Since any probability distribution σi (.) over a finite set of
elements (a finite state space), in our case Si , must satisfy two conditions:

Pi (si ) ≥ 0 for all si ∈ Si , and

σ
1

si ∈Si σi (si ) = 1.
2

That is, the probability of any event happening must be nonnegative, and
the sum of the probabilities of all the possible events must add up to one.
Notice that every pure strategy is a mixed strategy with a degenerate
distribution that picks a single pure strategy with probability one and all
other pure strategies with probability zero.
Example?
Example - Mixed Strategy

In the Matching Pennies game, for each player i, Si = {H, T }, the

simplex which is the set of mixed strategies, can be defined as follows:

∆Si = {(σi (H), σi (T )) : σi (H) ≥ 0, σi (T ) ≥ 0, σi (H) + σi (T ) = 1}.

We read this as follows: the set of mixed strategies is the set of all pairs
(σi (H), σi (T )) such that both are nonnegative numbers, and they both
sum to one. We use the notation σi (H) to represent the probability that
player i plays H and σi (T ) to represent the probability that player i plays
T.
Example - Mixed Strategy

Now consider the example of the rock-paper-scissors game, in which

Si = {R, P, S} (for rock, paper, and scissors, respectively). We can
define the simplex as

∆Si = {(σi (R), σi (P), σi (S)) : σi (R), σi (P), σi (S) ≥ 0, σi (R)+σi (P)+σi (S) = 1}

There are now three numbers, each defining the probability that the
player plays one of his pure strategies.
A pure strategy is just a special case of a mixed strategy. For example, in
this game we can represent the pure strategy of playing R with the
degenerate mixed strategy: σ(R) = 1, σ(P) = σ(S) = 0.
Continuous Strategy Sets

{ Definition } Given a mixed strategy σi (.) for player i, we will say that a
pure strategy si ∈ Si is in the support of σi (.) if and only if it occurs with
positive probability, that is, σi (si ) > 0.
For example, in the game of rock-paper-scissors, a player can choose rock
or paper, each with equal probability, and not choose scissors. In this case
σi (R) = σi (P) = 0.5 and σi (S) = 0.
We will then say that R and P are in the support of σi (.), but S is not.
{ Definition } Let Si be player i’s pure-strategy set and assume that Si is
an interval. A mixed strategy for player i is a cumulative distribution
function Fi : Si → [0, 1], where Fi (x) = Pr {si ≤ x}. If Fi (.) is
differentiable with density fi (.) then we say that si ∈ Si is in the support
of Fi (.) if fi (si ) > 0.
Example?
Continuous Strategy Sets

Consider the Cournot duopoly game with a capacity constraint

of 100 units of production, so that Si = [0, 100] for i ∈ {1, 2}.
Consider the mixed strategy in which player i chooses a
quantity between 30 and 50 using a uniform distribution.

0
 for si < 30;
si −30
Fi (si ) = 20 for si ∈ [30, 50];

1 for si > 50.


0 for si < 30;

1
fi (si ) = 20 for si ∈ [30, 50];

0 for si > 50.

 A continuous mixed strategy in the Cournot game.
.
Chapter 6 Mixed Strategies

F(si) f(si)
1

1
—
20

si si
30 50 100 30 50 100

FIGURE 6.1 A continuous mixed strategy in the Cournot game.

mixed strategy σi (.), it will be useful to distinguish between pure strategies that are
chosen with a positive probability and those that are not. We offer the following
Beliefs and Mixed Strategies

{Definition:} A belief for player i is given by a probability distribution

πi ∈ ∆Si over the strategies of his opponents. We denote by πi (s−i ) the
probability player i assigns to his opponents playing s−i ∈ S−i .
A belief for player i is a probability distribution over the strategies of his
opponents. Notice that the belief of player i lies in the same set that
represents the profiles of mixed strategies of player i’s opponents.
For example, in the rock-paper-scissors game, we can represent the beliefs
of player 1 as a triplet, (π1 (R), π1 (P), π1 (S)), where by definition
π1 (R), π1 (P), π1 (S) ≥ 0 and π1 (R) + π1 (P) + π1 (S) = 1.
The interpretation of π1 (s2 ) is the probability that player 1 assigns to
player 2 playing some particular s2 ∈ S2 . Recall that the strategy of player
2 is a triplet σ2 (R), σ2 (P), σ2 (S) ≥ 0, with σ2 (R) + σ2 (P) + σ2 (S) = 1,
so we can clearly see the analogy between π and σ.
Expected Payoffs

