GAME THEORY

Contents
Introduction History Applications Prisoners Dilemma Dominant Firm Strategy Mixed Strategies Case Study Evolutionary Game Theory Approach to Combat Money Laundering

Game Theory Introduction

Game theory is the branch of decision theory Game theory is the formal study of conict and cooperation The concepts of game theory provide a language to formulate, structure, analyze, and understand strategic scenarios The problems are called games and the participants are called players
game theory deals with any problem in which each player s strategy depends on what the other players do.

Few Examples where Game Theory helps

Friends choosing where to go have dinner Parents trying to get children to behave Commuters deciding how to go to work Businesses competing in a market Diplomats negotiating a treaty Gamblers betting in a card game

History
Game Theoretic Analysis was used in study of duopoly in 1838 by Antoine Cournot The mathematician Emile Borel suggested a formal theory of games in 1921, Theory of parlor games was introduced by the mathematician John von Neumann in 1928 Game theory was established as a field in its own right after the 1944 publication of the monumental volume Theory of Games and Economic Behavior by von Neumann and the economist Oskar Morgenstern In 1950 John Nosh introduced concept of Equilibrium point

Principles of Game Theory

Each decision maker has two or more choices or sequences of choices ("plays"). All possible combinations of decisions or plays result in a clear outcome: win or lose. The scenarios have a well-defined outcome and decision makers receive a "payoff (the value of the outcome to the participants). That is, participants will gain or lose something depending on the outcome. The decision makers know the rules of the game as well as the payoffs of the other decision makers. The decision makers are rational: when faced with two alternatives, players will choose the option that provides the greatest benefits.

Applications of Game Theory

Artificial Intelligence:
Difference between human and machine is ability to make independent decisions based on environmental stimuli In future AI may make new decisions unplanned for by their creators. This would require the programs to be able to generate new payoff matrices based on the observed stimuli and experience

Economics:
Many of the interactions in the business world may be modeled using game theory methodology A example is that of the similarity of the price setting of oligopolies to the Prisoners Dilemma

Applications of Game Theory

Resource Allocation and Networking:

Computer network bandwidth can be viewed as a limited resource. The users on the network compete for that resource

There competition can be simulated using game theory models

Biology:
There is cooperation exists between different species. The reason behind this coexistence can be modeled using game theory For example, birds called ziczacs enter crocodiles mouths to eat parasites. This symbiosis allows crocodile to achieve good oral hygiene and allows the ziczacs to have a decent meal

Politics

The study of military strategies

The study of international negotiations Bargaining

Prisoners Dilemma
MR A Confess MR B ( -8, -8) ( 0, -15) Dont Confess

Prisoners Dilemma

Confess

Dont Confess

( -15, 0)

( -1, -1)

Prisoners Dilemma
MR A MR B Confess Dont Confess

Prisoners Dilemma

Confess

( -8, -8)

( 0, -15)

Dont Confess

( -15, 0)

( -1, -1)

Prisoners Dilemma

Conclusion: Mr. B will confess And Mr. A?

Prisoners Dilemma
MR A Confess MR B ( -8, -8) ( 0, -15) Dont Confess

Confess

Dont Confess

( -15, 0)

( -1, -1)

Prisoners Dilemma
MR A MR B Confess Dont Confess

Confess

( -8, -8)

( 0, -15)

Dont Confess

( -15, 0)

( -1, -1)

Prisoners Dilemma
Conclusion: Mr.A confesses also

Prisoners Dilemma

Both get 8 years, even though if they cooperated, they could get off with one year each For both, confession is a dominant strategy: a strategy that yields a better outcome regardless of the opponents choice

Prisoners Dilemma Prisoners Dilemma

What would the B and A decide if they could negotiate? They could both become better off if they reached the cooperative solution.
which is why police interrogate suspects in separate rooms. Equilibrium need not be efficient. Noncooperative equilibrium in the Prisoners dilemma results in a solution that is not the best possible outcome for the parties.

