UNIT 11 INTRODUCTION TO GAME THEORY AND ITS APPLICATIONS Objectives fier reading this unit, you should be able to + explain that decisions will have to be taken in a competitive environment * describe that the outcome of one business enterprise depends on what the competitor will do * discuss the conceptual framework on the strategies to be adopted in a competitive situation, * apply various methods to select and execute various optimal strategies to win the game. * use dominance rules to reduce the size of a game payoff matrix and compute value of the game with mixed strategies. + apply minimax and maximin principle to compute the value of the game, when there is a saddle point. Structure 11.1 Introduction 11.2. Definitions and explanation of some important ant terms 11.3 Saddle points 11.4 Dominance 11.5. Mixed strategies: Games without saddle points 11.6 2xngames 11.7 Exploiting an opponent's mistakes 11.8. Solved problems 11.9 Summary 11.10. Self-assessment Exercises 11.11 Further Readings 11.1 INTRODUCTION In general, the term ‘game’ refers (o a situation of conflict and competition in which two or more competitors (or participants) are involved in the decision-making process in anticipation of certain outcomes over a period of time. The competitors are referred to as players. A player may be an individual, individuals, or an organization. A few examples of competitive and conflicting decision environment, that involve the interaction between two or more competitors are: 223Resource Allocation Models, 224 * Pricing of products, where sale of any product is determined not only by its price but also by the price set by competitors for a similar product. * The success of any TV channel programme largely depends on what the competitors presence in the same time slot and the programme they are telecasting. * The success of'a business strategy depends on the policy of internal revenue service regarding the expenses that may be disallowed, © The success of an advertising/marketing campaign depends on various types of services offered to the customers. For academic interest, theory of games provides a series of mathematical models that may be useful in explaining interactive decision-making concepts, where two or more competitors are involved under conditions of conflict and competition, However, such models provide an opportunity to a competitor to evaluate not only his personal decision alternatives (courses of action), but also the evaluation of the competitor’s possible choices in order to win the game. Game theory came into existence in 20th Century. However, in 1944 John Von Neumann and Oscar Morgenstem published a book named Theory of Games and Economic Behavior, in which they discussed how businesses of all types may use this technique to determine the best strategies given a competitive business environment. The author's approach was based on the principle of “best out of the worst’ 11.2, DEFINITIONS AND EXPLANA SOME IMPORTANT TERMS ON OF A competitive situation is called a game if it has the following properties: ‘he competitive situation will be called a game, if it has the following properties: (i) There are a finite number of participants called players (ii) Each player has a finite number of strategies available to him.(iii) Every game results in an outcome Players: There are finite numbers of competitors called ‘players, Strategies: Each player has a finite number of possible courses of action called strategies, Minimax principle: Minimax is the maximum of row minima, Maximin principle: Maximin is the minimum of column maxima. Pay-off matrix: The figures shown as the outcomes of strategies in a matrix form are called the pay-off matrix Saddle point: A saddle point is a position in the pay of matrix, where the maximin value is equal to the minimax value. Value of the game: If the game has a saddle point, then the value of the cell at the saddle point is called the value of the game,‘Two-person zero-sum game: In a game with two players, if the gain of one Introduction to player is equal to the loss of another player, then that game is called a two- Game Theory and Its person zero-sum game. plications Types of Games Pure Strategy Games: If there is any saddle point (common point) for the given pay-off matrix, then it is called a pure strategy. In this pure strategy, both players adopt one strategy each. Mixed Strategy Games: If there is no saddle point (common point) for the given pay-off matrix, then it is called a mixed strategy. In this mixed strategy, both players adopt more than one strategy each. 2x2 Games without saddle point: Consider a 22 two zero-sum game without any saddle point, having