Anatomy of a Game: Survivor!

by Shannon Appelcline

May 23, 2002 – It's been two years now since the first episode of Survivor came on the air. At first I — like many people in my circle — dismissed it fairly out-of-hand. Reality TV. Pfah! But, after my wife decided to watch the fourth or so episode of Survivor I, I realized there was more to the show than just sheer voyeurism.

At heart, the show is about sixteen people coming together in a remote location and voting each other out one by one, until only one remains. As a writer I was intrigued almost at once by the larger-than-life characters — and how they would react to their changing situation. On the other hand, as a game designer I was fascinated by the complex interactions created by a very simple set of game rules.

With the season finale of Survivor IV now a few days in the past, I feel like I really have enough information to finally talk about the game design of Survivor in depth, and so that's what I'll be doing this week — discussing a specific game (Survivor) and how its game design influences its gameplay.

The Game Is What the Game Is

For those of you who have been living in a cave, I should probably take a moment to explain the basic game design of Survivor. There have been some quirks and changes from season to season, and there are some additional details that don't affect the dynamics that much, but overall it goes something like this:

16 people are stranded in a remote locale. They are initially divided into two teams (tribes) of 8 members each. These two tribes are separated from one other and each begins to form its own community out in the wild.

However, every three days there is a Tribal Council which one tribe must attend in order to vote out, by simple majority, one of the members of their tribe. Which tribe must attend Tribal Council is determined by an immunity challenge. It's most frequently physical, though a few mental or creative challenges are included on occasion. In every challenge, one tribe wins, one tribe loses, and the losers then decide which of their members must leave.

When the total number of survivors gets down to 10, everything changes. The two tribes merge into one, and the immunity challenges become singular. Rather than a tribe winning immunity, an individual does. The tribal councils continue every three days, and this time only the one person with immunity is safe.

There's a minor quirk when the total number of survivors reaches 9. At that point all additional members voted out of the tribe join the "jury". We'll get to them in a second.

When there are only two survivors left they can no longer vote a member out of the tribe — because each would vote for the other. That's where the jury comes in. At the final tribal council each of the 7 jury members votes for who they want to win the game; the person with the most votes becomes the winner.

The only other rule of any relevance is the tie breaker. Suffice to say: there are means to break a tie at tribal council, and those means have changed from game to game.

Although not a rule, it's important to note how the game is typically played. People protect themselves from getting voted out by gathering other people around them and agreeing to vote together. These alliances are the core of the game. We'll get to how they work a bit further along.

That's the game in a nutshell.

As you may recall, back in Trials, Triumphs & Trivialities #38, The Game Is What the Game Is, I offered one of the most integral rules of game design. Paraphrasing the title of that article: You encourage specific styles of play in a game through the systems that you design to run that game. Few places is that as clear as in the television game of Survivor.

Indeed, based on the simple game design I outlined above, a specific style of gameplay has evolved. It goes something like this:

1. Stage One: Vote Out the Weak (Members 16-11). Early on, immunity is tribally based. An individual player cannot be voted out of the game if his tribe is not forced to go to council, and thus an individual player benefits from his tribe remaining strong. As a result, in the first stage of Survivor players tend to be voted out if they are weak — dragging their tribe down (and to tribal council).
2. Stage Two: Vote Out the Charismatic (Members 10-9). When the tribes merge there is a time of societal destability. The players have already formed their alliances to help protect themselves from getting voted out, but the merge threatens those groupings. New leaders — or want-to-be leaders — try to gather members from other alliances. Thus, shortly after the merge — and as a result of the destabilization caused by the merge — leaders and manipulators tend to get voted out.
3. Stage Three: Vote Out the Strong (Members 8-6). As the leaders go down, the dominant alliance then begins to turn to rest of the non-allied members. At this point threatened players become desperate to win immunity, and game leaders try to ensure that their enemies do not do so. Thus, whenever a strong player does not win individual immunity, it becomes very advantageous to vote them out, lest they engage in an "immunity run" (winning 4-6 immunity challenges in a row, something which happened in the first two shows).
4. Stage Four: Vote Out the Cute (Members 5-3). As the game drops down to its last few members, the dynamics of the game begin to revolve around the upcoming jury vote. Members begin to assess who they can beat in a final tribal council; as a result players who are cute, nice, or well-liked become top candidates for expulsion.

At least, that's the theory — that's how the game dynamics of Survivor generally suggest that the game play of Survivor should go. But, watchers of the television show already know... it doesn't always work out like that.