Equilibrium

Nash Equilibrium: Neither player has an incentive to change strategy, given the other players choice
Both confess is a Nash Equilibrium Both dont confess is not a Nash Equilibrium, rival will always want to renege

Dominant Firm Game

Two firms, one large and one small Either firm can announce an output level (lead) or else wait to see what the rival does and then produce an amount that does not saturate the market.

Dominant Firm Game

Dominant Lead Follow

Subordinate

Lead

( 0.5, 4)

( 3, 2)

Follow

( 1, 8)

( 0.5, 1)

Dominant Firm Game

Dominant
Lead Follow

Subordinate

Lead

( 0.5, 4)

( 3, 2)

Follow

( 1, 8)

( 0.5, 1)

Dominant Firm Game

Dominant
Lead Follow

Subordinate

Lead

( 0.5, 4)

( 3, 2)

Follow

( 1, 8)

( 0.5, 1)

Dominant Firm Game

Conclusion:
Dominant Firm will always lead..

But what about the Subordinate firm?

Dominant Firm Game

Dominant Lead Follow

Subordinate

Lead

( 0.5, 4)

( 3, 2)

Follow

( 1, 8)

( 0.5, 1)

Dominant Firm Game

Dominant Lead Follow

Subordinate

Lead

( 0.5, 4)

( 3, 2)

Follow

( 1, 8)

( 0.5, 1)

Dominant Firm Game

Conclusion:
No dominant strategy for the Subordinate firm. Does this mean we cannot predict what they will do?

Dominant Firm Game

Dominant Lead Follow

Subordinate

Lead

( 0.5, 4)

( 3, 2)

Follow

( 1, 8)

( 0.5, 1)

Dominant Firm Game

Conclusion:
Subordinate firm will always follow, because dominant firm will always lead.

Minimax Theorem
A pure strategy defines a specific move or action that a player will follow in every possible attainable situation in a game. A mixed strategy is an assignment of a probability to each pure strategy. This allows for a player to randomly select a pure strategy. Since probabilities are continuous, there are infinitely many mixed strategies available to a player, even if their strategy set is finite. The Minimax Theorem guarantees the existence of exactly one equilibrium point for any 2-player zero-sum matrix game. However, the equilibrium point may be the result of the use of pure or mixed strategies by either one or both players.

To find equilibrium points for 2x2 zero-sum matrix Game

Determine if the equilibrium point is associated with pure strategies: To do this we determine if the row players maximin strategy and the column players minimax strategy coincide in the same outcome.

If pure strategies do not produce an equilibrium point, we proceed as follows: For each player we find the probabilities that will provide the lowest expected payoff for the other player.

Finding an Equilibrium Point 2x2 Zero Sum Game

When a pitcher and batter face each other in a baseball game, we can consider each pitch as a simultaneous move zero-sum game. The pitcher only has two strategies: throw a curve or a fastball. The batter only has two strategies: swing expecting a fastball or swing expecting a curve ball. The outcomes we associate with each pair of strategies is not whether the ball is hit or not but rather the probability the batter has of hitting the ball.
Pitcher

Fastball
Batter

Curve .200 .400

Fastball Curve

.300 .100

Finding an Equilibrium Point 2x2 Zero Sum Game

What strategy (mixed or pure) should each player (batter and pitcher) adopt to optimize their payoff?
We are first checking for an equilibrium point in pure strategies
Pitcher

Fastball
Batter

Curve .200 .400

.400 .200 .100

Fastball Curve

.300 .100
.300

First, find the minima for the row player. Now what is the best he can do assuming the pitcher will play to achieve these values? The maximin strategy for the row player is .200. This is the best of the worst outcomes.

Finding an Equilibrium Point 2x2 Zero Sum Game

Pitcher

Fastball Fastball
Batter

Curve .200 .400

.400 .200 .100

.300 .100
.300

Curve

Now we search for the best strategy for the column player. Find the minimax strategy for the column player. For the pitcher, by choosing always to throw a fastball, the worst he can do is give up a .300 average for this batter. For the batter, by choosing always to swing for a fastball, the worst he can do is average a .200.

