It is our choices, Harry, that show what we truly are, far more than our abilities.
-- Albus Dumbledore, Harry Potter and the Chamber of Secrets

Some time back a company hired me to consult on the design of a large shooter game that was planned to have several different endings. On reading the design document, I discovered that they were all uniformly grim. I told the designers that commercial shooters are not Kafka novels and that players who had made it through to the end of such a game will feel entitled to at least one moderately positive ending. They took my point.

	Thus spake Dumbledore.

Artistically, my position was wrong: designers should certainly create the videogame equivalent of Kafka novels if they want to. But commercially, I think it's sound. This company had asked me for my opinion about a commercial game, and it's my opinion that the game-buying public is not yet ready to spend $50 on a triple-A shooter in which every possible ending is a downer. Winning and losing is a matter of definition, but a bad ending feels like a loss, and a happy ending feels like a victory. Over the years we have trained our players to assume that they will eventually win a long game sometime, though they may have to try repeatedly to do so. The longer a game lasts, and the more a player has invested in it, the less likely she'll be to tolerate an unhappy ending, as I told my clients. Adventure games or mysteries might get away with it, but if a game about survival doesn't end with the player surviving, she's going to be annoyed.

However, the episode got me thinking about the whole issue of multiple endings. It's a basic question for anyone who is designing a videogame: do you want the game to have one ending, or multiple endings? What should they be like, and how do you decide?

Game Endings in Different Competition Modes

Well, if you shoot a game of pool with someone in a bar, there are two outcomes: either you win or you lose. If you sit down to play poker, when you get up you have either more money than you started with, or less money: a final score, which you can then use to define victory or defeat. With no more analysis than that, we might conclude that videogames should either have a way to win and a way to lose, or a final score. However, a great many videogames - perhaps even most of them - are not analogous to pool or poker. Because there are so many kinds of games in the world, it's difficult to create one set of design rules that applies to all of them. I'm going to try to pare the problem down by considering the different competition modes of gameplay, because for a good many types of games the answer is straightforward.

We can answer the question quickly for team games and multi-player competitive games, whether we're talking about video pool or Counter-Strike. Competitive games obviously need the standard types of endings. That's what players expect when they're in a competition with each other. Multi-player cooperative games, where they all play together against the computer, are a bit less clear: sometimes they all win together; sometimes they all lose; sometimes one player drops out while the others go on. Ultimately, the game really only has three possible endings, though: either none, some, or all of the players have won. (Usually "some" and "all" are combined into one ending.)

We can also answer the question quickly for most massively-multiplayer online games, especially persistent worlds and MMORPGs. Generally, they don't have any sort of ending at all. The whole point is to continue playing indefinitely. Players try to make sure they don't "lose" (die), but death usually isn't permanent, and in any case the game still goes on for the other players. There's no such thing as "winning" in a persistent world unless that world has a defined endpoint and a victory condition, and I don't know of any that do.

In Sim City and most construction-and-management simulations, you can lose, but you can't win, so the game only has one ending - and like MMORPGs, "winning" isn't the point of playing. Classic arcade games like Space Invaders have no victory condition either; the game just gets harder until you lose. The only way to "win" is to end with a higher score than everyone else who has ever played. But whether you define that as winning or losing doesn't matter. The game still has only one ending.

Short single-player games (from a few minutes to three or four hours, say) are pretty straightforward too. Most short single-player games aren't very complicated, and they typically fit into a binary win/lose pattern like solitaire, or they just give you a final score and let you decide whether you won or lost.

Long Single-Player Games

That leaves us with long single-player games, and they tend to be more complicated. First let's consider what happens if the player's avatar dies in the middle of a long game. This is clearly "losing," but I believe it doesn't really constitute a separate "ending." It ends the player's experience of the game, but it isn't the real conclusion to the game. The player knows she hasn't seen the whole thing, and should go back and try again. Also, there's no point in crafting a lot of content to accompany avatar death, because the player will be seeing it a lot and get bored with it. For the purposes of this discussion, then, I'm going to exclude premature endings caused by early avatar death.