One of the strengths of Survivor is that its simple game mechanics can allow for complex game play which can change and evolve. Or, to look at it from the game designer's perspective: when you're building game systems to try and influence game play, you need to think about them very carefully, because players will do all kinds of wacky things.

As the players of Survivor have.

Modifying the Game I: Token Look-Ahead

Here's the first game design corollary suggested by Survivor: Players will sometimes think one step ahead.

Although the outline above suggests the most obvious gameplay based upon the game mechanics at the core of Survivor, in actuality it's a fairly naive structure. It presumes that players will never look ahead — that they will always act upon the needs of the moment. That isn't always the best long-term strategy in Survivor, as can be seen when examining an ideal (though abstract) optimum strategy for Survivor. It goes something like this:

At start an individual player of Survivor finds himself in a tribe of 8 members. He can immediately assure himself that he will reach the merge (10-member) point by forming an alliance with 4 other tribe members. His 5-member alliance will have clear majority in an 8-member tribe.

At the merge a player can find himself in one of three situations: an advantageous merge (where his tribe has 6 or more remaining members and the opposing tribe has 4 or less), a balanced merge (where both tribes are at 5), or a disadvantageous merge (where his tribe is the one with 4 or less members). In a balanced or disadvantageous merge, a player's fate is up in the air. On the other hand in the case of an advantageous merge a player can assure himself that he goes on — as long as he has convinced at least 1 member of his tribe that is not a member of his 5-person alliance that they really are. Thus requiring the 5-member alliance to actually have 1 pansy/supporter.

Once a player has gotten past the pivotal merge vote, and the tribe's size is decreased to 9, he is once more in the driver's seat. Because he has a 5-member alliance, and thus once more majority, he assures himself of getting down to the 5-person tribal council.

In order to win victory at the 5-person tribal council, a player must have formed a suballiance of 3 within his alliance, while at the same time convincing the additional 2 members that they are actually the core members, so that they continue voting with the group until that 5-person tribal council. This therefore requires the 5-member alliance not only have the one pansy/supporter mentioned above, but also 2 fall guys within the alliance.

This assures the player of reaching the 3-person tribal council. At this point forming a 2-person subsuballiance is insufficient, because the one person outside the subsuballiance could win immunity, and vote the strategist out. Thus the only way to assure himself of victory is for a player to form two subsuballiances, one with each other member. The truly gutsy player then purposefully loses the immunity challenge before the three-person tribal council. Whomever wins considers the manipulator his true partner and brings him along — at the same time voting out the manipulator's other partner, but not forcing the manipulator to do so himself (which would make him look bad to at least one jury member).

Finally the manipulating player attends the two-person tribal council, explains how cleverly he played the game, and wins.

Although there were some glitches along the way, this isn't a bad description of Richard Hatch's strategy in Survivor I. He:

• Formed an alliance of five.
• Thus stayed in control until the merge.
• Took victory in a balanced merge because the other team didn't realize that they were facing an alliance.
• After destroying the other tribe, voted out his alliance's fifth wheel.
• Let one of his partners vote out the fourth member, taking the heat herself.
• Purposefully lost the three-person immunity challenge, knowing that either winner would take him to victory.
• Gloated about his Machiavellian strategy and won.

Tina did much the same in Survivor II, though in her case she let other players believe they were the alliance leaders.

However, looking carefully at the strategy outlined above, it becomes clear that it doesn't fit perfectly with the "obvious" gameplay I described in the previous section. Most obviously a player might actually decide to vote out either a strong player or a charismatic player very early in the game, if he thought doing so would help solidify his own alliance — despite the fact that the game mechanics seem to indicate that's a dumb move.

In addition, because a player needs to fool between two and four of his alliance members into thinking that they're more valuable than they actually are, a manipulative player may decide to purposefully gather the weakest members (who he can fool) around him as the members of his alliance — another strike against the strong, even early in the game.

The bottom line is, quite simply, that the players of Survivor (or any game) will adopt different gameplay than is immediately obvious based on the game mechanics at a specific time, because of the fact that they're going to be looking ahead at how the entire game will be played.

Modifying the Game II: Evolution

Here's a second game design corollary suggested by Survivor: Game play will evolve as a game continues to be played.

In a poorly designed game, one strategy will eventually arise as the clear winner, and at that point all future iterations of the game will be simple mechanical exercises to see who can best (and most quickly) use that strategy. In Survivor, we've instead seen strategies evolve from one season to the next. It's sort of like rock-scissors-paper. No strategy is the best one, simply the best against what else is being offered at the time.