the pay-off matrix for player A as yy a asa The optimum mixed strategies, 5, PL =P; +p2= 1 then (xi taa2)=(er2 a1) p=1- Py 2 +4:=1 then Activity 1 Give a comprehensive explanation of the term "game theory" Activity 2 Define with the help of an example, the following i) Players ii) Pay-off matrix 225Resource Allocation Models, 226 iii) Outcome Activity 3 Define the following; 1) Strategy 2) Pure strategy 3) Mixed strategy 11.3, SADDLE POINTS The, simplest type of game is one where the best strategies for both players are pure strategies. This is the case if, and only if, the pay-off matrix contains a saddle point. A saddle point is an element of then e4Tix which is both the smallest element in its row and the largest element in its column. We shall see why pure strategies the best strategies in this case by reference to Example 2 are. Activity 4 For the pay-off matrix given in Example 1, determine the saddle point. Example 1 : The pay-off matrix for a two person, zero-sum game below. Fine the best strategy for each player and the value of the game. Player B 1 M1 Iv v 1 0 0 5 3 Player IT 1 2 2 nit 3 0 2 6 IV 3 4 2 6 Solution: On examining each of his pure strategies Player A notes that ‘the’ worst outcome if he plays his course of action I throughout occurs when B also plays I and the resulting gain to A is -2. Similarly the worst outcomes when A plays II, III and IV are 1, -4 and -6 respectively. Player A, therefore, can guarantee a gain of at least I by using his courses of action IT throughout. Looking now at the game from the point of view of player B; his worst outcomes, taking each pure strategy in turn, are 5, 3; 1, 5, and 6 (N.B. As the entries in the matrix represent gains to A, the larger the entry the more unfavourable the result is toB) Player B can guarantee a loss of at most | by using his strategy TIT throughout. If A can guarantee a minimum gain of. 1, B cannot have a better strategy than one which precludes A from gaining more than 1, Similarly, if B can guarantee a maximum loss of |, A's best strategy is that which ensures that B loses no less than | The solution is, therefore; Player A uses his ‘course of action 11 throughout. Player B uses his course of action III throughout. The value of the game is 1 The solution is stable, in that when player A realises that B is playing 111 throughout he cannot afford to change his own strategy, 1 being the largest element in column 11.1. For similar reasons B has no incentive to change his initial strategy when he discovers how A is playing. These choices are completely ‘spy roof * in the sense that neither of the player can profit from, knowledge of the other's choice. It will be noticed from Example I that a saddle point occurs when the element which_is the largest of the row minima equals the element which is the smallest of the column maxima. The solution of such a game lies in player A choosing consistently that course of-action corresponding to the row through the saddle point and player B choosing always that course of action corresponding to the column through the saddle point. The element at the saddle point determines the value of the game. Example 2 : Determine the solution of the following game Player B I i UL I 2 15 2 Player A IL 5 6 4 ri 5 20 8 Solution: The row minima, corresponding to player A's courses of action I, TM and III, are -2, -6, and -8 respectively. The column maxima, corresponding to B's course os action are -2, 20 and -2. The largest of row minima and the smallest of the column maxima are both equal to -2. However, there are two elements whose value is -2 and hence the game has two saddle points. The solution to such a game is not unique: Player A uses his course of action 1 throughout. Player B may use either of his courses of action 1 or Ill throughout, or a nixed strategy which uses any combination of these courses of action. 11.4 DOMINANCE ‘We can sometimes reduce the size of a game's pay-off matrix by eliminating a course of action which is so inferior to another as never to be used. Such a course of action is said to be dominated by the other. The concept of dominance is especially useful for the evaluation of two-person zero-sum games where a saddle point does not exist. Introduction to Game Theory and Its Applications 227Resource Allocation Models, 228 Dominance property: In some games, it is possible to reduce the size of the pay-off matrix by eliminating rows(or columns) that are dominated by other rows (or columns)respeetively. Dominance property for rows: (x V (if Player B adopts strategy 2) pitpel pi p20 By dividing the above throughout by V, we obtain + Pe v + wm 21 v + Bowl v v Define x1 = Pad am = v v Therefore, we