Class #12 Highlights

In this class, we learned about the theory of achieving coordination in repeated games. The idea is to combine carrots (promises of rewards for good behavior) with sticks (threats of punishment for bad behavior) to obtain cooperation.

This required: 

1. A payoff difference from the future consequences of bad versus good behavior. 

We saw that, when the game is only repeated twice or a small number of times, it was not possible to make credible promises to reward niceness in the last round of the game. As a consequence, good behavior breaks down. The lesson is that, as the endgame draws near, cooperation fails. 

2. Timely detection of bad behavior

We saw that, when an individual could get away with bad behavior for a couple of periods, sustaining cooperation became much more difficult. Likewise, delaying punishment also makes cooperation more difficult.

3. Proportional/Credible Punishments

If a punishment is so outrageous that the individual won't follow through, then it serves no role in getting good behavior. It must be credible that the punishment will be meted out. 

4. Forgiveness

Mistakes happen. An unforgiving punishment scheme suffers from two problems. It is disproportionate, so has credibility problems. It is also problematic since it cannot recover from errors. 

Together, these four keys allow for leveraging the future to achieve cooperation.

Revised Schedule

We are now officially one full class behind schedule. Accordingly, I have updated the syllabus to reflect the amended schedule. One topic, Principles of Bargaining, has been deleted and Judo has been moved into its slot. Please check the course website for the amended schedule. In terms of the near term roadmap:

Monday: GE v Westinghouse + OPEC Summit
Wednesday: Browser Wars

Following week: Mid Semester projects

OPEC Round 2 - Do Over

Dear Gamers,

Here are the results for round 2 (the do-over):

Market A: Price = 98.47, Quantity = 74,255
Market B: Price = 99.62, Quantity = 75,272

Round 3 due by Friday Wednesday at 1159pm. (Thanks for Billy Hwan for the alert correction.)

Class #11 Highlights

In class #11, we talked about building mental models. The case offers several frameworks that build up to form the "right" game to analyze. The first step, of course, is to gather relevant data. This is data not just about the financial aspects of a decision, but about the behavior of rivals as well. Competitor Analysis takes this data and uses it to paint a "portrait" of the rival. Sketching out assumptions and goals helps to determine which aspects of the payoff matrix on which to place weight. For example, Jerry Yang's founder status indicated that he highly valued control of Yahoo over other financial considerations. By studying current strategy and capabilities, one obtains a list of possible responses to your actions. For instance, if you are a startup contemplating a business model is some aspect of the social/mobile space, understanding the capabilities of Facebook is obviously essential in developing a non-threatening business strategy where their large user base will not be deployed against you. 

After completing a competitor analysis, it is useful to undertake a behavioral analysis. Are there ways that you can change the frame of a negotiation to get what you want? How does the positioning of the status quo affect the reaction to various negotiated outcomes. It is also useful to apply this same analysis to yourself. Is your data gathering subject to confirmatory bias? Are you overconfident in determining the appropriate scenario analysis to perform?

These behavioral factors can influence the perception of payoffs as well as the types of strategies your rival is likely to pursue. This is critical to building a mental model. Finally, game theory puts all of these analyses together to formulate what game is being played, what strategies are available and how payoffs are evaluated. As we saw in the Coors case, sometimes it is enough to know the relative payoff rankings, and these analyses can be especially useful in informing this aspect of payoffs. 

Together, these steps constitute the basic framework for assessing a rival's response to a new product introduction, a change in positioning, a new pricing strategy, and so on.

OPEC Do-Over

Dear Gamers,

I found the error in the simulator. Basically, the sum of all the OPEC quantities was coded as fixed rather than as reflecting the actual sum.

Here is the proposed fix. Since round 1 numbers were submitted legitimately, I reran the simulator with those quantities. The revised price results are

Market A: Price = $92.31, Quantity = 71,525
Market B: Price = $95.66, Quantity = 70,393

We will redo round 2 tonight at midnight. All subsequent rounds are the usual Mon, Wed, Fri at midnight schedule.

Sorry for the confusion.

OPEC Round 2

Round 2 results:

Market A: Price = 160.68, Quantity = 61, 576
Market B: Price = 165.80, Quantity = 62, 712

Market B enjoys a boom in demand while amenic demand plagues market A. Round 3 due by Monday at 1159PM.