• In Survivor I, there were two core strategies: one was to form an alliance to vote out opponents (the Rich strategy), another was to vote individually based on one's own ethics and morals (the Pagong strategy). The alliance strategy won.
• In Survivor II, there were two core strategies: one was to create a strong, visible alliance (the Jerri strategy, following in Richard's footsteps); another was to form a quiet, secret alliance (the Tina strategy). The secret alliance strategy won.
• In Survivor III, there was a chaotic mishmash of strategies, as, in the wake of the first two Survivors, almost everyone tried to form alliances. This inevitably led to two core strategies: one was to act as the leader in scheming and forming alliances (the Carl or Silas or Lex or Brandon strategy); the other was to quietly follow a leader and let him take the heat (the Ethan or Kim J. strategy). The strategy of quietly following won.
• In Survivor IV, we saw a return to the two core strategies of Survivor I: one was to form an alliance to vote out opponents (the John strategy); another was to vote individually based on one's ethics and morals (the Maraamu strategy). Miraculously, this time the individual strategy won — because the individual players knew about the alliance strategy and were willing to temporarily adopt it until the alliance strategists were gone, at which point they could return to voting individually.

Overall the evolution of survivor shows us three things: first, that over time strategies will change, with new ones rising to the forefront and often defeating old ones; second, that much of this evolution is based upon knowledge of other strategies; and third, that this knowledge can even turn losing strategies into winning ones.

Modifying the Game III: Losing Victories

Here's a third game design corollary suggested by Survivor: Winning strategies will change based upon how people play the game.

Much of this corollary has already been suggested by the above description of game evolution. In Survivor IV, forming an alliance was not a winning strategy exactly because of the fact that it had been in previous shows. However, the converse of this rule is also true: a losing strategy can become a winning one based on different game play.

In Survivor I, Survivor II, and Survivor III, it was considered the kiss of death to betray an alliance or break your word — primarily because of the jury vote at the end. If you lied or broke your word, people would be less likely to vote for you as the ultimate winner, and thus you would discover that you'd been playing for second place.

In Survivor IV, something unique happened: both of the players in the final two had broken alliances and betrayed their word; thus it became less of a stigma. The winner of the show was actually the player who had broken at least two alliances (Vecepia) rather than the one who had broken only one (Neleh).

This unlikely condition was the result of a reverse Prisoner's Dilemma puzzle. As long as no one had broken their word, it was disadvantageous to do so because the oath breaker dramatically decreased their chance of achieving a final victory. However, as soon as one person did break their word, it became more advantageous to do so, because the second oath breaker would know there was one person who he would have a fair chance to defeat in the jury vote. In actuality, this could create an avalanche: the more people who broke their words, the less disadvantageous it was to do so (from a game play perspective at least).

Within Survivor IV, four people all broke their words (including both of the players who made it to the final two) by jumping alliances at the exact same time, thus demonstrating a way that a previously losing strategy could clearly be turned around in a simple, repeatable manner.

Modifying the Game IV: Different Victories

Here's a fourth game design corollary suggested by survivor: People will invent their own victory conditions, and each individual's victory conditions will affect their individual game play.

It's fairly standard for a game designer to presume that all the players will be playing for the victory he has defined — bankrupting the other players, getting three X's in a row, winning a million dollars, or whatever. However human beings tend to make up their own victory conditions. They'll then play the game in a way which might seem bizarre, because their goal is different from every one else's.

This has been seen in every season of Survivor:

• In Survivor I, the entire Pagong tribe's goal seemed to be simply to have fun. As a result, they never thought much about game design, mechanics, or game play, and when they met players who had (the Rich-led Tagi alliance), they were quickly wiped out.
• In Survivor II, it appears that Jerri's goal was to get herself noticed so that she'd be able to get acting jobs after the show was over. As a result, she was bitchy, arrogant, and overassertive. It got her voted off the island much more quickly than she would have been if she'd instead been nice... or just quiet.
• In Survivor III, it appears that Frank's goal was not win the game, but rather to win the game honorably. As a result he refused to team up with players he didn't respect, even when doing so could have handed him at least a trip to the final four, and thus he joined the long list of Survivor losers.
• In Survivor IV, Gabe clearly stated that his goal was to form a community. He didn't care about winning the game at all (at least by the victory conditions set by producers Mark Burnett). He got voted out very early, because other players considered him untrustworthy — because he wasn't playing the same game they were.

The Pagongers never should have ignored the game; Jerri never should have been obnoxious; Frank never should have been honorable; and Gabe shouldn't have tried to build a community. At least, not if they were trying to win the game by the victory conditions that most other people were using.