Finding an Equilibrium Point 2x2 Zero Sum Game

Pitcher

Fastball
Batter

Curve .200
.200
.100

Fastball

.300

Curve

.100

.400
.400

.300

Because the maximin and minimax strategies do not equal, the equilibrium point is not found with pure strategies. The batters worst case average is .200 and the pitchers worst case average is .300 using pure strategies. Now by mixing strategies, each player will try to get as much of that difference as possible. What strategy should each adopt assuming best play by the opponent?

Finding an Equilibrium Point 2x2 Zero Sum Game Pitcher p

1-p

Lets find the optimal mixed strategy for the pitcher first.

Fastball Fastball
Batter

Curve .200 .400

Let p equal the probability the pitcher will throw a fastball thus 1-p equals the probability he will throw a curve.

.300 .100

Curve

We write the expected value (which is the average payoff in repeated trials) for the batter based on the probabilities p and 1-p, for each of the batters strategies. EF = .300p + .200(1-p) is the expected value the batter would receive choosing fastball. EC = .100p + .400(1-p) is the expected value the batter would receive choosing to swing for a curveball.

Finding an Equilibrium Point 2x2 Zero Sum Game

The pitcher wants to find the probability p, for throwing a fastball, that will minimize the batters expected value. We simplify each expression above, which gives expected value (for each strategy) as a function of probability, p, graph each function, and find a minimum value, as follows: EF = .300p + .200(1-p) = .3p + .2 - .2p = .1p + .2

Likewise,
EC = .100p + .400(1-p) = .1p + .4 - .4p = -.3p + .4 Note that these expected value functions are always linear in p, and are thus easy to graph

Finding an Equilibrium Point 2x2 Zero Sum Game

We have expected value functions for each strategy of the batter, for a given choice of p for the pitcher. These functions are: EF = .1p + .2 and EC = -.3p + .4 Graphing these functions, we find expected value (payoff to batter) .4 EF .1 EC p 1 The choice of p that corresponds to the minimum payoff for the batter will occur at the intersection of these two lines.

Finding an Equilibrium Point 2x2 Zero Sum Game

We have expected value functions for each strategy of the batter, for a given choice of p for the pitcher. These functions are: EF = .1p + .2 and EC = -.3p + .4 Graphing these functions, we find expected value (payoff to batter) .4 .4p = .2 EF .1 EC 1 Because the payoff of .25 ( a batting average of .250 ) corresponds to the batters (row players) payoff, this is the value of the game. which gives a solution of p = 1/2 p to find the intersection point, we solve the equation: .1p + .2 = -.3p + .4 (1/2, .25) yielding

Finding an Equilibrium Point 2x2 Zero Sum Game

The conclusion we have is that the pitcher should throw fastballs with probability p = 1/2 and curveballs with probability 1-p = 1-1/2 = 1/2 to minimize the batters average payoff. We also have found the value of the game, which is .250.

But we have not yet determined the strategy the batter should use to maximize payoff assuming best play by the pitcher. This is done the same way for the batter as it was done for the pitcher but the probabilities associated with an optimal mixed strategy may be different for each player. Nevertheless, assuming best play by each player, there is exactly one outcome, which is the value of the game. Well now find the optimal mixed strategy for the batter

Finding an Equilibrium Point 2x2 Zero Sum Game Now for the batters best-play strategy Let q be the probability for the batter of swinging for a fast ball and thus 1-q will be the probability for the batter of swinging for a curve. Pitcher

Fastball
q batter 1-q

Curve

Fastball
Curve

.300
.100

.200
.400

Well calculate expected value functions for the pitcher for each strategy which depend on the value of q. Then well determine the value of q that is best for the batter assuming best play by the pitcher.

