Thursday 9 December 2010

The Theory Behind In-Game Failure

As a topic for theory, failure states aren't new. If you're providing the player a challenge, the traditional way to handle his failure to live up to that challenge is a


followed by a discretionary amount of replaying from the last checkpoint. To a certain degree I'm sold on this concept. Particularly in these days of casual, persistent and console gaming, developers are always seeking ways to avoid those ugly words,


and allow the player to proceed without breaking the fiction (see my discussion on Bioshock's Vita-Chambers). Sometimes this cuts short your options: if you're escorting an NPC how do you handle that NPC's potential death? Invulnerability? Branching story lines? Massive text redundancy?! Sometimes a good old fashioned


is all you need.

On the flip side, I love games that incorporate player failure not just into their fiction, but into their gameplay. In fact I think they're far more rewarding experiences. I believe this on the basis that its not just challenge that is central to good drama, but failure. With that in mind, in order for failure states to be meaningful they need to be incorporated into the both the game's fiction and, more importantly, the world of the game's mechanics.

I see three ways in which failure is modelled:

1. Game over
2. Forced fails
3. Gameplay incorporated

Game over (yeah, line breaking for 'game over' is getting annoying now) definitely qualifies as a failure, but in most cases it's mechanically unrelated to everything else in the game. You step back in time and turn that failure into a success in order for the narrative and interaction to continue.

Forced fails are when the story demands the character strike out, and the game forces the player to embody this. There's a time and a place for this, but clearly as soon as you're scripting player action to reflect the story, rather than the other way around, you've given up a core tenet of interactive narrative. In both game over and forced fail, failure acts as a barrier to interaction rather than a part of it.

The better approach, as I see it, is the smaller scale stuff that's handled by the game mechanics. That's things like getting shot, taking too long, or making poor decisions. These are things which, combined, might lead to a Game over, but which can usually be taken on the chin. I played a lot of RTS and management games when I was a kid, and I think one of the appeals of those genres is that game over is far rarer than in action games, and that failure is Incorporated into the flow. Messing up in Theme Hospital doesn't make the level unplayable; you don't have to redo the half hour since the last check point. What it means is having to hire new specialists. What it means is having to stare at the monstrosity of a Bloaty Head treatment machine you just built when all your patients are dying of Broken Hearts. What it means is being punished proportionately, in a way that's cohesive with the fiction and the game structure, and in a way that forces you to work harder to overcome your self-wrought challenges.

What it arguably means is a far truer interpretation of interactive drama.

Of course, there remain sticking points. How do we tell a story which revolves entirely around a protagonist's failure, but keep the player motivated? How do we tie this approach not just into the gameplay, but more thoroughly into a complex narrative?

I'm going to go play Theme Hospital (well, its open source clone) and think about it some more.


  1. Freespace 2 had an interesting, if simplistic take on this topic. In the final mission you escort survivors to a jumpgate as your home system counts down to destruction. The human race are refugees.

    You need to save the survivors in order to avoid game over and complete the game. But you don't need to save yourself. If you - through fault or will - stay behind to hold back the bad guys, or leave it too late and don't make the jump, the conclusion will reflect your 'sacrifice'. It's a very small change which doesn't affect the plot, but it's sharp: it brings to the fore the relatively inconsequntial nature of the protagonist's role in the grand space opera, and makes dramatic sense of a failure that's 100% a part of the gameplay.

  2. I found your previous post on failure states to be very interesting and the same is true with this one. It remains pertinent for me as it's something I still have yet to fully decide on in the game/mod I'm designing. Your Theme Hospital example was useful. I guess it's an example of a 'soft fail' as opposed to a 'hard fail' from which there's no coming back from - i.e.a game over and I think the use of 'soft fails' seems to be a good way to go with the fiction just expanding to accommodate whatever the gameplay consequence is naturally.

    What I'm still trying to rationalise is that even with these 'soft fails' do you not need to inconvenience the player for choosing the sub-optimal route which lead to their 'failure' so that they get the appropriate feedback for their actions that makes sense for the consistency of the gameworld? This could easily lead to those players that least grasp how to play the game best then having fewer resources to tackle the rest of the game with and lead to a vicious cycle in which they're ultimately going to fail eventually. Many games solve this by setting the bar really low and ultimately don't punish the player hardly at all but the gameworld then very quickly loses it's sense of reality which can be a bad thing if that's what you're aiming for.

    Maybe it's possible to just let the game play out differently rather more difficultly after a player has a soft fail - i.e. not a 'punishment' in terms of the difficulty of gameplay thereafter at all but nevertheless a difference in the narrative that still provides feedback tothe player - the world has adjusted itself aroudnd their actions. You can only provide these splitting pathways and soft fails without punishment so many times given the constraints that the fictions of many stories allow. I think I will ultimately work very hard to prevent failure in the first place by training the player of the failures they're looking to avoid and then just hope almost all will make it through the whole game with only a few soft failures and maybe subsequent minor punishments that the game narrative is able to accommodate before having to reach a complete dead end (game over).

    It's good to look for the clever ways to mitigate these things rather than taking the obvious paths but eventually there's probably only so much you can do.

