An Ever Ending Story

When I wrote my interactive fiction game “20-4-7” (currently only available in German, translation is in process), I wanted the players to experiment with possible consequences of their actions. The game is set on an airship with a bomb in it, and there are several ways to dismantle the bomb to save everyone aboard. It is also possible to have just one or two characters escape the ship with parachutes, or the player could just ignore the story line with the bomb and focus solely on the interpersonal relationships of the main characters. This would mean in most cases that the bomb goes off, killing all PCs. While this would be considered a proper ending for a linear narration (Titanic would not lose anything of the interpersonal story if all characters died at the end), it implied much more emotional investment for an interactive experience.

The players wanted to save the main characters. And while replaying the game and experimenting with different decisions and outcomes was the exact gameplay I had in mind when designing the game, focusing solely on the bomb story arch was not. So I added a disclaimer to the game, informing the players that the end with all main characters dying was not the “bad end” (and no “game over”), that they did nothing wrong in that case and that a catastrophe was still a decent end for such a fiasco-loaded game. Form that moment on, the players accepted the death of the main characters as a story solution and started to focus on the major part of the game that happens before the bomb goes off, instead of completely concentrating on the game’s ending.

And this got me thinking.

Game Endings Should Be Valued Less

“Multiple endings” is a big selling point for games these days. The number of different endings for a game seems to act as a criterion on a game’s interactivity. Just as in the example above, players seem to feel cheated of their freedom when all their efforts lead to the same outcome anyway. Regardless whether the meaningful choices they made led to no difference or make for two extra minutes at the end – as long as there are different endings, the game appears more interactive. This probably ought to enhance the game’s replayability, although five different lines of dialog at the end alone certainly do not do this. And though it could solve a lot of larger-scale story problems (like stories of game prequels, expanded universes or game worlds with heavy lore), a fixed ending for a game is considered bad. But why is that?

In my opinion, the ending of a game is too much in focus of both developers and players. Contrary to movies where all actions work towards the ending, and the ending itself is quite a big part of the experience – probably at least one 60th of screentime – the focus of a game lies in its middle part. And somehow this movie expectancy gets carried over to games, where suddenly the short ending – probably one 1000th of play time and in most cases not even interactive – is valued so much that an undesirable ending spoils the whole game for some.

Since developers know that quite well, they work satisfying endings into their games  – endings that many players will never see, for a large number of them does not continue playing a game to its very end. Should some writer dare to include a cliffhanger or open ending with unsolved questions into a game, he is soon forced to develop a sequel that leads the waiting community to a satisfactory ending (as happened to Scratches, for example). Also, unusual endings are often considered the “wrong” or “bad” ending for a game, which means that in fact there is only one right ending, the others just being punishments to the player for not playing right.

My dearest hope is it that endings of games will leave the focus of attention in terms of that people stop expecting happy endings – for the emotional investment games build up over their long play time can be used quite well for other intense ending releases like sadness (Final Fantasy X; Red Dead Redemption) or puzzlement (Bioshock Infinite).

Game Endings Should Be Valued More

As soon as we reach a point where the ending of a game is more than a mere means to reward the player for his or her investment a last time before shutting of the game, we can start concentrating on telling meaningful stories through a game’s ending. When we are not bound to please the player in a satisfactory way, we can start treating him as a mature entertainment customer that also endures sad, revealing or open endings.

This also means that I argue for not shying away for writing the occasional, mandatory “bad ending” – with dying main characters, world destruction, betrayal, confusion, and so on. At the moment, such endings seem to just disgust players. A great example for this was the ending of Quantic Dream’s Fahrenheit (you might know it as “Indigo Prophecy”) – eventually, the character will obtain god-like powers which he uses to destroy all of existence and create a new world. The ending was not well received, because of which Quantic Dream returned to rather conventional resolution in their follow-up game Heavy Rain. In fact, many players would not like Fahrenheit just because the ending did not please them, although it changed nothing about their play experience with the game (not even re-contextualize it).

Of course, some of my arguments do not refer to all types of (narrative) games. Some open world games like TES V: Skyrim do not have an ending. Sure, each story arch has an eventual resolution, but the game itself never ends. Assassin’s Creed games on the other hand use to have two endings per game: An in-animus ending (that usually concludes with a positive resolution) and an off-animus ending, that is usually an open-ended cliffhanger. Even crazier deviations from the classical endings are conceivable: Imagine a game with an interactive ending or with one that needs the player to break an eternal cycle (which the Stanley Parable does quite well).

So What Now?