{Definition:} The expected payoff of player i when he chooses the pure

strategy si ∈ Si and his opponents play the mixed strategy σ−i ∈ S−i is
X
vi (si , σ−i ) = σ−i (s−i )vi (si , s−i ).
s−i ∈S−i

Similarly the expected payoff of player i when he chooses the mixed

strategy σi ∈ Si and his opponents play the mixed strategy σ−i ∈ S−i is

 
X X X
vi (σi , σ−i ) = σi (si )vi (si , σ−i ) =  σi (si )σ−i (s−i )vi (si , s−i ) .
si ∈Si si ∈Si s−i ∈S−i

Example?
ors beats paper, and paper beats rock. If winning gives the player a
d the loser a payoff of −1, and if we assume that a tie is worth 0, then
Expected Payoffs: Example
be this game by the following matrix:

Player 2
R P S
R 0, 0 −1, 1 1, −1

Player 1 P 1, −1 0, 0 −1, 1

S −1, 1 1, −1 0, 0

ightforward toInwrite
thedown
R-P-S thegame,
best-response
assume correspondence for player
that player 2 plays σ2 (R) = σ2 (P) = 1/2;
ieves that player 2 will play one of his pure strategies as follows:
σ2 (S) = 0. We can now calculate the expected payoff for player 1 from
any of his pure strategies,
⎧
σ2 ) s2==1/2
⎨ vP1 (R,when R × 0 + 1/2 × (−1) + 0 × 1 = −1/2
s1(s2) = vS1(P,when s =
σ2 ) 2 = P1/2 × 1 + 1/2 × 0 + 0 × (−1) = 1/2
⎩
v 1(S, σ2 ) s2 ==S,1/2 × (−1) + 1/2 × 1 + 0 × 0 = 0.
R when

(symmetric) list
It would
is easybetotheseebest-response
that player 1correspondence of player
has a unique best response to this mixed
the two best-response
strategycorrespondences immediately
of player 2 (playing P with implies that there
probability 1). Clearly if his beliefs
ategy equilibrium,
aboutjustthe
likestrategy
in the Matching Pennies game.
of his opponent The reason
are different then player 1 is likely to
g with any pairhave
of pure strategies,
a different bestat response.
least one player is not playing a
Expected Payoffs: all-pay auction
All-pay auction: here players have strategy sets that are in intervals.
Ttwo players can bid for a dollar. Each can submit a bid that is a real
number (we are not restricted to penny increments), so that Si = [0, ∞),
i ∈ {1, 2}. The person with the higher bid gets the dollar, but both
bidders have to pay their bids. If there is a tie then both pay and the
dollar is awarded to each player with an equal probability of 0.5. Thus if
player i bids si and player j ̸= i bids sj then player i’s payoff is

−si
 if si < sj ;
vi (si , s−i ) = 1/2 − si if si = sj ;

1 − si if si > sj .


Imagine that player 2 is playing a mixed strategy in which he is uniformly

choosing a bid between 0 and 1. That is, player 2’s mixed strategy σ2 is a
uniform distribution over the interval 0 and 1, which is represented by the
cumulative distribution function
(
s2 for s2 ∈ [0, 1];
F2 (s2 ) =
1 for s2 > 1.
Expected Payoffs: all-pay auction

The expected payoff of player 1 from offering a bid si > 1 is (1 − si ) < 0

because he will win for sure, but this would not be wise.
The expected payoff from bidding si < 1 is

v1 (s1 , σ2 ) = Pr {s1 < s2 }(−s1) + Pr {s1 = s2 }(1/2 − s1 ) + Pr {s1 > s2 }(1 − s1 )

= (1 − F2 (s1 ))(−s1 ) + 0(1/2 − s1 ) + F2 (s1 )(1 − s1 )
= 0.

Thus when player 2 is using a uniform distribution between 0 and 1 for

his bid, then player 1 cannot get any positive expected payoff from any
bid he offers: any bid less than 1 offers an expected payoff of 0, and any
bid above 1 guarantees getting the dollar at an inflated price.
Expected Payoffs: continuous strategy space

Consider a game in which each player has a strategy set given

by the interval Si = [si , s̄i ] . If player 1 is playing s1 and his
opponents, players j = 2, 3, ..., n, are using the mixed
strategies given by the density function fj (sj ) then the
expected payoff of player 1 is given by
Z s̄2 Z s̄3 Z s̄n
... vi (si , s−i )f2 (s2 )f3 (s3 )...fn (sn )ds2 ds3 ...dsn .
s2 s3 sn
Mixed-Strategy Nash Equilibrium