have. axi tex 1 bx: + dx22 1 mites v AAs Player A wants to maximise V, itis equivalent to minimising v ‘Therefore, the linear programming problem can be written as Maximize Z = V or Minimize W= | = x, ¢ x2 Subject to the following constraints ax $e 21 “ butdu 2! xi, 220 ‘The objective of Player B is to minimise its expected losses. ‘This can be reduced by minimising the value of Vi. it might Jose less than V if player A adopts @ poor strategy. Hence the expected loss for Player B will be as follows: aq + bqr SV (if Player A adopts strategy 1) eq: + dgs S V (if Player A adopts strategy 2) and qitqel qa S0 Dividing the above throughout by V, we obtain su el v v He Met v vResource Allocation Models, 234 eg fo v vv Define 1 2 i a sy and Bey v v Therefore we have, ays + by S 1 oy td yoy t Asminimising V is equivalentto maximizing | , the resulting linear programming problem can now be given as Minimize V or Maximise Fyity: ‘Subject to the constraints by:<1 ot dys yinye20 Please note that problem (A) is the dual of the problem (B). Therefore, the solution of the dual problem can be obtained from the optimal simplex table of primal Activity 6 For the problem given in Example 6, determine the optimal strategies for each Firm and the value of the game using Linear Programming Method. 11.6 GRAPHICAL METHOD FOR Mx2, OR 2xN GAMES Graphical method for 2xn Gam Step1: Reduce the size of the pay-off matrix of player A by applying dominance property, if it exists Step2: Let x be the probability of selection of alternative 1 by player A and 1- x be the probability of selection of alternative 2 by player A derive the expected gain function of player A with respect to each of the alternatives of player B. Step3: Find the value of the gain, when x=0 and x=1 Step4: Plot the gain function on a graph by assuming a suitable scale. keep x on x-axis and the gain on y-axis, StepS: Find the highest intersection point in the lower boundary of the graph —maximum point Step6: If the no of lines passing through the maximum point is only two, form a 2x2 pay-off matrix. - Go to step &; otherwise go to step 7 Step7: Identity any two lines with opposite slopes passing through that point. then form a 2x2 pay-off matrix.Step8: Solve 22 game using oddments and find the strategies for players A and B and the value of the game. Graphical method for mx2 Game: Step]: Reduce the size of the pay-off matrix of player A by applying dominance property, if it exists. Step2: Lety be the probability of selection of alternative 1 by player B and I- y be the probability of selection of alternative 2 by player B derive expected gain function of player B with respect to each of the alternatives of player A. Step3: Find the value of the gain, when y=0 and y=1 Step4: Plot the gain function on a graph by assuming a suitable scale. keep y on x-axis and the gain on y-axis. StepS: Find the lowest intersection point in the upper boundary of the graph minimum point Step6: If the no of lines passing through the minimax point is only two, form a 2x2 pay-off matrix. - Go to step 8 ; otherwise go to step 7 Step7: Identity any two lines with opposite slopes passing through that point. then form a 2x2 pay-off matrix. Step8: Solve the 2x2 game using oddments and find the strategies for player A and B and also value of the game. Example 7: Determine which course of action Player B will not use in the following game, Obtain the best strategies for both players and the value of the game. Player B I fi m Player A I 3 -1 7 bit 4 1 2 Solution: Since there is no saddle point, and since no course of action dominates any other, we consider each 2x2 sub-game in turn and find its value. As it is player B who has to make the decision of which course of action to choose, he will select the pair of courses of action corresponding to the 2x'2 sub-game whose value is least. Considering each 2x2 sub-game we obtain the following values: @) Player B I ii PlayerA I 3 =I 0 4 1 Value V1 = 1 Introduction to Game Theory and Its Applications 235Resource Allocation Models, 236 (by Player B I I Player A I 3 7 Ib 4 2 the value Vi = (73)(~2)-7*4 7A (-3-2) -(7+4) 8 Player B u Player A I 3 7 TL 4 2 11.7_ EXPLOITIN OPPONENT'S MISTAKES We have so far considered the best strategies for both players in a two person game. The definition of "best strategy” implies that the game is being played against a rational opponent whose object is to maximise his own gain. It is possible for a player A to take advantage of the knowledge that player B is not using his best strategy. Consider the following game of matching coins. Two players, A and B, each put down a coin. 11' the coins match, i.e. both are heads or both are tails, .A collects them both otherwise B collect them both. The pay-off matrix for this game is given below: Player B I (Heads) | 1 (Tails) (Heads [2 dt Player A | M(Tails) | -1 1 Player A's strategy of ([q2) is optimal against a shrewd opponent in that it protects him against loss. II', however, B is observed to play "heads" more frequently than “tails”, A can increase his gain by also playing "heads" more frequently than "tails". If, for example, B plays "heads" twice as often as "tails", i.e. a strategy of (2/3, 1/3) A increases his gain by also choosing for his strategy, (2/3, 1/3), The gain to A under these circumstances being. 