Procedural Narrative: Left 4 Dead

Happy New Year everyone!

Left 4 Dead got procedural narrative.  What’s interesting is that it did it in a way no one had really thought about – the gamey way – and what that means.

It had never occurred to me before I heard about L4D, but I thought horror was the perfect game genre for procedural narrative.  Pace is absolutely critical in horror, spawning and difficulty control a major component, and narrative events and discovery define the storylines, over conversations or non-deterministic player decisions.  But L4D took it even further then that.  Valve used procedural narrative to provide the game of their product, not just enhance it.

The reality of L4D is that it’s just not a systems deep game.  Game systems deep.  See zombies run.  Shoot zombies.  Shoot bad zombies.  If zombies close, shove, then shoot.  Try and not shoot to much.  Don’t shoot teammates. Stay near teammates.  Classic, yes, but also simple compared to other modern shooters.   But L4D doesn’t need deep systems, because it goes for deep narratives.

The AI Director’s procedural narrative doesn’t just give the respawns variety and change the pace, it provides narrative depth that the systems lack, player stories that change every time you play.  It keeps the game itself from being simple, by varying these mechanics in such a way that they are always new and interesting – in a story way.  Ever been a sniper who fell a bit behind?  Dozens of times, right?  Sure, but the one time a smoker got me from the woods, just as zombies mobbed the rest of my friends as they ran for the safehouse.  There’s nothing mechanically unusual about that situation.  But it is completely memorable.  And there are tons of situations like that.  The horror setting, the tight team dynamic and situations, the swaying difficulty, the orthogonally designed super-zombies, all compound to push the players alongside the pacing and set up dramatic conflict and hard, human choices.  As opposed to hard, systematic choices.  Even though there’s barely a traditional “narrative” at all in L4D, the mechanics and the experience manager seem to create a framework that allows the player’s to build their own unique story, but in a different way then the Sim and Civilization-likes designers refer to it.  This is player story built on structure, rather then player story built in a sandbox.  And we know from experience that structured storytelling creates something special.  Something that keeps players coming back, again and again.   In effect – narrative game fun.  That works.  This is something I’m going to study and continue to learn from – there’s likely a whole field of game design here that we’ve only been cracking pieces of in the past.

I’m not sure they were the first to do this, but now that they have, how well did it work?  Is the narrative actually a good, entertaining, deep one?  I’m not actually sure.  For the first dozen times, sure, L4D holds up.  The vast quantities of art help a lot.  But what I find most fascinating abou L4D is that it still ultimately feels shallow to me.  There’s just not enough “mechanics” to make it “skill-deep”.   But I think the goal was this to be narrative “mechanics”, not skill “mechanics”, and so falling short is maybe as much a function of theme and setting as it is of game.   It’s ironic – choosing the genre of horror allowed them to apply mechanically tropes that simplified the problem and allowed them to explore procedural narrative, yet gave them such simplistic themes that the narrative itself was shallow.

I take back a lesson about narrative’s gamey potential and the need for narrative sophistication.

Patrick thinks that structured procedural storytelling creates deceptive learning – that we should instead (I gather) be pushing for sandbox player stories.  I think the best way to find out is to explore this medium, even more.

17 thoughts on “Procedural Narrative: Left 4 Dead

  1. I’d be interested to know how often you played L4D with other people? I’ve been playing it quite a bit since it came out and I’d have to say that I always find it compelling to coordinate with other people to get through a stage on ‘Expert’. I think you’re absolutely right that they kept the game simple, but I think their hope was that the complexity would arise from the interactions between players.

    You’re right on the money with your other points. I think that L4D isn’t getting the reputation it deserves as an experiment in narrative.


    “… hard, human choices.”

    This is probably the best description of the game I’ve read anywhere.

    • Hi Charles! It’s good to hear from you. I enjoy your work very much.