OPEC Round One

The round one results are now available.

Goals of OPEC

I've been asked in various ways what your goals should be in the OPEC game. Here's my view of "best practices" in regard to OPEC and all the games we do in the class:

Your objective should be to make as much money as possible using realistic strategies. In OPEC, that means coming up with creative solutions to cooperation that would be workable in the REAL WORLD context of international competition. Cheesy workarounds that violate the spirit but not the letter of the rules I set out may help you to achieve the monetary objective, but will not provide you with usable lessons in a business context.  

If you're uncertain about the legitimacy of a given practice, you should ask yourself whether whatever tactic you have in mind is at all sensible in the real world version of the environment being simulated. If it is not, then you should refrain from that practice.  While your narrow goal is to make money, the broader pedagogical goal of the experiments is to give you the opportunity to try out various tactics and hone skills useful in the real world. Tricks that take advantage of the fact that it is a simulation, rather than reality, may net you a short-run monetary gain don't help you a bit in becoming a better strategist/tactician/negotiator/leader in the real world. Presumably the reason you took the class is to become a better leader and not to learn how to "game" a simulation. 

OPEC Clarifications

Dear Gamers,

Two important clarifications for the OPEC game:




1.  Your colleague, Rodrigo Donoso, brought to my attention a critical typo in the OPEC Data sheet. The total reserves for each country is off by a factor of ten. In other words, the Saudi total reserves should be 528,000 rather than 52,800. I have now updated the spreadsheet to reflect this fact. 


In your calculations, the total reserves should not be binding unless the game lasts for a really long time. For instance, the Saudis will not run through their capacity, even if they are producing all out, in fewer than 44 turns. 


2. All production units are in 000s of barrels. Thus, if Saudi produces 12,000 in period 1 and the world oil price is $100, then it earns revenues of $1,092 million. 


Let me know if you have any additional questions. 

Class #6 Highlights

In class #6, we studied the question of how to value the assets of a company. This is a fundamental question. Indeed, acquisitions and divestments are the two most significant decisions any CEO can undertake in guiding firm strategy. Conventional inward thinking suggests that we study the cash flows generated by the asset, the sale value, the replacement cost, or other such metrics in valuing an asset. Our game theory approach was to apply the "it's a wonderful life" rule to determine the strategic value of an asset. 

Under this rule, we study the value of the company with an without the asset. In the case, we showed situations where this analysis leads to the conclusion that an asset can have a negative value. In the case, having excess capacity prevented the smaller firm from undertaking a price cutting strategy without provoking a price war. By shedding the asset, the smaller firm became less threatening and unilateral price cuts were more likely to be tolerated by the larger firm. So long as the small firm had enough remaining capacity to add share, these price cuts were, in fact, profitable. 

The shedding of assets to seem less threatening to a rival is called the puppy dog ploy. A key part of firm strategy is calculating where the breakeven point is in terms of share grabbing that avoids provoking a costly price war. Even though one might think that excess capacity provides option value, in some circumstances, it can actually reduce a firm's competitive options and thereby destroy value. 

Class 9 Highlights

In class #9, we concluded our unit on dominance with a couple of observations.

1. In auction settings, the only way to get truth-telling and efficiency in equilibrium is to apply Vickrey's Law. No other auctions will work. For example, If we use the rule that in the A,B setting, the winner of A simply pays the amount of the losing bid, then bidding above your value can be profitable. Likewise, in the multi-unit auction setting, using a uniform price auction will also not work in that a bidder's own bid can determine the price of units won. Therefore, bidders have an incentive to bid below their value for units further down the demand curve.

One practical implication of this second finding relates to auction IPOs. Most auction IPOs, such as those run through Hambrecht & Co. use the uniform price auction. For bidders seeking small numbers of shares, bidding is approximately truthful, but large institutional investors seeking many shares will have incentives to shade down bids. Thus, this auction does not completely do away with the "pop" on initial trading as the bid determining the price probably reflects a discount below fundamentals owing to these incentives.

The highlight of the second half of the class was a solution to the teams problem. We first noticed that, in a conventional teams problem, free-riding was an important consideration. In effect, the game was a prisoner's dilemma. Efficiency wages, i.e. paying wages above the market rate, can help to some extent by activating reciprocal motives on the part of employees. For years, this was IBM's strategy. It has also been routinely demonstrated in laboratory settings.

