TCS Daily

When Lady Luck Plays Moneyball

By David R. Henderson & Charles L. Hooper - October 20, 2006 12:00 AM

Most sports fans and sports analysts, for all their hours examining their teams, are fundamentally wrong about one important aspect of sports. Most of these spectators decide which teams are good or bad right now based on their winning and losing streaks. They shouldn't.

How often have you seen sports analysts castigate a team for losing four straight games or a player with a .300 batting average for going hitless for five games? It's pretty frequent, right? But this frequency alone should tell you something. The fact is that the best teams and the best players have slumps. And, to take the baseball season as an example, in a 162-game schedule, the best teams and players are even likely to have more than one slump per season. These slumps aren't necessarily due to the team or the player doing anything different or wrong. Rather, they're based on the laws of probability. To put it in slightly jargony terms, "there's much randomness in the world." And randomness doesn't make an exception for sports.

The best professional baseball teams, the ones that make the playoffs, generally win about 60 percent of their games. Using probability theory and Monte Carlo simulations, we've proved that the probability of a team that wins 60 percent of its 162 games having at least one losing streak of five or more games is 80 percent. Such a team will have, on average, 1.2 such losing streaks a season. In other words, it's almost a certainty that a playoff team will have had at least one substantial losing streak during the regular season. So it shouldn't have been shocking -- and was hardly informative about their future -- that the Detroit Tigers ended the season on a five-game losing streak. The worst teams, in comparison, generally win only 40 percent of their games. (For example, in the 2006 baseball season, Tampa Bay and Kansas City won 37.7 and 38.3 percent of their games, respectively.) Using probability theory and Monte Carlo techniques, we've shown that the probability of such a team having at least one winning streak of three or more games is virtually 100 percent. Such a team will have, on average, 6.4 such winning streaks a season. In other words, even the worst teams can expect to have winning streaks. That's why it shouldn't have been so shocking that the Detroit Tigers, with the third-best American League regular-season record, ended the season by losing all their games against the Kansas City Royals, the second-weakest AL team.

It seems that only a few people in sports get this basic point. How else can one account for their surprise after the St. Louis Cardinals had an eight-game losing streak at the end of the season and, yet, went on to demolish the San Diego Padres in the first round of the National League playoffs? Sports commentator Joe Morgan, who often talks about the importance of momentum, surely doesn't get it. Morgan almost invariably explains how he thinks baseball teams will do in the next game based on how they did in the previous few games. Now there could be such a thing as momentum. But simply probability theory can explain many of the streaks we observe in baseball. What Joe Morgan and others should really say is, "Team A has been lucky during its last four outings, but it is still a weak team, and so we shouldn't expect it to be lucky today."

One person who distinctly does get the point is Oakland A's General Manager Billy Beane. Beane, whom we discuss in our book Making Great Decisions, recognizes that his job is to get his team to the playoffs, where whoever wins has more to do with luck than skill. As Michael Lewis wrote in Moneyball:

Pete Palmer, the sabermetrician and author of The Hidden Game of Baseball, once calculated that the average difference in baseball due to skill is about one run a game, while the average difference due to luck is about four runs a game. Over a long season the luck evens out, and the skill shines through. But in a series of three out of five, or even four out of seven, anything can happen.

As Billy Beane put it, when asked by Lewis why he was so detached during the 2002 playoffs, "My s**t doesn't work in the play-offs. My job is to get us to the play-offs. What happens after that is f***ing luck."

So what does this have to do with business and with life in general? A lot. There's a huge difference between decisions and outcomes. Decisions can be good even if bad outcomes follow or bad even if good outcomes follow, all because of the role of chance and luck. Luck and probabilities have a huge role. Here's an example from our Making Great Decisions in Business and Life.

Imagine that you own two uranium processing plants, and you believe that good decisions should be measured solely by the good outcomes they produce. The plants generate the same revenues, but one plant (Plant A) has 30 percent lower labor costs than the other (Plant B). Plant A certainly is enjoying a better outcome -- higher profits, quarter after quarter. For the last six years, Plant A's outcomes have been better than Plant B's by any objective measure. So it follows that Plant A's managers have made better decisions, right? How could anyone argue with this? If you had a bonus pool to distribute, would you give more to the Plant A's or Plant B's managers and workers?

Before you hand out the bonuses, disaster strikes. Uranium processing Plant A suffers a horrible accident, killing four workers and nearly releasing enormous amounts of toxic radiation into a nearby community. Lawsuits bury your company. How could this happen? Simple. Plant A workers skipped many safety procedures in their quest to reduce labor costs, increasing their risk of disaster from infinitesimal to perhaps one accident every 100 months. Each day, they were able to speed their work by skipping burdensome and "unnecessary" safety procedures.

The Plant A managers were playing Russian roulette, yet you were about to reward them for their good outcomes during the last six years. Had you studied their decisions instead of their outcomes, you would have realized that they were making horrible tradeoffs to achieve their objectives. The question is: which processing plant made the best decisions? The best decisions are based on the cost of adhering to the safety procedures versus the expected cost -- or risk -- of skipping them. Managers who base their rewards purely on outcomes will unwittingly encourage risky behavior that focuses on the short term at the expense of the long term.

