Introduction to Game Theory MATH60141/MATH70141

Lectured by Dr Sam Brzezicki∗

Autumn Term 2023

Chapter 1: Prelude
1.1 About this Module
Welcome to Introduction to Game Theory! This course will provide you with a broad understanding of the
basics of game theory and investigates several of the different types of games that we can analyse using the
theory. We will endeavour to build up our knowledge in as ‘hands on’ a way as possible with an ‘active’
learning style, often by playing games in the classroom to understand the ideas and concepts. There will also
be recommendations of games to play as homework as well as collaboration on group assessed tasks (more on
that later). Throughout the module applications of the theory will be presented; we’ll see applications in the
economic behaviour of market competitors, competition between biological species, how to make informed
military decisions, politics and how to run for prime minister, traffic and congestion, bargaining and coming
to collaborative agreements, whether you should pay your taxes, applications in sport and also where to
meet for a date (and many more applications)!

The course has been split into seven chapters, starting with chapter 1, the prelude, where we are now,
which will be a short excursion exploring a few games and talking about what constitutes mathematical
game theory. Chapter 2 then goes on to look at the concepts of dominance, best response and equilibria.
In chapter 3 we introduce the notion of a mixed strategy and prove Nash’s theorem on the existence of an
equilibrium in finite games. Chapter 4 concerns zero sum games, or strictly competitive games, where we
will understand the concept of max-min and min-max strategies and see Von Neumann’s Minimax theorem.
Chapter 5 deals with an introduction to cooperation in games, and how players may improve their success
by cooperating on an agreed joint strategy. We will see the Nash bargaining solution arise from a set of
axioms. In chapter 6 we will focus on congestion games, or traffic games, a type of potential game. In this
context we will see how equilibria emerge and investigate whether these equilibria are always socially good
by using a ratio known as the price of anarchy. Finally we arrive at chapter 7 on combinatorial games. In
this chapter we will focus on the game of Nim and see how it plays a central role in the theory of these
games.
∗ [email protected]

1
1.1.1 Accessability

This module requires few prerequisites and has been designed with one aim of trying to ‘open the door’ for
more cross-departmental study into the mathematics department (which is typically difficult). As such, any
third/fourth year Mathematics, Physics or Computer Science student will be well equipped for studying the
module. For students slightly further reaching, like third/fourth year Engineers, Chemists and Biologists,
it is likely that you will be able to take the module with your current mathematics knowledge but it is
recommended that you speak with the lecturer about suitability so you (and me) are put at ease: there may
be one or two gaps in your knowledge for which a small amount of supplementary reading can be used to
bridge them.

1.1.2 Prerequisites

Though we have said the module is accessible, for those wanting a more detailed breakdown the following
topics should be considered prerequisite knowledge:

• Basic Probability and Statistics;

• One-dimensional Calculus;

• Mathematical notation;

• Vectors and Matrices;

• Logic and Proof techniques; we will utilise several different methods of proof and logical arguments,
this is perhaps the only area which could be lacking for students beyond Maths, Physics and Computer
Science. As mentioned - if you are unsure on your current level of knowledge, please get in touch with
the lecturer!

1.1.3 Study Materials and Reading Lists

There are seven problem sets to accompany these lecture notes; one for each chapter, which may build on
understanding from previous chapters as we develop our knowledge further. The problems will generally
start simple and progress in difficulty throughout each sheet. The majority of the problems should enrich
your understanding of the concepts and prepare you well for examination (there will be nothing included
that would constitute a waste of your time). Occasionally, problems with one of two symbols by the question
number may appear. Firstly star problems, which are considered very difficult and not of exam standard.
Often the solution of these may include references to research papers for instance. Nevertheless they are
included for interest and often take the theory in the notes a little further. Secondly diamond problems,
which are not necessarily of any specific difficulty, but are open problems that have deliberately been left
unsolved (no solutions will be provided for these). These are included to provide ideas for possible group
coursework topics (see soon). Usually these are ideas and problems the lecturer thinks are interesting, but
hasn’t had a chance to investigate yet!