Modifying the Game V: Irrational Humans

Here's a fifth game design corollary suggested by Survivor: People will not always follow the best strategies, because they're irrational humans.

In the heat of the moment players often do things that look irrational to pundits looking on from outside (in the case of Survivor, us, the viewers). It's human nature, and it's shown itself many times in Survivor. Just to offer a few examples:

• In Survivor II, Kim J. was never voted out, despite the fact that she was clearly weak early in the game, and therefore she kept dragging her team to tribal council. The problem: she was nice and people liked her. Irrational.
• In Survivor IV, Peter was the first person voted out of the game, despite the fact that he was strong and a hard worker. The problem: he was weird and bossy. Again, irrational.

It's interesting to note that the irrationality of human beings can cause some of the other basic game play modifications noted here, through a very Darwinian process. Irrational human responses to game design stimuli mimic the mutation process in biological organisms. In turn, some of these new, accidental strategies prove themselves to be more fit than their predecessors, and thus strategies evolve — losing strategies win, and winning strategies lose.

Manipulating the Game

I've spent much of this article identifying a basic system of game play suggested by the game design of Survivor, then showing how it can change, evolve, and mutate. However, before I close up I'd like to note one final thing. My basic rule, that The Game Is What the Game Is, is really stronger than I'd previously suggested. In other words, the designer of a game can carefully manipulate the game at a very detailed level in order to produce the exact results (game play) that he desires.

Two Survivor examples spring immediately to mind.

By the time he was filming Survivor II, producer Mark Burnett clearly knew it was boring for the audience to watch an alliance form, then cut its opposition down, one member at a time, week after week. Thus, in Survivor II, we started seeing challenges which involved players specifically knocking opponents out of that challenge (thus, making them illegible for immunity that week).

This didn't really pay off until Survivor IV. Prior to the 9-player tribal council, an immunity challenge took place which involved cutting down opponents' coconuts. At the time there was a 4-player alliance and 5 individual voters, at least 2 of whom had been brought into the 4-player alliance as 5th and 6th wheels. However, during the coconut challenge the 4-player alliance worked as a team, knocking their opponents out of the challenge one by one in the exact order that they planned to vote them out at tribal councils.

All of the individuals, including the extra wheels, saw their fates laid out for them. They caught on, banded together, and then destroyed the alliance (one by one).

By seeding this directly oppositional challenge at an appropriate time, Mark Burnett had encouraged a specific type of game play: unifying against the dominant alliance. And it worked beautifully. (Parenthetically it's worth noting that this tactic hadn't offered any sparks in the past, as with the sling challenge in Survivor II, but then players won't always react the way you'd like, for all the reasons already noted.)

Another good example of how Mark Burnett has manipulated game play at a fairly low level can be seen in the immunity challenge before the 12-person tribal council in Survivor II. At the time, the Ogakor tribe was at a severe disadvantage, only having 5 members to the other team's 7. It seemed likely they'd face a disadvantageous merge and then be voted out of existence.

But the 12-person immunity challenge involved a maze. The members had to stay together while collecting idols and returning them to the maze's center. Because of the complexity of coordinating a larger group of people for this task, the 7-person tribe was actually at a disadvantage, and thus the smaller tribe won.

This type of challenge, which disadvantages a larger tribe, has been seen at other times in Survivor. In addition other late challenges are often either creative or mental — the exact type of thing that a tired, disadvantaged, or discouraged tribe would be more likely to win.

Some people accuse Mark Burnett of quickly preparing challenges of this sort when one tribe is getting behind, but a good game designer actually doesn't need to do that. Knowing that the game will be most interesting if tribes are balanced, Mark Burnett can ensure that the last few challenges before a merge are the exact type that will always give the advantage to a smaller or more discouraged team. Thus far it looks like he's done exactly this — and that this strategy has worked well — because out of four Survivors, three of them included a balanced merge.

Shape your game play and its results through game design — that's what the axiom that I started this column off with is all about.

Survival Myths

I've got some other things I'm tempted to write about the game design of Survivor — how winning challenges can actually be beneficial or detrimental at different times, how different levels of assertiveness can be effective, how reward challenges affect the overall results in minor ways, etc — but I've pretty much decided that what remains is about me being a fan-boy rather than a game designer, so I've decided to let them rest. I hope what I've laid out thus far, however, will prove generally useful in demonstrating how a simple game design can shape game play and how many things can modify that game play.

Next week I'm back to the topic I started a month ago: mythology. I want to turn the generalities of previous weeks into specifics, building on the work of Campbell, Jung, and others.

I'll see you then.