So, leaving aside early death, how many endings should a long single-player game have? The answer has to do with the psychology of gameplay and the purpose that different endings might be intended to serve. First let's consider some of the oldest long single-player games, side-scrollers. Such games throw a steady stream of challenges at the player and make her work for her very survival. Side-scrollers typically have one ending, which is presented in such a way as to feel like a triumph. It's the player's reward for playing her way all the way through the game without dying along the way. Multiple endings would be superfluous; they would only reduce the player's sense of achievement. If one ending offers a greater reward than the others, the player is bound to view that as the "win" ending - the best ending - and the others as inferior even if they're not actually presented as "losing." There's not much point unless you want to distinguish between "good" play and "brilliant" play with something like an extra level or a special movie.

Long Games With Stories

Suppose we add a story to a long, single-player game. Starcraft, for example, had a linear storyline with a single ending. The way in which you played the game didn't change the story. As with side-scrollers, its gameplay was about survival and success, and the player's reward for succeeding was to hear more story. It didn't offer the player any actions that would change the storyline, and in fact, the story was mainly used to tie together a variety of types of missions. It was possible to play all the way through it without paying any attention to the story at all; achieving the missions was its own reward.

[CAUTION: There's a minor spoiler for Half-Life in the next paragraph.]

I figure that games like Starcraft only need one ending, and it should be presented as a win (though, of course, leaving something open for a sequel). To play all the way through a linear story and then suddenly give the player options for different endings is amusing but ultimately superfluous, and I think this was a (very minor) weakness in Half-Life. G-man's offer to Gordon turned out to be an offer he couldn't refuse, so why bother making it? It's a phony choice and a waste of the player's time. And if it had been a legitimate offer, it presumably would have required two different sequels, one for each option.

However, players who are really interested in a story, and enjoy taking an interactive role in it, want to feel as if their actions matter. They believe their actions should influence future events in the story, and that potentially includes multiple endings as well. A good example of this sort of game is Wing Commander. Wing Commander was famous for being a space-combat shooter with a real, non-linear, plot. If you failed a mission, it didn't always result in a premature ending; rather, the story simply took a different turn. Success or failure in the missions became a plot element, and indeed, Wing Commander had different endings.

However, I don't feel that this is absolutely essential. It depends on the nature of the story itself. A single soldier in a large war has little chance of making a really big difference in the war. He can make a series of small, cumulative contributions, and his choices will have major implications for his own experience; but they won't necessarily change the outcome of the entire war. The object of the game is to survive the war, and different players may approach the problem in different ways and experience different storylines, but ultimately the ending is the same: you did survive and (perhaps) the war is over. So you can legitimately have a non-linear storyline that has a single ending, if the ending is characterized in large-scale terms rather than personal ones.

Challenges versus Choices

When you're trying to decide how many endings you want, take into account the types of actions that the player can use to influence the story. If what influences the story is success or failure at challenges, then it makes sense for the story to have some sort of notion of achievement built in, with endings that vary on the basis of that achievement -- winning and losing endings, or perhaps a final score with a number of different consequences. Players are legitimately entitled to think that overcoming a challenge is good and should be rewarded by some kind of positive outcome, whether it's loot (as in RPGs) or a positive ending to a story; whereas failing at a challenge should produce either no outcome or a negative outcome. This is what happened in Wing Commander, where challenges (succeeding at missions) controlled the storyline for the most part.

The other major type of action that can influence a story is a choice. The difference between a challenge and a choice is that a choice does not require that the player work to achieve anything (although the choice can be difficult to make up your mind about). As a result, there's no built-in presumption that a good choice should produce a reward, or even that there is such a thing as a good choice. Of course you can build in right and wrong choices leading to good and bad endings, but be careful -- taken too far, this leads to a moralistic, preachy game. A more interesting approach is to offer the player morally ambiguous choices and explore some of the possible consequences of them.

The easiest sort of choice to implement in software is the simple choice: the player makes a particular decision and the storyline goes one way or the other based on that choice. When they make another choice, the storyline branches again. If the branch occurs immediately, this feels rather mechanistic: the player can sense the presence of a game tree. (Although they're indispensable for RPGs, this is one of the weaknesses of tree-based dialog in games: you can actually see the game tree unfolding in the dialog.) You can hide this mechanistic quality somewhat by delaying the consequences of a player's simple choice until later in the game (the system records the player's choice but the actual branch occurs later). However, this also has its drawbacks. If the player regrets her choice, she has to go back a long way to correct it, and she may not have a saved game that old. The Hitchhiker's Guide to the Galaxy was roundly cursed for having an apparently-meaningless choice near the beginning that prevented the player from winning the game near the very end. As a general thing, a choice should not appear to be meaningless, its consequences should not be too distant, and the player should know (or be able to deduce) when and why the consequences are occurring. And of course a choice need not produce only one consequence; it may have numerous effects throughout the game, such as choosing an evil alignment in Dungeons & Dragons.