The course has been designed so that it is entirely self-contained; meaning that these lecture notes, along
with the seven problem sets contain everything that you will need to prepare for the examination, so if you
don’t learn efficiently from using books and other resources don’t worry, everything you need is included.

2
Nevertheless, many students like to learn from the use of textbooks on the topic so a short reading list is
included below. There are many books on game theory and really any textbook introducing the topic will
contain similar information, so feel free to browse the library if you’re looking for a different choice. Some
useful texts include:

• Game Theory Basics (Bernhard von Stengel): For those looking for a companion book for the course
this should be your number one choice. We will very closely follow parts of this book throughout
several chapters of this course. In fact this book really re-shaped some parts of this module, so if you
prefer to learn from textbooks then the corresponding chapters from this book are my recommended
reading;

• Introduction to Game Theory (Peter Morris);

• The Theory of Games (Jianhua Wang);

• Introducing Game Theory and its Applications (Mendelson, Chapman and Hall);

• Game Theory: A Nontechnical Introduction to the Analysis of Strategy (Roger McCain).

Below are some other resources that I recommend (of course I have developed these notes because I want
you to use them, but hearing different explanations of the same concept can help you to master it):

• There is an excellent series of lectures on Game Theory by Ben Polak from Yale university on Youtube
(YaleCourses). The lecture series is delivered primarily to Economics majors (eww), meaning the level
of mathematics is generally lower than what we will employ, but nevertheless Ben does a fantastic job
of explaining the concepts of the theory - for those that are interested in more detail on the social,
political and economic aspects of the theory, then these lectures are highly recommended. Many of the
classroom games we will play have also been taken from Ben’s game theory course.

• The lecture notes on Algorithmic Game Theory by Tim Roughgarden (these are also written into a
book called ‘Twenty Lectures on Algorithmic Game Theory’). For our course they prove an excellent
resource for Congestion Games.

• Finally, and most importantly, the lecture notes and module on the Dynamics and Learning of Iterated
Games (which you should certainly consider taking if you are interested in this module) taught by Pro-
fessor Sebastian van Strien here in the Mathematics department at Imperial is an excellent companion
to this module. This module has been designed so that there is as minimal overlap as possible with this
companion. Where we have some gaps on the theory of repeated games and dynamics within games,
such as evolutionary game theory, that module delves deep into these concepts.

If you find a textbook or any resource you believe to be excellent, please let me know!

1.1.4 Assessment

The module will have two assessed group coursework tasks, each worth 5%. These will take place during the
Autumn term. More details about these closer to the time! There will be a 2-hour exam with four questions
worth a total of 90% in the May-June exam season. For Master’s students the weightings remain the same
but this exam will be 2-hours 30-minutes long and contain an extra fifth question aimed at Master’s level.

3
More details regarding best exam preparation will be given later in term. It is likely I will produce several
‘mock’ papers for the module ready for April time that you can use as preparation for what to expect in the
exam (since this is the first year the module has taken place and there are no past papers to study from).

1.1.5 Final Administrative Things

There will be a weekly ‘office hour’ (I’ll be in my office and free for one hour to answer any questions you
might have about anything for this module) that you are welcome to come and drop by during if you have
any questions that you want to ask me. My office is room 655 in Huxley. During the first lecture we will do
a vote as to when the office hour will take place.

The course has 30 hours of scheduled lectures; usually 3 hours per week for 10 weeks. We will use ap-
proximately 25 of these as lectures, where we will move through these lecture notes and cover new content.
The remaining 5 hours we will dedicate to being ‘problem classes’; spread approximately one every 2 weeks,
where we will spend the time going through some of the problems from the problem sheets. We may also
play some additional classroom games during these sessions that didn’t quite fit into the main lecture notes,
so don’t miss out!