2/3 (1/3) + 1/3 (-1/3), i.e. 1/9. It can be shown that if the opponent's strategy is, known in advance, a player achieves his maximum gain by employing a pure strategy. Thus in this example, A could increase his gain to one-third by playing heads throughout. Unfortunately, if A did this, B would almost certainly notice and be -led to modify his own strategy. The problem for A in pra exploit B's blunders without showing him the error of his ways. In the foregoing discussion, the concept of a number. of mixed strategy implies that a game is to be played a number of times. What of games that are playedonly onces? It is true that the implication has been that games are repeatable. Introduction to ‘The reason for this is that in a mixed strategy it is intuitively easier to imagine Game Theory and Its the probabilities of the various choices as the frequencies with which each Applications course of action would be chosen if the game were repeated many times. The meaning of the term "expected gain” is also more readily understood under this assumption, In fact, the results hold whether a game is played only once or repeated many times. If a game is played only once and the solution suggests a mixed strategy for a particular player, this implies that he must not in any event, disclose in "advance his proposed course of action. By using a chance device, the player ensures that he chooses a course of action that his opponent cannot anticipate, The probabilities given by his mixed strategy determine the characteristics of the chance device he should use, Activity 7 In managerial economics, you have studied price determination under a duopoly. Can you apply game theory in this problem if you are one of the duopolistic? (Assume you can know in advance the strategy of the opponent.) Example 8: A has two ammunition stores, one of which is twice as valuable as. the other. B is an attacker who can destroy an undefended store but he can only attack one of them. A can successfully defend only one of them. A leams that Bis about to attack one of the stores but does not know which. What should he do? Solution: The value of the smaller store is 1, the value of the larger store is 2. If both stores survive, A loses nothing. Ifonly the larger survives, A loses 1 If only the smaller survives; A loses 2 Since both parties have only two p is as follows. ible courses of action, A's pay-off matrix B Attack Attack smaller store larger store Defend 0 2 smaller store A Defend ll 0 Larger store Before analyzing this problem as a game, let us consider how A and B might otherwise reason in a real situation. A would argue that he cannot afford to take the risk of losing the larger ammunition store and therefore, he must defend this 237Resource Allocation Models, 238 one, -if B is shrewd, however, he would anticipate A's reasoning and attack the smaller. store with a certain gain of 1. Thus A's "logic" would result in a gain of 1. The problem may be solved by the methods outlined earlier. A's best strategy is the mixed strategy (4/3, 2/3) and the value of the game is -2 3. In order to make his decision, therefore, A should roll a die. He should defend the smaller store ifa 1 or a2 tums up; otherwise, he should defend the larger one. This strategy leads to a more favourable expected outcome than the "logic" mentioned above. Having illustrated how a "one-off" decision should be reached, we know that decision in practice is not made in the above manner. An executive would probably not inspire confidence in his ability if it were known that he made his decision by selecting a number at random; and what manager would ascribe to the toss of a chin the credit for a decision which proved successful-—-or catastrophic? There is an important difference between games and many of the decisions that have to be made in practice. By definition, in a game, nothing is known of an opponent's strategy, save that he is an extremely shrewd player. In a decision-making situation, this is not generally so. Even where there is a component the decision is often based on some knowledge of his likely strategy. 