Another form of choice is the cumulative choice, in which the player's cumulative actions affect the storyline, but no single decision is responsible. You can see this in games that give you the opportunity to take good or evil actions, without requiring that you specifically choose to be good or evil at the beginning of the game. Black and White is a good example. The cumulative consequences of your chosen actions affect the way the game plays and the way your pet creature (it's not actually an avatar) looks.

One thing we can be sure of: if the story changes significantly on the basis of the player's choices, she's going to want to go back and see what happened if she made a different choice. It's human nature to be curious about what might have been. Do your players a kindness, and allow them to save the game right before any important choice - particularly if the decision might have disastrous consequences that they could not have predicted.

Multiple Storylines and Multiple Endings

When you make a game full of choices, the story can branch out in a combinatorial explosion of possible lines and, theoretically, have a vast number of endings. In practice, though, I don't think there's much point in this. For one thing, it's prohibitively expensive to create all the content, as they found out with Wing Commander; for many years Wing IV was the most costly game of all time. But beyond that, I don't feel that it's actually necessary. A player who is motivated by story is seldom interested in having complete freedom to obtain any sort of possible ending. A player who wants complete freedom will play one of the Grand Theft Auto games and ignore the missions entirely, simply running around and committing mayhem.

What a story-motivated player wants is a satisfying, credible, narrative conclusion to the experience she has just had. An ending that relieves the dramatic tension of the story and winds it up in a way that is coherent. This doesn't necessarily require a large number of possible versions. Remember, as you learned in high school literature class, a story should have a theme - a message. The theme is the soul of the work, and it should touch the player's heart and mind, make her think and feel. The presence of a thought-provoking theme is the hallmark of great literature, and is what sets a work above mere escapism. A story can have an ambiguous theme, but it should not have a flatly contradictory or illogical one. Much of that message is delivered right at the end, in how the story concludes: happy, sad, disturbing, ambiguous. Even if a story is interactive, this still holds true. If the choices in the story offer too many options, too many storylines, and above all too many endings, you risk the theme being confused or lost.

Planescape: Torment, a game that I otherwise love, had about eight different conclusions, each based on a particular choice the player could make at the very end of the game. I saved and reloaded until I had seen them all (I think). I understand the designers' motivations for including them: the game was full of difficult choices, and they wanted the player to make one more big one and show the consequences arising from it. The endings were meaningful and in some cases profound. And yet I didn't feel they were all necessary. One ending felt like the "best," and the others - while clearly not "losing" - were less satisfactory. The game's theme became slightly muddled at that point. I think it would have been more powerful if it had had only one ending, or if the ending had been based on a cumulative choice made over the course of the whole game, rather than a simple mechanistic decision at the end.

Conclusion

To summarize, never create multiple endings simply because you can: such endings are a waste of money and your game is unlikely to be improved by them. Rather, create multiple endings when they serve a higher purpose. Some of those purposes are straightforward: to reward the player for achievement and punish her for failure; to create a judgment about the quality of her gameplay. Winning and losing are obviously important endings for many types of games.

If your game includes an interactive story, you may want multiple endings to reflect important actions that your player has made in the course of the game, actions that she expects to have meaningful consequences. If your story is largely influenced by challenges, then the endings should reflect her triumphs and tragedies. With choices rather than challenges affecting the story, you can go in many more directions, but be doubly careful here. The emotional power of an important event is diminished if you can go back in time and unmake that event. Starcraft ended with a heroic character sacrificing himself for the good of the galaxy, and the player could do nothing to prevent it. That moment would have been far less touching if the player could simply reload and choose a happier outcome. There are times when a player shouldn't be given a decision; it reduces the impact of your game.

	Starcraft sports a linear story with a single ending.

Be sure that each ending is coherent with the storyline that produced it, however varied it may have been. Make certain that each ending works to create a sense of narrative satisfaction, the "ahhh" feeling that you get when you come to the end of a good book or movie. And take care that all your endings reinforce your theme, rather than muddle or distort it.

I'm not saying not to make multiple endings. I'm saying to make them carefully, and for a reason.

______________________________________________________