1.2 The Everybody Bids Game

Let’s play a game. We’ll involve everybody in the class (I’ll play too!). Everyone in the class is going to be
given a piece of paper, and on that paper you need to write down your name and a whole number between
0 and 100 inclusive. Don’t let anyone see your number and don’t discuss your decision with anyone! Once
everyone has decided on their number, the papers will all be collected and the average (the mean) of the
numbers we all chose will be calculated (any bids that are not whole numbers within the range given will be
discarded and not count towards the average). Whoever has bid closest to one half of the class average will
be declared the winner of the game! If there are multiple people with the same closest bid then they are all
declared winners of the game.

Activity: Play the game!

1.2.1 Discussion

There are a few useful concepts to discuss here. We will generalise and formalise some of these ideas later
in the course, but for now let’s identify them and appreciate how they might influence dictating the play of
the game.

Let’s think about choosing a number in the range from 51 to 100. We can notice that this is not going to
be a good choice (it’s a bad choice actually, in fact a strictly dominated strategy - we will define what
exactly a strategy is and what we mean by these terms soon), why? Well let’s think, even if everyone in
the class bid 100, then one half of the average would be 1/2 × 100 = 50, so 50 is definitely going to be closer
to one half of the class average than 51, or 52, or 53, or anything up to and including 100, regardless of
what everyone else is doing. So we can conclude that choosing any number from 51 to 100 is certainly not
optimal, since choosing 50 would give us a bid strictly closer to one half of the average, a better choice.

4
Now that we know this, what can we say about the game and the players. Well it’s not just us that this
knowledge applies to, we are not playing against beings that can’t think for themselves, it applies to every
player in the game. So no rational player (or no intelligent player that wants to try and win) will have
chosen a number between 51 and 100. This means that we can, in essence, re-state the game. It’s the same
game as before, except now we know that the players are going to choose numbers between 0 and 50 (we
can ‘delete’ or ‘remove’ the numbers 51 to 100 from the game as no rational player will choose them). What
does this mean about a play of the game?

Well, let’s take our logic from earlier one step further, we can conclude that all the numbers from 26 to 50
are now bad choices, since even if everyone bidded 50, one half of the average would then be 25. So now no
rational player will choose any number from 26 to 50. Can you see what is going to happen?

Indeed if we continue the above argument, after each iteration removing all the ‘bad’ choices, we will even-
tually reach the situation where everyone is choosing the number 0. There is then no better any player can
do, since the average is now at 0 (if everyone is bidding 0), so bidding any other number will cause the player
to be further away from one half of the average than if they were to bid 0. Each player is now responding
best to what all the other players are doing. This situation is an important concept in game theory, known
as an equilibrium, or a Nash equilibrium (named after John Nash, one of the most famous mathematicians
to progress the theory). We will talk about these equilibria extensively throughout the course - in fact, so
powerful is the concept of an equilibrium in a game, that one can even think of classical game theory as
being the search for equilibria in games.

1.3 The Sweets Dilemma

Let’s play another game. This game is played between two players and each player has the choice to play
strategy 1 or strategy 2. To involve everyone in the class we are going to do the following: on your sheet of
paper write your name and whether you want to play strategy 1 or strategy 2, we will then collect in all the
papers and randomly assign you against each other in pairs to see what the outcomes are. If you want to
discuss your ideas with those around you before you make your choice feel free to do so! Before that I need
to tell you what the outcomes of the game are under each possibility.

Each player has two options available to them, strategy 1 or strategy 2. If both players choose strategy 1,
then they each receive nothing. If both players choose strategy 2, then they each receive 1 sweet, but if
one player chooses strategy 1 and the other player chooses strategy 2 then the player who chose strategy 1
receives their sweet and their opponent’s sweet (a total of 2 sweets!) and the player who chose strategy 2
receives nothing.

Activity: Play the game!

1.3.1 Discussion

Let’s discuss the game a little bit as there are a few useful things to note here. First, to try and help visualise
the game, let’s put the information into a table, shown in figure 1. Let’s call our first player A and our

5
second player B and denote the strategies as s1 and s2 (strategy 1 and strategy 2).

