A famous game is found in the example of Prisoners Dilemma. In this game two people have committed a crime, have been captured and are now being interrogated by the police. The two players can choose one of two strategies. Either, they can cooperate (stay silent) or they can defect and tell the police that on top of the burglary, they had been planning on arson. The pay-off is – (time spent in prison), as they are trying to limit the extent of their incarceration. If both cooperate, they both get one year in prison. If both defect, they both get 10 years on prison, for burglary and attempted arson. However, if one defects and one cooperates, the person who defects gets to walk away free, for honesty, and the player who cooperates gets a longer prison sentence of 15 years, for withholding information. Here, the Nash Equilibrium is Defect, Defect as neither can improve their situation by cooperating individually, in this situation. So, the best strategy, regardless of the other player’s strategy, is to defect. However, the collective best strategy would be to both cooperate, as the collated prison sentence is shortest. So, prisoner’s dilemma proves individual rationality and group rationality are not necessarily equal.
Coordination games are another form of game, in which there are multiple pure strategy Nash Equilibria. Take the example of 2 friends, collaborating on buying PC’s. When they come to buying keyboards, they face a choice. They can buy a QWERTY keyboard, or a keyboard laid out in a fashion that makes it more efficient to type (call it Optimal keyboard). If the two buy different keyboards, they have a low pay-off, because it is inconvenient to buy different keyboards. If they both buy QWERTY keyboards, they have a higher pay-off, and if they both buy Optimal keyboard’s they have the highest pay-off, as the Optimal keyboard is the most efficient. So, here there are two Nash Equilibria, when they either both buy the QWERTY, or the Optimal keyboard. In these situations, neither can improve their pay-off by individually changing the keyboard they buy. So, you have two Nash Equilibria, one with a higher pay-off then the other. So this illustrates society can be trapped in a ‘bad’ equilibrium, of which it is difficult to escape because everyone needs to simultaneously be committed to shifting strategy to reach the ‘better’ equilibrium.
The presence of multiple Nash equilibria can lead to self-fulfilling prophecies. Take the advent of HD and Blu-ray DVD’s, which are incompatible with DVD players designed for the other DVD design. There are two Nash Equilibria, one where everyone uses HD, and one where everyone uses Blu-ray. The two compete for a larger market share. As more people switch to one form of DVD (say Blu-ray), others believe the ‘end is nigh’ for HD DVD’s, and they also switch to Blu-ray. This causes Blu-ray to become the dominant DVD, and a Nash Equilibrium is reached, based on people’s expectations- a self-fulfilling prophecy. Predicting which Nash Equilibria will prevail is difficult, as it depends on people’s expectations of what others are doing. In this case, people choose their strategy, to switch to Blu-ray, based on a belief others are doing likewise.
So, why and how do people come to play Nash equilibrium? Well, there are three main reasons. One is simply due to rational reasoning. Take the Prisoners Dilemma, for example. Via rational reasoning alone, a player can come to the conclusion that the best strategy is to defect. If both are rational, and come to the same conclusion that the best strategy, regardless of what the other prisoner does, is to defect, an equilibrium is reached. However, usually rationality is not, alone, sufficient to reach Nash Equilibrium. Correct beliefs of others behaviour is also vital. That the combination of rationality and correct beliefs of others behaviour is needed to reach Nash Equilibrium is a central concept.
To form these correct beliefs, a second way people come to play Nash Equilibrium is through pre-play communication. Players communicate beforehand and reach an agreement on how to behave. Because of the nature of Nash Equilibrium, this is a self-enforcing agreement. Neither has an incentive to individually deviate from the agreement, so the behaviour agreed on is guaranteed, and the Nash equilibrium (probably one of many possible depending on the individual situation) is reached. Take the example of battle of the sexes, another famous example. A man and woman plan to meet up, with a choice of two locations, a shopping mall and a football stadium. The pay-off for each situation is shown in the pay-off table of figure 1.
man |
woman | ||
Football | Shopping | ||
Football | 3,2 | 0,0 | |
Shopping | 0,0 | 2,3 |
Figure 1 – pay-off table in ‘Battle of the sexes’ example
As you can see, if they go to different locations, they both have a pay-off of 0, as they don’t meet. If they both go to the football, both are happy. The man is the happiest, as they are at his venue of choice. The opposite is true if both go to the Shopping Mall. Here, rational thinking is not sufficient to determine the best strategy, as there are multiple Nash Equilibria. So, they may arrange to meet, say in the Shopping Mall, the day before. This agreement, as the table shows, is self-enforcing. If either deviates from the agreement unilaterally, they don’t meet and they therefore gain a lower pay-off-they have no incentive to deviate.
The third way people come to play Nash Equilibrium is via trial and error adjustment. By accumulating experience, they gain knowledge of which strategies are good, or bad, allowing them to reach Nash Equilibrium.
Dynamic adjustment (another name for trial-and error adjustment) is the most common way in which Nash Equilibrium is reached. However, there is no guarantee the adjustment process always converges to reach a Nash Equilibrium. So, is it useful? Well, yes. This is because of the existence of Stylized facts, a stable mode of behaviour that is repeatedly observed. An example of a stylized fact is convention on escalators. If you are standing, you keep to the left. If you are in a hurry, you therefore have space to walk or run up the right side of the escalator. This forms a sort of Nash Equilibrium. If you defy the convention and stand on the right, your pay-off drops, as you are harassed by people who are looking to hurry past. Thus, the existence of stylized facts that act as a Nash Equilibrium illustrates why Nash Equilibrium is useful, in social sciences for example. Furthermore, it is useful to analyse behaviour in market competition.
All that has been mentioned before now involves pure strategy Nash Equilibria. However, there are also ‘mixed strategy’ equilibria. Take rock-paper-scissors. No matter what strategy you use, you have a 1/3 chance of winning. You have an equal average pay-off, no matter what your strategy. So, in this situation, the best strategy is to adopt a random strategy, and to be unpredictable. Thus, you reach the ‘mixed strategy’ equilibrium.
Fixed point theorem is another game theory concept relating to mixed strategy. Take the game of matching pennies. Players 1 and 2 simultaneously show 1 side of the coin. If both show the same side, player 1 wins. If they show different sides, player 2 wins. Thus you get a cycle that converges towards a fixed point, in terms of the probability of choosing heads. If Player 2 has a high probability of choosing heads, player 1 raises his probability of choosing head, so as to maximise the probability he wins. But if this occurs, player 2 will lower his probability of choosing heads, to maximise his chances of winning. Thus, slowly, the graph depicting of the probability of player 1 choosing heads, against the probability of player 2 dong likewise, converges to a fixed point, a Nash equilibrium where neither can improve their average pay-off by changing their probability of choosing a head individually.
Finally, Nash Equilibrium (possibly in mixed strategies) exists in all games, as long as two criteria are fulfilled; there are finitely many players and finitely many strategies. Of course, there are practically always going to be finite players. However, there may not always be finitely many strategies. Take the location game, where the vendor has infinitely different locations to set up shop. However, by dividing the street into finite slots, you can bypass this issue, approximating the equilibrium by a model of finite strategies. As such, almost all games fulfil the two criteria.