Finding an Equilibrium Point 2x2 Zero Sum Game Pitcher

Fastball
q batter 1-q

Curve

Fastball
Curve

.300
.100

.200
.400

If the pitcher chooses the fastball strategy, the pitchers expected payoff will be EF = .300q + .100(1-q) = .3q + .1 - .1q = .2q + .1

And if the pitcher chooses the curveball strategy, the pitchers resulting expected value will be
EC = .200q + .400(1-q) = .2q + .4 - .4q = -.2q + .4

Finding an Equilibrium Point 2x2 Zero Sum Game Expected payoff values for the pitcher, for each pure strategy, are as follows: EF = .2q + .1 and EC = -.2q + .4

We seek the value of q that maximizes payoff to the batter (the worst case for the pitcher). That is, what should the batter choose assuming best play by the pitcher.
Graphing these functions, over values of q from q = 0 to q = 1, we have

expected value (payoff to pitcher)

.4 .3 .2 .1 EF

EC
q 1

Finding an Equilibrium Point 2x2 Zero Sum Game Again, the choice of q that will maximize payoff for the batter assuming best play by the pitcher will occur at the intersection of these lines. Solving .2q + .1 = -.2q + .4 we get .4q = .3 and thus q = 3/4.

Therefore, the strategy for swinging for a curve is 1 q = 1 3/4 = 1/4.

Conclusion: The batter should mix strategies of swinging for a fastball and a curve ball with probability 3/4 and 1/4, respectively. Intersection of payoffs for pure strategies at the point (3/4, 1/4) EF

expected value (payoff to pitcher)

.4 .3 .2 .1

EC
q 1

Finding an Equilibrium Point 2x2 Zero Sum Game We have found what the Minimax Theorem guaranteed exists: the equilibrium point for a 2x2 zero-sum matrix game. In this case that equilibrium point is found at mixed strategies for both players. The equilibrium point is the combination of strategies with probabilities (1/2, 1/2) for the pitcher and (3/4, 1/4) for the batter of pure strategies of fastball and curveball, respectively.

That is, these probabilities will yield each player the best possible payoff assuming best play by the opponent. The pitcher should mix fastballs and curveballs with probabilities 1/2 and 1/2 for each. The batter will maximize his average by swinging for fastballs 3/4 of the time and swinging for curves 1/4 of the time.
The value of this particular game is the payoff to the row player (the batter) which is .250.

After foreign exchange and the oil industry, the laundering of dirty money is the worlds thirdlargest business.
Jeffrey Robinson, The Laundrymen
How Much Is Laundered? IMF estimate - 2-5% Global GDP

What is Money Laundering?

Definition: The process of disguising the proceeds of crime in an effort to conceal their illicit origins and legitimize their future use Objective: To conceal true ownership and origin of the proceeds, a desire to maintain control, a need to change the form of the proceeds Techniques: They can be simple, diverse, complex, subtle, but secret

Proceeds = any economic advantage derived directly or indirectly from criminal offenses

Money Laundering Cycle

1. Predicate Crimes
Corruption and Bribery Fraud Organized crime Drug and human trafficking Environmental crime Terrorism Other serious crimes

4. INTEGRATION
The last stage in the laundering process. Occurs when the laundered proceeds are distributed back to the criminal. Creates appearance of legitimate wealth.

2. PLACEMENT
Initial introduction of criminal proceeds into the stream of commerce Most vulnerable stage of money laundering process

3. LAYERING
Involves distancing the money from its criminal source: movements of $ into different accounts movements of money to different countries Increasingly difficult to detect

Money is laundered through

Banks Brokerage firms

Other Examples: Insurance companies, Money remitters, Cash intensive businesses, Brokerage firms, Realtors Crooked LAWYERS and ACCOUNTANTS

Financial services

Simple Bribe and Money Laundering Transaction

Country 3
Country 1

Company A Needs to generate $1 million for bribe to Finance Minister. Uses invoices from company in Country 2

Company owned by Ministers cousin

Company Bank Account

$500,000 - Purchase of Real Estate Country 4

$500,000 - Purchase of Bearer Share

Country 2

Money laundering, why do we care?