Figure 1: The Sweets Dilemma

We will see another example with more detail in section 1.7, but this is a common way to display certain
two-player games. The rows correspond to each strategy of player A and the columns to each strategy of
player B. In each cell a pair of entries are given: the first the amount of sweets player A receives if that pair
of strategies is chosen, the second entry the amount of sweets player B receives. These values are widely
referred to as the payoffs to the players in the game.

Now that the information is more clearly summarised we might notice something interesting - that playing
strategy 2 feels ‘a bit weaker’ in some way. Indeed, notice that if our opponent plays strategy 2, then we
will get more sweets (we’ll get 2) by playing strategy 1 than if we played strategy 2 (where we’d get only
1 sweet), and if they play strategy 1, then we will get nothing regardless of what we choose. So it feels
like strategy 1 ‘does better’ for us than strategy 2 (indeed, as we will define more rigorously later, strategy
1 weakly dominates strategy 2). By this logic it feels like, since strategy 1 performs either the same or
better than strategy 2, we should play strategy 1. But if both players follow this line of thought then they
end up with no sweets at all! However, this is an (and in due course you will see how to show this is the
only) equilibrium of this game played in this manner.

There is more to say here however. It is likely upon playing the game in class that actually some players
did choose strategy 2, and moreover some of those players were paired up together and received a sweet
each. Why? Strategy 1 seemed at least as good as 2 if not better, so why would some players still choose
strategy 2? Analysing the game in the way we have in figure 1 is making a judgement on our players - that
what they care about is getting as many sweets as possible for themselves, hence their payoffs are exactly
the number of sweets received. For some of you this might be a really bad assumption on which to model
your payoffs! Some of you might not like sweets at all, so might see even 2 sweets as worthless. Some players

6
may value the thought of sharing and negatively value the thought of backstabbing their classmates, making
them uneasy about the choice of strategy 1, resulting in a tendency to go for strategy 2. All of these ideas
would end up changing the payoffs for a player, so it is very important to have a clear concept of what each
player’s payoff is in a game to be able to analyse it effectively!

Remark: In this course we will not look at the concept of a player’s utility, but those interested in how
payoffs can take account for these personal ‘values’ of a player might be interested in reading more about
utility theory.

A final point to mention here is that of communication. In the game communication was allowed beforehand
with those nearby; people you would unlikely be assigned against. Even if you could communicate with the
person you were going to be playing against, would it help? Notice that even if a player says they will play
strategy 2, they are still incentivised (if they care about maximising their own amount of sweets) to lie about
this and switch to strategy 1 as it will give them the chance of getting more sweets than strategy 2 could.
So under the premise that this game is being played non-cooperatively, communication doesn’t help here
and players are incentivised to try and convince each other to play strategy 2 so that they can profit from
playing strategy 1. Ultimately this will result in both players playing strategy 1 most of the time!

However, if we could change the rules slightly, and allow the players to play the game cooperatively with
an enforced or binding agreement on their strategy choices, i.e they had to agree on a pair of strategies
and then were obliged to see that agreement through, then they could collectively do better, for example,
by agreeing on playing strategy 2 and each receiving 1 sweet. This is much better than both players trying
to backstab one another and ending up with nothing. We will see more of this idea later in chapter 5 when
we discuss co-operative game theory.

1.4 Principles, Elements and Classification of Games

As we have highlighted through our discussions of the two games we’ve played so far, in the real world
people play games with a variety of principles behind their actions and at times without thought or rationality.
Although this is a very interesting philosophical and ethical problem to discuss, to build up an understanding
of the core mathematical principles of games, we will make some assumptions based on how we expect our
players to act which will pertain throughout the analysis we do in the course!

1.4.1 Principles of Game Theory

Some principles of mathematical game theory that we will assume throughout the course are:

• Values: In mathematical game theory we always analyse games as if every one of the players is trying
to win or optimise their own score/winnings. That is, we assume the players are rational and greedy.
The players don’t put any value on the ‘fun’ of playing the game or on any exterior societal implications
or morals. The players don’t act in irrational ways for any reason. In any games we model we will
assume that a player’s payoff will already factor in their values. For example, in the case where a player
would gain one sweet but their opponent nothing, a player who cares about others receiving sweets
too will have a lower payoff than a player who cares only about getting as many sweets as possible for
themselves. This moral value is then factored into the payoffs for the game.

