Someone recently bought our

students are currently browsing our notes.

X

Game Theory 2 Notes

Operational Research Notes > Operational Research Techniques Notes

This is an extract of our Game Theory 2 document, which we sell as part of our Operational Research Techniques Notes collection written by the top tier of LSE students.

The following is a more accessble plain text extract of the PDF sample above, taken from our Operational Research Techniques Notes. Due to the challenges of extracting text from PDFs, it will have odd formatting:

Lecture 7: Decision Theory / Game Theory II 19 November 2010

Topics

* Decision Making Under Uncertainty: When Nature's Probabilities are Estimated

* Game Theory for Non-Strictly Determined Two-Person Zero-Sum Games

Reading

* Taha (13.2 and 13.4)

Decision Making Under Uncertainty: When Nature's Probabilities are Estimated

* We may be sufficiently confident to assign specific subjective probabilities

* i.e. using method of relative frequencies of occurrence

* We would then have a probabilistic decision matrix

* This is not the same as using the Laplace criterion

* We could apply the "folding back" method but this is strictly for objective probabilities

* Two further methods for decision making under uncertainty:

* Expected Value Criterion

* Expected Value Adjusted for Risk Criterion Expected Value Criterion

* We evaluate the expected pay-off for each of the player's strategies
* Expected Pay-off:

* Perhaps the best criterion for making long run decisions

* Under achievement and over achievement counter balance each other
* Example of the expected value criterion Expected Value Adjusted for Risk Criterion

* A decision maker may wish to minimise the risk of getting a low value in addition to maximising the expected value

* i.e. they might have two objectives:

* To maximise the expected value

* To minimise the risk of getting too low a value

* One way to achieve this is to minimise the standard deviation

* We can combine this and the maximising expectation:
? Maximise

* Where k is an arbitrary (but suitable) non-negative constant

* The higher the value of k, the more the decision maker wants to avoid risk Example values of k

*

* Example of the expected value adjusted for risk criterion Game Theory for Non-Strictly Determined Two-Person Zero-Sum Games

* Saddle Point = A cell of the pay-off matrix which is the minimum in one direction and the maximum in the other direction

* A Strictly Determined Game = A game where there is a saddle-point

* The value of the game is the value in the saddle-point cell

* Not all two-person zero-sum games are strictly determined

* A zero-sum game is fair only if its value is zero Dominance

* Strictly Dominant = If every value in a row of the pay-off matrix is greater than the corresponding values of another row, the former strategy strictly dominates the latter one

* Dominant = If every value in a row of the pay-off matrix is greater to or equal to the corresponding values of another row
* Example demonstrating dominance Non-Strictly Determined Games and Mixed Strategies

* A mixed strategy = A combination of pure strategies in a certain proportion

* A mixed strategy can push up the values of the minimum that A can guarantee to win

* Conservative Strategy = A strategy where from all the available strategies it is the one where the smallest payoff A can receive, A receives the largest
* An example of a game with a mixed strategy solution

* Any pair of pure strategies will not be optimal and will not provide a stable solution if the two players are playing rationally An example demonstrating that any pair of pure strategies will not be optimal and will not provide a stable solution if the two players are playing rationally Finding Optimal Mixed Strategies

* We generate a pay-off table for the m x n game

* This in an LP problem:

* m+1 unknowns

* A's Problem:

* The solution that the value that B will pay-out will beCourse Notes Page 11

v:

Key Points

* Decision making under uncertainty

* Expected value criterion

* Expected value adjusted for risk criterion

* Saddle points

* Strictly determined games

* Dominance

* Non-strictly determined games

* Mixed strategies

* Finding optimal mixed strategies

* Non-strictly determined games when m=n=2

* Non-strictly determined games when m>2, n=2

* Non-strictly determined games when m=2, n>2

* Rugby post diagrams

Definitions

* A mixed strategy = A combination of pure strategies in a certain proportion

* A Strictly Determined Game = A game where there is a saddle-point

* Conservative Strategy = A strategy where from all the available strategies it is the one where the smallest payoff A can receive, A receives the largest

* Dominant = If every value in a row of the pay-off matrix is greater to or equal to the corresponding values of another row

* Saddle Point = A cell of the pay-off matrix which is the minimum in one direction and the maximum in the other direction

* Strictly Dominant = If every value in a row of the pay-off matrix is greater than the corresponding values of another row, the former strategy strictly dominates the latter one Formulae

* Expected Pay-off:

* Maximise

Buy the full version of these notes or essay plans and more in our Operational Research Techniques Notes.