To sum my point up, I would wish both players and developers to devalue the ending of a game so to value it more. Taking a step back from emotional investment and the connection of all of the game’s worth to the ending would enable the developers to tell more meaningful endings and derive them from a much broader range of tropes, while it would in turn enable players to value the sheer narrative and artistic worth of the ending instead of just conceiving it as a source of pleasure or possibly just another reward for their achievement.

Lines, Trees, and Clusters

Some of the concepts presented here are partly based on Justin Alexander’s node-based scenario approach to writing Pen & Paper RPG adventures.

Though the industry talks much about interactive storytelling and non-linearity in games, most games available to this day actually still follow pretty linear approaches to leading the player through the game’s environments. And though one might argue that open world games offer freedom of choice, these choices appear to be choices of different linear paths to follow rather than open interconnections. The reasons for this linearity can be manifold, but the most-quoted of them is probably the manageability of linear content. When there is a predefined route for the player to follow there is little doubt that the player will encounter all content placed there for him or her to find. The argument goes that each larger piece of content that will not be encountered during playthrough is lost development time.

What I am suggesting here is that there are two conventional ways to address this problem:

  • Following the linear approach
  • Branching the story into a tree-like structure

And then there is another solution that I clearly prefer above the others:

  • Keeping the story open in a cluster-like environment


A linear story path

A linear story path

In the shown example, each of the squares might stand for a point in a dialog, a place the PC is to visit or an in-game interaction during a quest. For example’s sake, let’s imagine we are talking about a single quest in a fantasy RPG where the player has to find and retrieve an abducted king. The developers prepare six interactions (talking to court members, catching a member of the kidnapper gang, interrogating him, finding the gang’s hideout a.s.o.), and the player will encounter any of these steps, one after the other. The story will appear well prepared, but the replayability is comparatively low due to the lack of diversions the player can take. This makes for 6 interactions (6 per playthrough), 0 choices, 1 way and 1 possible outcome.


A branching story tree

A branching story tree

To offer players greater freedom, branching decisions are often implemented into games. This results in a tree-like structure that allows for the player to decide at certain points to follow a different path. Players feel more empowered with this non-linear approach, especially when comparing to other players, and replayability is kindled as well. On the downside we have the enormous development cost – the example requires the developers to implement 22 different interactions, but upon each playthrough the player will encounter but four of them. In the said example this could mean that the player can decide between interrogating court members, searching for clues in the king’s bedroom, visiting a seer and so on, with another two critical decisions later on, based on the taken path, but no backtracking. All in all, this makes for 22 interactions (4 per playthrough), 3 choices, 11 ways and 11 possible outcomes.


An open story cluster

An open story cluster

Following Justin Alexander’s insights on Pen & Paper RPGs, the next logical step is to open the system up for an almost indefinite number of player choices. Such a clustered approach keeps up the impression of freedom of choice for the player, while having this freedom emerge from the system itself rather than from predefined content. The cluster allows players to freely move between the nodes of the cluster, going back and forth and backtracking as he or she pleases until a defined outcome is reached or it is interrupted by the player. In the fantasy RPG example, this could mean that the player might wander around the court, talk to different people, search for clues and so on. From the gathered information can then be concluded where the hideout of the kidnappers is. This could happen through cutting the hideout-related information up into different parts (like the surroundings – forrest, desert, mountains, – the cardinal direction, and the secret keyphrase to enter) and then have the player uncover only parts of these from each interaction until the information is complete (doubling the pieces and hiding them in different nodes is therefore necessary). The player will maintain full control as to where he or she goes and how the investigations proceed, finally uncovering the truth using his or her very own way of doing so. To sum this up, the Cluster approach will require the designers to come up with 9 interactions (any number of them per playthrough) and an almost indefinite number of choices, ways and outcomes, as all these depend on the player’s decisions throughout the quest.


This post argues that using a cluster-based approach, development cost can be held comparatively low while still providing the player with an interactive and replayable experience depending on his or her playstyle with almost no lost content. Any of these techniques can also be used during a conversation with an NPC – with the Line approach equaling a cutscene, the Tree approach corresponding to usual dialog trees and the Cluster approach providing a technique where (using only a few generic interconnections), topics could be addressed in a freer and more natural way, without endangering the player to have missed a critical branch of the dialog.

Please note that other than in a tree-based approach, the cluster does not withhold any node from the player – he or she can access any node at any given time, provided the correct interconnections are taken. This way, no content is lost to the system, and the better part of the story emerges from the interactivity itself.