7
 • Rules: Games in this course and those which can be studied by game theory have strict rules. In the
real world, sometimes thinking ‘outside the box’ to discover a ‘new move’ or option in a situation is
a great skill, but in this course the games will always have specified moves which are available to the
players - the skill comes in determining which are the best moves!

• No Cheating: Linked to the last point, in the real world cheating can happen: you can swap a card
with one kept up your sleeve, you can change the state of a game while your opponent is distracted,
etc. To study games analytically we will assume no cheating takes place.

1.4.2 Elements of a Game

Some games we encounter will involve different elements, possibly there are random moves incorporated or
the physical location of the game matters. There are three core elements within all games we will study that
dictate our analysis. These are:
• Players: the number of players in the game needs to be given.

• Moves: the list of all possible moves at each point in the game for all players in the game and the
order in which these moves will take place (many games are asymmetric and different players will have
different moves available to them).

• Payoffs: The list of all payoffs (outcomes at the end of the game) for all players needs to be given.
This information will usually be given at the start of a game (in everything we will look at this will be the
case), though there are some interesting examples of games where this is not the case.

1.4.3 Classification of Games

There are many ways in which we can classify games, often related to the manner in which they are played.
Some of the main ways are shown here:
• Number of players: games can be played with different numbers of players, even single player where
the player is playing against ‘nature’, ‘the casino‘ or some set ‘algorithm’ - things that don’t have
conscious choice but follow an algorithm or have chance elements.

• Whether moves are simultaneous or sequential.

• Complete or incomplete information: a game is said to have complete information if all players know
the structure of the game: the amount, and order in which the players move, all possible moves in each
position and the payoffs for all possible outcomes for all players (Chess and Tic-Tac-Toe for example).
If any of these are not known the game has incomplete information (Bridge and many other card games
for example).

• Zero-sum (or constant-sum) games: these games have the property that the sum of all the payoffs to
all the players for each possible outcome of the game equals zero (or a constant).

• Communication: are the players allowed to communicate before/during the game or not at all?

• Cooperative or Non-cooperative: some games may be played cooperatively where the players try to
work together to maximise their payoffs. Other games are non-cooperative where the players are
playing against each other.

8
1.5 Strategies
Now is a good time to introduce an important definition, what we mean by the strategy of a player in a
game. This is different from the concept of a move in a game and also different to an outcome of a game,
which we will define fairly loosely here (as their meaning can become more specific depending on the type
of game being played).

Definition 1.1. A move refers to the action a player must make on their turn to progress from one game
position to the next position.

Definition 1.2. An outcome of a game refers to the final result (which may be win/lose/draw or a
combination of payoffs for all players) of a game once the game has been played.

Definition 1.3. A strategy for a player involves a complete description of all the moves that will be
made in any game position, including responses to any random moves (tossing coins, shuffling cards), and
the opponents’ moves. In other words, a strategy is a program which can be followed to play the game
mechanically.

There is one last definition we make now of a pure strategy. These will be the type of strategies considered
throughout this and the next chapter. Later, in chapter 3, we will generalise these pure strategies to include
mixed strategies (which we will define formally later but give an example now to clarify the difference
between these and pure strategies).

Definition 1.4. A pure strategy is a strategy that doesn’t involve any self-imposed random chances of
playing any moves.

This definition is based on negation, so here’s a quick example to try to clarify it slightly: a fair coin is tossed
and you have to guess whether it will land as heads or tails. Your pure strategies are to choose either heads
or tails with absolute intent. A mixed strategy would constitute actively deciding upon a probability that
you will choose heads. For example choosing heads with probability 1/2 (and hence tails with probability
1/2) would constitute a mixed strategy - for now don’t worry about these!