  3. Another aspect of failure states is to convey them through cutscenes, i.e. beyond the player's control (as a subset of forced fails). Mass Effect is a game that does this. At the start, you're dropped on Eden Prime with a mission to stop the geth invasion and recover a Prothean beacon. From a gameplay standpoint, you succeed in this. The geth are eliminated and the beacon is recovered. Good job! But then the story comes wading in with a cutscene and blows up the beacon, which is then summarily blamed on you. You're a loose cannon, Shepard!

    I'm no fan of BioShock's Vita-Chambers (I'd quickload before dying until a patch came that allowed me to disable them entirely), but I do always appreciate the effort to incorporate failure states into the story and/or gameplay. I've touched upon how an Inception game would allow this here: Jubert's post made me think of a way to handle NPC deaths which I neglected to include, so I'll do that here.

    When an NPC dies in a dream, he would disappear (being kicked up one level). After a while, he'd reappear good as new, having reinserted himself into the present dream. When the dream is deep enough that death drops you down into Limbo instead, the NPC would seemingly die when killed, but immediately get back up again. This would be an abstraction of that NPC having dropped down into Limbo and finding a way to kill himself there. This seems instantaneous because of time dilation. Dynamic NPC banter could cater to this, like asking how an NPC killed himself this time. "Jumped off a building". "Whoa!"

  4. Damn, I started writing about this topic some time ago and never picked it back up. All very interesting Tom. I prefer it when a game can keep moving on within the fiction without stopping and restarting, even if the ways of mitigating it are cosmetic like in Queens where you're given 'lives' but they're masked as new individuals.

    You mentioned the Sands of Time a few posts back and that had a great way of dealing with death. I suppose Braid wasn't too dissimilar in that Tim never really died either. MMOs and games like Demon's Souls, which borrow heavily from them, have devices which build respawning into the fiction of the world, however clumsily.

    The problems come when the game's set in a realistic world and follows the story of a single character like Uncharted or GTA. How do you smooth over death when it's (supposed to be) absolute and there are no other characters to play as? Pretend that they didn't die from the bullet to the head or the 200ft drop? There're are only so many times you can watch Niko come strolling out of the hospital before it starts to get silly.

    It's easy enough to cite Heavy Rain but that wasn't exactly a death defying high octane action assault.

  5. @matthew
    The vicious cycle of bad player = punishment = worse player is certainly a problem. I suppose one solution to this might be dynamic difficulty, but the idea of simply morphing the player's experience (ala Heavy Rain), rather than inconveniencing it, is also an interesting one. I'd just be wary of making the way your failure affects the story too arbitrary.

    @everyone else
    I'm glad you guys are finding this as interesting a topic as I am. I think re: SoT and Inception we need to be careful to draw a line between failure being knitted into the fiction and failure being central to both the fiction and the game world. Where SoT effectively explains away its failure states, they're still really just game overs. Where L4D and an Inception game might temporarily inconvenience players without resorting to game over, they need to be careful to keep that failure central to the drama.

    How do you handle this theory in real world settings? Tough one. The petulant answer would be "Why don't we stop making games set in the real world about unreal things?" It's largely just death that causes big problems for continuity; stop making games about one man killing machines and you get rid of the problem.

    Failure in a relationship, or as a father, or as a teacher is far more flexibly represented, and much more interesting to boot.

  6. You could make an action game with failures in it that let the player avatar live no matter what. The model of a field commander that can lose her troops isn't unprecedented, it's just that most any game that includes such a character makes them less vulnerable instead of invulnerable.

  7. Great post and comments! Very interesting.

    And amen to this : "stop making games about one man killing machines". Game designers should read more French comicbooks. All of the pulp (and then some), none of the absurd, repetitious violence of teenage pop culture.

    Games about something else than gun- or sword-play are very much the solution to that fail-state dilemma. Minimize recourse to "Death" as the main thematic representation of failure, and all sorts of new design avenues open themselves.

    In this perspective, Inception is the status quo : for all its imaginative premise, it comes down to guns and death - but maybe I'm being uncharitable.

  8. Death, as a part of action driven gameplay, is almost an inevitability when it should be respected as a tool. Death is a device that can evoke a certain response, or attitude toward a particular game from the player by making them aware of the risks involved in their strategy and their actions. The issue arises when, as has been mentioned, when this particular tool turns the flow of gameplay into experience like being hit in the teeth with an iron bar, jarring you out of the world that invested in your immersion.

    I think that rather than creating more plausible uses of failure within games (although that is an interesting discussion), the game should focus more on ways in which the player's experience can be enhanced by drawing attention to their mortality, rather than writing off death as a triviality. Even in some of the most basic formulas as such as Call of Duty 4 online, they have created a mode that grants you only a single life; reflecting your mortality and creating some of the most intense experiences I've had with an online shooter. Of course this is a considerable distance from what a narrative could achieve in respects to a player's mortality, but perhaps something like CoD is the only experience devoid enough of character connection to get away with it. After all, you're only a number.

  9. If the Permanent Death playstyle was forced in Fary Cry 2 or other games, it could give it a lot of drama. Far Cry 2's use of buddies as your only way of surviving death would have a lot more weight then- you'd really owe them your life, versus your quicksave/quickload reflexes.

  10. The Far Cry 2 iron man, permanent death play through is an interesting read: