I try not to take leisure advice from computers. Their sense of fun generally doesn't mesh with mine. Calculating Pi to the thousandth decimal? Not my bag. Optimizing the cluster allocation of a hard drive? Not so fun. But, recently, I let a computer talk me into watching an entire season of TV about multiple agents working out competition and cooperation strategies in pursuit of a scarce incentive, and it turned out to be pretty entertaining.

The computers advising me were the ones who run the recommendation algorithms for Netflix. Logging on recently, I was surprised to see that they suggested Survivor on DVD. Survivor? Some TV shows make sense on DVD—a six-hour Sopranos bender is way preferable to following the story in weekly dribbles. Ditto Arrested Development. But Survivor? Why would anyone want to sit through a game show when they already know who wins?

Because Survivor is no ordinary game show. It's a competition, but it's impossible to win unless you can cooperate. Stripped of its tiki torches and castaway iconography, the show is about weighing consequences, and calculating how much something will benefit you or hurt your foes. I submit that this is the key to the show's longevity-- time has been kinder to Survivor than it has to Joe Millionaire because the former has some substance lurking under the surface. Survivor is, in essence, a televised exercise in game theory.

Game Theory arose as a field of study in 1944 when John von Neumann and Oskar Morgenstern published Theory of Games and Economic Behavior. John Nash, popularized in A Beautiful Mind, was another early pioneer. I like to think of Game Theory as the fun branch of Economics (although labor economics are always good for some yuks). After all, economics is concerned with how people make choices involving something that is scarce. Game Theory extends these ideas to situations in which multiple agents are competing for some incentive.

Sounds pretty dry, until you start subbing real words for the variables. If two teams of nine people compete for an incentive called RBIs, it's exciting enough to draw thousands of people into a teflon bubble to watch them. If the incentive is surviving a nuclear exchange, you've got enough warheads to end human life on Earth, sixty years of computer research devoted to crunching launch strategies, and a surprisingly decent Matthew Broderick movie. If the incentive is a million dollars, you've got a reality TV cash cow where pretend castaways backstab each other for our amusement.

(Survivor is hardly the only game show with a deeper structure worthy of analysis; Steven Levitt and Stephen Dubner's Freakanomics , a rambling tour through some novel applications of economics to the real world, includes an examination of The Weakest Link. The book is highly recommended, even if the game show is second-rate.)

Following the Netflix computers' advice, I chose the eighth season of Survivor, the All-Star edition wherein contestants from previous seasons were dumped onto an island off of Panama. Why settle for pedestrians when you can watch the masters in action?

Survivor All-Stars kicks off with a flurry of strategic pairings between veteran players who know the value of having an ally. One notable deviant from these match-ups was Survivor's first winner, Richard Hatch. Recognizing that all of his co-contestants had watched the show's first season, in which he schemed and ruthlessly plotted to remove competition (and periodically cackled with glee over his manipulations), Hatch knew that he was seen untrustworthy and poor alliance material. He launched a campaign of “crazy” behavior calculated to establish himself as a harmless clown. This approach-- consisting mainly of refusing to wear clothing and spouting non sequitors-- appeared to be an unorthodox-but-sound approach for a while. Eventually, though, performing a full-frontal naked grind on abrasive Wisconsinite Sue Hawks got him bounced and raised the possibility that maybe Hatch is more of sociopath than a deep strategic thinker.

As mentioned before, Survivor presents an interesting strategic situation because of the blended needs of cooperation and competition. To have a chance of winning the final prize, a player has to spend the early rounds proving themselves valuable to their team. This balance closely resembles one of the stock situations studied by game theorists, the Prisoners' Dilemma.

The Prisoners' Dilemma runs like this: two men have been arrested, and each is offered a deal by the police. If one agrees to testify against the other, and the other refuses to cooperate, the testifier goes free and the refuser serves ten years. If both testify, both serve two years. If both refuse to cooperate, both serve six months. Thus, you get your best outcome if you screw the other guy, but only if he declines to screw you. You could refuse to sell the other guy out, hoping to get the second-best outcome, but that leaves you exposed for the worst possible outcome depending on how the other guy chooses.