{ Definition } The mixed-strategy profile σ ∗ = (σ1∗ , σ2∗ , ..., σn∗ )

is a Nash equilibrium if for each player i, σi∗ is a best response
∗ . That is, for all i ∈ N,
to σ−i
∗
vi (σi∗ , σ−i ∗
) ≥ vi (σi , σ−i ); ∀σi ∈ Si .
 llowing classic zero-sum game called Matching Pennies.1 Players
penny on a table simultaneously. If the two pennies come up the
or tails) then player 1 gets both; otherwise player 2 does. We can
Example:
e following matrix: Matching Pennies

Player 2
H T
H 1, −1 −1, 1
Player 1
T −1, 1 1, −1

cludes the best-response

Let p be the choices of each that
probability playerplayer
using the method
1 plays H and (1 − p) the
ection 5.1.1 toprobability that he plays T . Similarly let qsee,
find pure-strategy Nash equilibria. As you can be the probability
ot work: Given a belief that player 1 has about player 2’s choice,
that player 2 plays H and (1 − q) the probability that he plays
match it. In contrast, given a belief that player 2 has about player
T.
ld like to choose the opposite orientation for his penny. Does this
Using the We
quilibrium fails to exist? formulas forsee
will soon expected payoffs
that a Nash in this game, we can
equilibrium
writetoplayer
we allow players choose1’s expected
random ies, and
payoff
strateg there
from willof
each behis two pure
to the proposed equilibrium.
actions as follows:
ies is not the only simple game that fails to have a pure-strategy
Recall the child’s gamev1 (H, q) = q × 1 +in(1which
rock-paper-scissors, − q)rock
× (−1)
beats= 2q − 1;
v1 (T , q) = q × (−1) + (1 − q) × 1 = 1 − 2q.
Example: Matching Pennies

Player 1 will prefer to play H over playing T if and only if

v1 (H, q) > v1 (T , q)
i.e., 2q − 1 > 1 − 2q,
i.e., q > 1/2.

Similarly playing T will be strictly better than playing H for

player 1 if and only if q < 1/2 . Finally, when q = 1/2, player
1 will be indifferent between playing H or T .
 p
BR1(q)
1
trategies
Example: Matching Pennies
p
BR1(q)
1

1 q
–
2
1

3 Player 1’s best-response correspondences in the Matching Pennies game.

ple analysis results in 1the best-response

q correspondence of player 1,
–
2
1
⎧
layer 1’s best-response correspondences inifthe 1
Matching Pennies game.
⎨p=0
⎪ q< 2
1
BR1(q) = p ∈ [0, 1] if q = 2
⎪
⎩
p=1
analysis results in the best-response if q > 21
correspondence of player 1,

ed in Figure 6.3.
CanNotice that this
⎧ write
you forisplayer
a best-response
2? correspondence, and
1 1
 ies and rock-paper-scissors games, the unique Nash equilibrium
y Nash equilibrium. It turns out that mixed-strategy equilibria
when they exist.equilibria
Multiple In fact when a game has multiple pure-strategy
ll almost always have other Nash equilibria in mixed strategies.
ng game:

Player 2
C R
M 0, 0 3, 5
Player 1
D 4, 4 0, 3

hat (M, R)Canand

you(D, for are
solveC) both
all the pure-strategy
Nash equilibria? Nash equilibria.
× 2 matrix games like this one, when there are two distinct
quilibria then there will almost always be a third one in mixed
 Multiple equilibria
d Strategies

p
1
BR1(q)

1
BR2(p)
–
6

3 q
–
7
1

URE 6.4 Best-response

For thiscorrespondences and Nash1’s
game, let player equilibria.
mixed strategy be given by
σ1 = (σ1 (M), σ1 (D)), with σ1 (M) = p and σ1 (D) = 1 − p, and let player
ing to see that all2’s mixed
three strategy
equilibria wouldbe given
show up inbya careful
σ2 = (σ 2 (C ), σ2 (R)), with σ2 (C ) = q and
drawing
ponse functions.σUsing
2 (R) = − q. functions v1(M, q) and v1(D, q)
the1payoff
Player
⎧
1 will mix when v1 (M, q) = v1 (D, q).
3
The⎪ p = 1 strategy
⎨mixed if q <Nash
7 equilibrium is:
BR1(q) =∗ p∗ ∈ [0, 1] if q = 3
{σ1 ⎪
, σ2 } = {(1/6, 5/6),7 (3/7, 4/7)}
⎩
p=0 if q > 73 .