Let’s look at some examples to understand strategies better:

1.5.1 Examples

1). Two companies, A and B, each produce and sell a product. Each wants to increase their market share.
Company A intends to do some of either:

– Spend 10% of profits on advertising,

– Reduce the price of the product, or
– Give a free gift on purchase of the product.

Company B intends to do one of either:

– Spend 5% of profits on advertising, or

– Reduce the price of the product.

9
 The companies act simultaneously in this game. How many pure strategies does each company
have?

Answer). Company A has 23 − 1 = 7 pure strategies. Company B has 2 pure strategies.

Exercise: If company B acts after company A (with full knowledge of company A’s choices), then how many
pure strategies does company B have?

2). A game consists of a fair coin being tossed showing either H (heads) or T (tails). Player A must then
make a move from a choice of n possible moves {m1 , m2 , · · · mn }. How many pure strategies does
player A have in this game?

Answer). A pure strategy in this game needs to tell player A what to do in either outcome of the coin toss, so
it is a map:

{H, T } 7−→ {m1 , m2 , · · · mn }.

For example, one possible pure strategy is: If the coin shows H, play m1 , if the coin shows T , play m2 .
So, in total, player A has n2 different possible pure strategies for the game, one corresponding to each
mapping.
Notice this is very different to the set of possible moves for player A, where there are n, and also
different to the set of possible outcomes, where there are 2n (each possible coin outcome and move
pairing).

In the two examples above, the players all had a finite number of pure strategies to choose from, but this
does not have to be the case - a player may have an infinite number of pure strategies, for example; a player
must choose a real number from the interval [0, 1]. This motivates the following definition that we will use
to describe games.

Definition 1.5. We call a game finite if all players in the game have a finite number of pure strategies. If
at least one player has an infinite number of pure strategies then the game is an infinite game.

1.6 Goofspiel (The Game of Pure Strategies)

Let’s look at another game; Goofspiel, or sometimes called the game of pure strategies. This is a
two-player game (let’s call the players A and B). At the start of the game a positive integer n is decided,
which is traditionally 13 (as the game is commonly played with a deck of cards and there are 13 cards of
each suit). Each player is then given n cards numbered from 1 through to n (so in a deck of cards each
player takes all the cards of one suit and then the ace is considered to take value 1 and the Jack, Queen
and King are considered to take values 11, 12 and 13). Let’s say player A has red cards and player B
has blue cards, just to help distinguish them. Then another set of n cards (again numbered 1 through n),
which are coloured black, are shuffled and dealt face up into the centre, one at a time. After each black
card is dealt, players A and B simultaneously bid for the card by playing one of their coloured cards into
the centre. The player who places the larger bid (the higher valued card) scores a number of points equal
to the value that was on the black card dealt into the centre. If the two bids are equal (for instance both

10
players play their 1 card), then the number of points awarded to each player is equal to half the value of
the black card (or, in some variants, both players score zero points here). This comprises one round of
the game, after which all three cards involved are removed from the game. The game then continues with
the next black card being dealt into the centre and the players bidding on its points value using their re-
maining coloured cards. Once all cards have been used (after n rounds) the game concludes and the player
with the higher points total is declared the winner. If the players have an equal score then the game is drawn.

Activity: Play a game of 5 card (n = 5) Goofspiel. Discuss what seem to be ‘good’ strategies in this game.

Now we have seen the game of Goofspiel in action, let’s use it as another example of calculating the number
of pure strategies available to us. First let’s consider 2-card Goofspiel and ask the following question... how
many pure strategies does each player have in this game?

Well, a strategy for a player tells the player how to respond to the first dealt card. After that the rest of the
game is fully determined (as each player and the face-down deck have only 1 card left in them each). So, a
pure strategy is a map:

{dealt card} 7−→ {players first bid},

i.e. {1, 2} 7−→ {1, 2}.

Thus there are 22 = 4 different maps and hence there are 4 different pure strategies.

Remark: You will investigate more on the game of Goofspiel in problem sheet 1.