This example, unfortunately, is not fictitious. In 1999, workers at the JCO Co. uranium processing plant in Tokaimura, Japan didn't follow proper procedures and mixed too much uranium -- 16 kilograms instead of the approved 2.4 -- with nitric acid in a storage tank and started a fission reaction that went temporarily out of control. All 310,000 residents in the city were evacuated, 21 people were sickened, and three workers were hospitalized. While this may have been an innocent accident, it probably was the result of a technique the JCO workers used to improve their efficiency.

Luck is like a great wind that blows randomly. It camouflages people's decisions and actions. Our job, if we want to make good, or even great, decisions, is to look past the wind that blows today for intrinsic quality and give credit where credit is due. Sometimes we can learn important lessons from sports like baseball. And we don't even have to be lucky, just clear-thinking enough to pay attention.

David R. Henderson is an associate professor of economics at the Naval Postgraduate School in Monterey, California and a research fellow with the Hoover Institution. Charles L. Hooper is president of Objective Insights, Inc. and a visiting fellow with the Hoover Institution. Their book is Making Great Decisions in Business and Life (Chicago Park Press, 2006).


The Cubbies
What about the Chicago Cubs? How do you explain them? They've had so much bad luck over the past century that they could win the next five World Series straight and it wouldn't even it out.

Sports and Medicine
This is exactly the way I encourage medical students and residents to think about medical decision-making. They are often surprised to hear me say that I'd rather make a good decision and have a bad outcome then make a bad decision and have a good outcome.

The reason I feel this way is that I know that good decisions will save more patients in the long run than bad decisions. An occasional bad outcome despite a good decision should never discourage us from making the same good decision in the future.

This is such a difficult concept to teach. Hopefully, your article will make it easier in the future. I'm going to encourage my students and residents to read it.



Maybe they're just bad?
The bad "luck" you postulate may instead be a manifestation of the fact that they're just consistently BAD.


Good stuff until here:

"Luck is like a great wind that blows randomly. It camouflages people's decisions and actions."

What camouglages people's decisions and actions are desired outcomes, not luck and not decisions and actions. When managers reward desired outcomes (success) regardless of unfunded risks rather than diligence, prudence and loyalty, they tempt luck to unmask their own ignorance, foolhardiness and disloyalty.

Simply put, the measure of folly isn't luck; it's wishful thinking.

Desired Outcomes...
There is nothing inherently wrong with rewarding desired outcomes. The authors appear to say so, but what is really the problem is the failure to properly identify the desired outcome.

In the listed example, the desired outcome of running a nuke plant is not just profits and low costs. It is also a good saftey record. Thus, as the authors state, an important consideration is also whether safety procedures are properly followed. The danger is in determining that a "safe" plant is one that has never had an accident. This is not necessarily so. The safe plant is one where established safety proceedure are followed, and constantly evaluated for improvement. Even a safe plant might have an accident (for example, due to a earthquake, meteor, or some unforseen and unknowable risk).

Thus, it is important to identify the correct performance measurements in order to evaluate a system.

Furthermore, ignoring ACTUAL outcomes is also a danger. Bureaucracies are notorious for forcing rule-following, but ignoring that the rules may be outdated, or fail to achieve their intended results.

The key is that the process must lead to the desired outcomes. Thus, the process must be followed, but constantly improved and tweaked, and sometimes even radically changed. The process should be enforced (because is should be the best that we have), but maleable. But it is also important that the process not be tweaked at the whim of the individual--it must be enforced until a change is required based on an identification of an erroneous match with the desired outcomes.

Finally, the outcomes being monitored must also be regularly evaluated. Choosing the wrong outcomes upon which the process is tailored can lead to disaster.

In the end, the keys are (1) to have a defined process, (2) to enforce the process, (3) to have defined desired outcomes (objectives), (4) to monitor whether the process is achieving the desired outcome, and (5) modifying the process when it does not lead the desired outcome, and (6) To constantly question whether the desired outcomes are the right ones.

It is (6) that I think most organizations fail to do adequately, although many fail at any or all of them.


I agree completely
The attitude, "Profits at all costs!" invites disaster because it underfunds prudence in return for *pop*!

A culture that asks, "What have you done for me lately?" must also be a culture that demands a scapegoat for every misfortune. This is our culture; America's culture.

Sports fans aren't stupid
Sports fans do not overestimate the importance of a winning or losing streak in a long season.

Go to and look at the probability of the Tigers winning the World Series. It stayed around 10% throughout the end of season losing streak. (Admittedly with a thin market)

Sports reporters and announcers talk about momentum because they get paid for talking.

I think you will find the same thing in the political betting market and punditry. The markets are relatively stable, but pundits are all over the place - again, they get paid for talking. Look at the market for House GOP control - a steady decline, with a sharp drop when the IM scandal came out.

I don't want to brag, but my understanding of markets turned a $250 Tradesports initial account into $1.57 in just a few months.

TCS Daily Archives