Something close to this pattern recurs as Survivor All-Stars winds on. It becomes clear that “Boston” Rob Mariano is playing the game on a level above everyone else, cornering swing voters before each week's Tribal Council, persuading them that he's on their side, and then leading a bloc to vote them off. In other words, he convinces the other guy not to cooperate with the police, and then sells him out for the best outcome. He may come off as a smarmy meathead, but Rob Mariano's reign of terror on Survivor All-Stars reveals him to be frighteningly good at applied game theory.

---

Von Neumann and Morgenstern developed Game Theory toward the end of World War II, and much of the early focus was militarily oriented. While the field has moved into the mainstream of economics and mathematics, national security applications still abound.

Consider the so-called “war on terror.” Despite vast disparities in resources and manpower, small groups like Al Qaeda are able to play something resembling an evenly matched game with a power on the scale of the United States and its allies, largely because of the strategic relationship between the two. Al Qaeda has an innate advantage because they're playing offense. The Department of Homeland Security (and its international counterparts) is forced to use resources to defend an essentially limitless range of potential targets; because of the nature of the game, they are essentially reactive. Al Qaeda, on the other hand, has the luxury of focusing resources on whatever they see as vulnerable. If American airline security has become impregnable, they shift to European rail transit. If railways get reactively improved, they move on to another target.

Or consider football, war's closest analog in the sporting world. One of the central truths of Game Theory is that if your opponent knows what you're going to do, you're screwed. So suppose you're the coach and your team has the ball. It's second down, three yards to go. Conventional wisdom says that you should run. But then your opponent expects you to run. So should you pass? Whatever you do, you shouldn't be consistent and predictable about it. (An interesting corollary to this is that the only strategy that your opponent can't suss out and counter is a truly random one. Suddenly Richard Hatch seems smarter.)

Minnesota Vikings fans have spent the past few years watching recently-fired coach Mike Tice piss all-over this principle. Tice began his tenure by announcing that he was bringing the team back to its core strengths; accordingly, he decreed that his playbook would revolve around receiver Randy Moss, to the tune of 20% of the offense. There's something to be said for getting back to basics, but the “Randy Ratio,” as Tice called it, announced to defensive coordinators around the league that one in five of his plays would be passes to a specific receiver. This took a bit of the challenge out of defending against the Vikings. Tice would have exactly zero chance of bringing home the million on Survivor, no matter how good he might be at spearing fish.

---

Looking at the underlying structure of a competition can even help you in your own home. My wife, our cat, and I recently found ourselves as multiple agents fighting over a simple but crucial incentive: sleep.

We wanted to sleep until seven every morning; Ella the cat was determined to eat at five, and adopted a strategy of meowing, knocking things off shelves, and generally raising hell to wake us up. While Ella lacked the ability to plan countermeasures to our moves, she made up for this shortcoming with sheer determination (risk-aversion and willingness to escalate are prime concerns of Game Theory). When we tried to quiet her by locking her in the basement, she countered by pounding on the door with her paws. When we rigged a cardboard barrier at the bottom of the basement stairs to keep her from the door, she spent hours every night picking at the tape (or velcro, or nails) holding the cardboard in place. No matter what measures we took, she defeated them and woke us up.

We may have been able to analyze the situation—but generally it just led us to work out how Ella would brute-force her way through whatever gimmick we came up with. We were being consistently outwitted by something with a brain the size of a walnut.

The impasse was broken when we looked at the terms of the game. Reviewing our failed stratagems, we realized that keeping Ella from the door was the wrong objective. We just needed to keep her from pounding on it. Several loops of contact paper were attached to the door, removing her ability to pound, and the larger-brained members of the household were immediately rewarded with vastly more peaceful sleep.

I'm not claiming that we were working on the level of von Neumann, or even Rob Mariano. But I bet we'd do better on Survivor than Mike Tice.