Is a global threat Is fuel to expand criminal enterprise Helps hide corrupt payments Uneven playing field for honest business Risks for financial systems & institutions-erodes integrity Regulatory
Reputational, credit and operational risk Market risk

Economic:
Deters private investment Destroys competition Revenue impact

Financial:
Perpetuates corruption, obstructs good governance Erodes confidence Destabilizes financial institutions

An evolutionary game theory approach to combat money laundering

From an evolutionary game between financial institutions and employees, it is possible to show that the decisions of banks and employees to cope against money laundering is determined endogenously

The two players in the game are employees and the Financial institutions

Contd..
Workers have two possible types or strategies.
committed to combat money laundering, being denoted h else l If they decide to combat money laundering, they face a cost C

financial institution that is committed to fight money laundering, denoted by F else I is the probability that an employee is being hired by F institution and 1- by I type be the out put of a bank operating legally and hiring a worker committed to combat money laundering be the out put of a bank operating legally and hiring a worker not committed to combat money laundering is the wage of a trained worker hired by a bank operating legally is the wage for an untrained worker hired by a bank operating legally

Contd
Let wi be the wage paid for both trained and untrained workers when banks decide to operate illegally be the cost for banks operating legally If the bank does not accept to follow this rule it is subject to be caught and pay a fee, expressed by m Let us assume that the probability of being caught is given by

Contd
The expected profit of the bank if it decides to cope against money laundering is given by

Where is the probability of hiring a worker that is committed with money laundering prevention If the bank decides not to cope against money laundering its expected pay-off is given by

 Bank that decides to operate legally has to incur in costs related to compliance which are denoted by In the illegal operation, the firm is also assumed to hire only one worker and also produces yi Let be the probability of getting caught and m be the pecuniary fine Accordingly, if the worker chooses the strategy of being trained to combat money laundering, then, her expected payoff is given by:

 For alternative strategy, the expected pay-off is given by:

A worker decides to cope with money laundering prevention if and only if Which means

 Representation of the game can be as below

The Evolutionary Game

According to the evolutionary approach the static game of the previous section may be evaluated through a dynamic game in which the players review their strategies in each period of time comparing their pay-offs with the average pay-off By using this approach, it is possible to introduce a mass behavior component in our model and emphasize the fact that the joint effort of banks, employees, and competent authorities is the key variable to combat money laundering

Contd..
N is the total no. of workers in financial system The worker choose one between following two strategies: cope up with money laundering or not to cope up with money laundering In this vein N=Nh+Nl Nh: number of workers who are willing with money laundering prevention N l: number of workers who do not cope with money laundering combat

L=number of banks Assuming each bank hires only one person for simplicity, L=N Let it L f denotes the number of banks that operate legally and L i the number of banks in the illegal operation. Hence, L=Lf+Li,

the number of workers who choose to cope against money laundering changes according to a comparison between the pay-off of choosing this strategy, namely U h , and the average pay-off of both strategies namely then the population of workers who decide to cope with money laundering prevention increases

Note that the crucial variables here are related to compliance costs, the probability of being caught and the ne to be paid in case of being caught in the illegal operation. If the expected loss due to punishment is higher than the compliance costs than banks decide to cope with money laundering prevention,

Case (i). If this loss is smaller than the compliance costs than banks decide not to combat money laundering,

Case (ii). If this cost is exactly equal to the compliance costs then some banks will decide to cope and others not

Case (iii). It is important to notice that, in this case, it is not possible to determine the fraction of banks that will cope and consequently the fraction that will not cope with money laundering prevention

Equilibrium states found

The five possible equilibria are (1,1),(1,0),(0,1),(0,0) and

Conclusions
the efficiency of anti-money laundering combat relies on the conjugation of factors such as a a) proper design of the anti-money laundering regulation b) Endogenous willingness of banks and workers to cope with this war Two way perspective: number of banks willing to ght money laundering affects the number of employees that also ght against money laundering number of banks that decide to cope against money laundering is also affected by the number of employees that are prepared or willing to ght it

Finally
level of combat against money laundering is an endogenous variable but may be strongly affected by design of the regulatory system

efficient combat to money laundering depends on the joint effort of competent authorities, banks and employees..

Notes
1. The insight behind these equations is straightforward: once a strategy has a higher t, measured by the its expected pay-off, than the average pay-off of possible strategies then the frequency of this strategy increases

2. Multiple equilibria can arise in a game when strategic complementarities are present, and spillovers exist between agents at the level of pay-offs