Incentives based solutions include profit sharing and bonus schemes. We saw that unless effort was extremely profitable relative to the personal cost to the employee, profit sharing schemes suffer from the same free rider problem as fixed wage schemes. Bonus schemes transform the problem into a stag hunt. The riskiness of the stag hunt depends on the aggressiveness of the target. The more aggressive the profit target, the more fragile the trust relationship and the less the likelihood of success. As a practical matter, a firm that strives to get more out of its employees may end up getting less than a firm that sets more modest goals. It is important to consider this tradeoff between trust and profits in determining appropriate incentives for a company.

As a practical example, some pundits have noted that the emphasis on "accountability" and ambitious short-term targets at Yahoo led to a culture where cooperation and trust across properties was diminished rather than strengthened. In stag hunt, one can think of the hare strategy as one of "silos" within an organization. Each group pursues its own goals rather than pursuing synergistic opportunities between groups. The lesson from game theory is that, accountability in the form of high powered incentives and severe punishments for missing targets creates exactly the conditions for silos to become the correct best response by managers in the firm. That is, we can use game theory to see how incentives can affect culture--sometimes in unexpected and undesirable ways.

Game Theory Class #5 Highlights

In this class, we studied the timing principle. In games where there is no private information about payoffs, a player moving first can always do at least as well as going at the same time. Moreover, in many situations, moving first can improve payoffs. The reason secrecy doesn't pay is that the only possible secret is your strategy. In equilibrium, individuals are engaging in mind reading, so even this is no longer a secret. When there are no real secrets, there is nothing to be gained by keeping moves hidden. Transparency, on the other hand, can shape the rival's strategic response. This option value is potentially useful. As we saw in the McCain-Schumer experiment, it was quite useful for the first mover. 

We also observed that the right first move depended on the competitive position of the first mover relative to the second mover. When an aggressive move from the first-mover provokes a retreat by the second, then being aggressive is optimal. On the other hand, when aggression is met with aggression (as in the case when the first mover has lower value than the second), then the right strategy is to ratchet down competition. This is referred to as the favorite and underdog effects, respectively. 

Class 7 and 8 Highlights

In these classes, we studied dominance. Recall that a dominated strategy is one where there is some other strategy that does at least as well (and sometimes strictly better) regardless of the strategy chosen by the rival. 

One can use this concept repeatedly to "dominance solve" some games, but each round of dominance requires ever greater levels of common knowledge of rationality. For instance, the equilibrium for the beauty contest can be solved using infinite rounds of iteration of dominated strategies. But, as you saw earlier, that prediction fares poorly. A good rule of thumb is the deletions up to about two rounds are same, but not thereafter. 

We then applied this concept to Vickrey auctions. These are auctions satisfying Vickrey's law: You pay the amount of the externality you inflict on others. This is a version of the It's a Wonderful Life Principle. First, compute the values to all other players if you were not present. Then compute their values when you are. If you pay this amount in winning an item or items, then it is a dominant strategy to bid truthfully. This is a powerful insight for designing proper incentives. 

In practice, there are two limitations: 1. It doesn't work well when there are a large number of options with synergies between them. 2. It doesn't work well when fairness is an important consideration. For instance, we saw how Vickrey auctions can lead to situations where the high bidder pays less (or even nothing) for an item while a lower bidder ends up paying more for the same item. 

Game Theory P&L

Please fill in your P&L from McCain-Schumer and Spectrum Auctions in the P&L Spreadsheet located on the Game Theory website. The results from both games are posted there as well next to the relevant experiment in the syllabus. In McCain-Schumer, your score in ecus is your raw score divided by 1 million. The same is true of spectrum. Thus, if you earned $110 million in Spectrum, this counts for 110 ecus.

Thanks!

Sports Geenius

Just a reminder to be a part of the experiments on the structure of incentives in tournaments. Please visit the site:

http://fantasyallnews.com/SportsGeenius/

and sign in using your email and the password I provided in class. This is a good chance to make decisions under contest incentives as well as to earn a little money. Any feedback you might have is also appreciated.

Please note: SportsGeenius is open Mon, Weds, and Fri. It is closed on other days. This is purely for testing purposes. The final product is intended to be open every day.