1.7 Displaying Games

Throughout this module we will encounter a wide variety of different games. Often there is a most convenient
way to display them. There are two common ways to display games and we will see both throughout the
course: normal (or strategic) form, or in extensive (or tree) form.

In normal form two-player games are represented in a table or matrix (for three or more players extra
dimensions would be needed). A schematic is shown in figure 2 and we will look at an example game soon
to explain the notation.

11
 Figure 2: Schematic of a game in normal form.

In extensive form games are described using a tree, as shown in the schematic in figure 3.

Figure 3: Schematic of a game in extensive form.

In this form strategies of the players form paths through the tree. The payoffs to each player are usually
given at the bottom of each branch through the tree.

The two different approaches are equivalent, though it is often more convenient to use one approach over

12
the other based on the game being studied. Let’s look at an example to better see how a game might be
displayed in these ways.

1.7.1 Example Game

Suppose players A and B are playing a game. Player A has two possible pure strategies (a1 and a2 ) and so
does player B (b1 and b2 ).

• If player A chooses a1 and player B chooses b1 , player A receives a score of 1 and player B gets 2;

• If player A chooses a1 and player B chooses b2 , player A receives a score of 2 and player B gets nothing;

• If player A chooses a2 and player B chooses b1 , player A receives a score of 5 and player B gets 1;

• Finally, if player A chooses a2 and player B chooses b2 then player A receives a score of 0 and player
B gets 3.

In normal form we would display this game as shown in figure 4.

Figure 4: The example game displayed in normal form.

Player A is placed on the left and their pure strategies are represented by the different rows. Player B is
placed at the top and their pure strategies are represented by the different columns. Inside the grid, each
cell contains a pair of payoffs corresponding to the two strategies played; the first payoff corresponding to
player A and the second to player B. If, for example, player A chooses a2 and player B chooses b2 the grid
shows 0, 3 representing that player A receives 0 and player B receives 3. This grid of payoffs is often referred
to as the payoff matrix for the game.

In extensive form the game could be displayed as shown in figure 5.

13
 Figure 5: The example game displayed in a possible extensive form.

This tree diagram has been drawn in a way where it looks as if moves are sequential. This could indeed be
the case but need not be - a tree diagram can still be drawn with pairs of simultaneous moves as in this
example. A pathway through the tree, always descending, results in a payoff at the end of a branch.

Exercise: In this game we could draw the tree diagram with B’s moves first then A’s and this would also
be an equivalent way of representing the same game. Can you draw this alternative extensive form of this
game?

1.8 The Solution of a Game

Before we move on to the next chapter, let’s pause for a moment to briefly discuss what we are actually
trying to achieve by analysing games, i.e. what constitutes a solution of a game?

We know for example, if asked to solve a linear equation, like 2x + 4 = 0, that we have a unique solution
to the problem given by x = −2. If we were tasked with solving a quadratic, like x2 − 3x + 2 = 0, then we
can have either two, one (repeated) or no real solutions (or two if we consider complex solutions), here we
have two solutions: x = 1 or x = 2. Maybe we are given a Diophantine equation with possibly infinitely
many solutions, or a system of equations. In short, we know what it means to solve an equation.

Question: But what does it mean to solve a game? What should a solution to a game consist of?

A solution to a game should tell the players playing the game how to play ‘optimally’, i.e a solution should
provide an ‘optimal’ strategy for each player in the game (notice this is for each player). With these
strategies we can deduce the resulting payoff for each player as part of the solution.

Notice that I have put quotation marks around the word optimal in the above. Indeed, what exactly we mean
by a strategy being optimal is a non-trivial question and is one we will work toward answering throughout

14
the module! For now, let’s just understand that to solve a game we must provide strategies for each player
that each need to fulfil some condition of optimality - precisely what this condition is we will develop in the
coming lectures.

Note that, as discussed above, equations for example, can have more than one solution. So what about
games? Can games have more than one solution? Or perhaps the more terrifying thought - some equations
have no solutions, so do some games have no solutions? Let’s eagerly press forward with these questions in
mind!

15