11.8 SOLVED PROBLEMS Problem 1: Solve the game whose pay ~ off matrix is given by, Player B B, B2. B3 Alf 3 1 PlytA A2| 00 4 3 A3 [1 at Solution: Player B Bi B2 B3 Rowminimum Al 1 3 1 1 PlyrA A2| 0 4 3 4 Aa] 1 soa 1 Column 1 3 1 maximum, Maxi(minimum) = Max (1,4, -1)= 1 Mini (Maximum) = Mini (1,5,1) Saddle point exists. The value of the game is the saddle point, which is 1. The optimal strategy is the position of the saddle point. Problem 1: Solve the game whose pay ~ off matrix is given by,Player B Introduction to Game Theory and Its Bl B2 B3 B4 BS Applications ail [2 0 3 3 PlayerA =A2 | 3 2 1 2 2 A3 4 3 0 2 6 Ad 3. 4 2 + Solution: Player B Bl B2 B3 B4 BS Rowmini al [2.0 0 3 3 2 PlayerrA A2 | 3 2 1 2 2 1 AZ 4 3 0 2 6 4 Aa | 5 3. 4 2 + 6 Column 5 3 1 3 6 minimum Maxi(minimum) = Max (-2,1, -4, -6) = 1 Mini (Maximum) = Mini (5,3, 1,5,6) = 1 Saddle point exists. The value of the game is the saddle point, which is 1, The optimal strategy is the position of the saddle point. Problems 3: Solve the following pay -off matrix. Also determine the optimal strategies and value of the game. Player B PlayerA AA 5 1 11.9 SUMMARY Atthe outset, a very comprehensive introduction to the theory of games has been provided with a clear conceptual framework and definition of important terms used in games. The pay-off matrix has been explained through an example. 239Resource Allocation Models, 240 The next point in discussion has been the concept of saddle point-an element of the matrix which is both the smallest element in its row and the largest element in its column, We have also mentioned that saddle point occurs in the case of pure strategies adopted by both the -players. Examples have worked out to illustrate the method. Mixed strategies have been covered as situations where the saddle points do not exist. The concept is well brought out through an example. The algebraic solution giving the mixed strategies for the players, and the value of the game has been derived so as to enable you to use the formulds for a given problem. The principle of dominance and its -importance in reducing the size ofa game's pay-off matrix has been dealt with through an example. 2xn game has been the next topic of discussion. In this, one player has only two options while the other has more than two courses of action. The solution of such games involves identifying combinations of 2x2 sub-games within which lies the solution of the larger game. One numerical example has been worked out to bring the essence. Exploiting the opponent's mistake is very crucial-in business involving competition where the rival firms can thrive on their opponent not using his best strategy. The relevant aspects have been succinctly explained through an example. 11.10 SELF-ASSESSMENT EXERCISES 1. Solve the following zero-sum two-person games, Obtain the best strategies for both players and the value of the games: Player B 1 0 PlayerA 1 ° u 0 2 Player B (sees sea Vey, ' af ell fda ct n of] 3 ais Player A IIL af2fafa]a wf ws} ad - o | 5 v of of 6 | o | o Determine the optimal strategies fr each firm and the value ofthe game. 2. Two players A and B match the coins If the cons match then A wins. ‘otherwise he loses. Detrite the optimum sirateis forthe plajers and the value ofthe game.3. “Game theory provides a tematic uantatvé approach for analysing competitive situations in wich the competitors make wse of logical proces and techniques inorder to determine an optimal strategy for winning" Comment. 44. Describe the strengths and limitations of game theory 5. Two companies A and B ae competing forthe same produt: Ther diferent strategies are giverin the following payoff mari: Bl. B Al Company AR Determine the best strategies and find the vale ofthe game. 5. Reduce each ofthe fllowing games by using the rule of dominance and then solve the reduced game by any of the method you have studied @ Bho “AUR | arsine See o BOBS | Oy ae - 2 a |g iy ey 4 Bl] mw 6 8&8 Bw 6 7. The following is the pay-off table for a zero-sum two-person game, with the pay-off being the amounts player B losses to player A Player B I n Player A 4 3 2 2 Determine the best strategies and find the value of the game. 11.11 FURTHER READINGS Sasieni, M., Yaspan, A., and Friedman, L. (1959)-Operations Research-Methods and Problems. John Wiley & Sons, Inc., New York. Baumol, W.J. 2nd Edition, 1965)-Economic Theory & Operations Analysis Prentice Hall, Inc., Englewood Cliffs, NJ. Gupta MP. & LK. Sharma-Operations Research for management, (2nd edition 1987), National Publishing House, New Delhi. Introduction to Game Theory and Its Applications 241