      I exclusively play L4D with others, usually full parties at work, primarily co-op, but versus as well. I find co-op to be much more interesting, surprisingly, probably for narrative continuity reasons. Switching sides to “re-play” a scene and usually not finishing each section just doesn’t feel right. One quirk is we aren’t very serious about it, so we aren’t training to get optimal, but I believe there will be a (small) versus community that forms around that. I personally haven’t treated Expert as a compelling personal goal (I wish they’d given me personal level times to beat, aka time trials, from a starting difficulty). Maybe that detracts from the complexity, aka Halo, and I’ll give it a try. The AI director does make me feel like story creation rather then skill challenge is the player goal, so it hadn’t much occurred to me – we’d rather finish then not in a lunch, even if we get dynamically bumped down.

      Hope that helps.

      • An interesting note. One of the things I did learn about myself from my player stories in that “hard, human choices” element of L4D is that I’m the guy who goes back, sometimes for good, usually for ill. When someone goes down, I usually rush to help them, even if the rest of us could otherwise escape, and usually end up bringing the group down with me because I shift my focus. I’m not sure I could have learned that and explored that facet and its consequences in another medium.

  2. I think horror is the most successful genre in video games, because fear and surprise are the two emotions games have the easiest to convey. The problem is that other media regard the horror genre as having less artistic value.
    Valve must be doing something right though, since Left4Dead is the only multiplayer game where I catch myself acting irrationally.

  3. I think it holds up pretty well (but to be fair I’m not past the dozen or so times you mention is needed to judge). 🙂

    However, one of the problems I have with the pacing – while very fluid and dynamic – seems to be fairly regular in it’s ups and downs. Anybody else notice any of that? I know I found myself wanting a really long creepy strecth in terms of the dramatic tension, but the game would inevitably give you another 30 seconds or so, and throw more zombies at you. Perhaps I’m just being picky and overly sensitive about the pacing.

    What’s most interesting is that it is a more well known feature of the game, and it does at least a decent job at creating a fresh experience. Even if it’s not hugely varied, at least other mainstream devs now can point their execs to L4D and say they’re going to do something similar, if they want to experiment with taking it a step further.

    • I’ve definitely felt that too – I’m really curious how much pacing arcing they’re actually doing. The difficulty seems to be more tied to the player setting and the set events in the maps then where you are in the “narrative”. I kept wanting the AI Director to defy my expectations – set me up for a big rush, and then have nothing happen, for example. Or have a few rooms just be completely empty and then do a huge mob. Maybe the predictability is fed by the lack of variety in AI Director spawn styles.

  4. I agree, I definitely feel a common pacing. Hopefully this is something that Valve will continue to work on. My interpretation of some of the commentary was that they really didn’t know how the game would go down with players, so they played it safe. Maybe now that it seems to be a success they’ll take some more chances and try more things on the next version.

  5. How interesting is it that we’re all able to pick this up so easily? It seems like on the surface it would be a really subtle thing. I think this goes to show how powerful a mechanic we’re playing with here.

  6. That’s a good point Dan – part of the tension of horror movies, and zombie movies in particular, is playing with what the viewer expects to create tension… When you think they’re going to attack and they don’t and vice versa, when that dead zombie really isn’t dead, and when you’re checking out a zombie to make sure it’s dead, and instead something else surpises you from behind.

    It seems like you could achieve that with the same types of spawn options, but I think it really takes a deep understanding of what makes horror movie pacing tick, if you will. Since what often happens (and I really wonder if it’s the case here) is that the developers get to a more complete understanding (if they’re attacking a new problem like this) only at the end of development, maybe they’ll play around with that some more. I’ll have to check the commentary too (thanks for reminding me Charles).

  7. Pingback: Left 4 Dead - this r srs lit review « A Pleasing Fiveness

  8. Pingback: Game AI Roundup Week #1-2 2009: 14 Stories, 2 Videos, 1 Job, 1 Paper, 1 Book —

  9. Something that I think is really important is the role that level design plays in this. We already know that Valve are experts at telling stories through their levels, and L4D is no exception — the smashed storefronts, writing on walls, doctors lying dead next to empty gurneys. This really helps, because there is already such a strong ambient context to everything, that our silly human

    But the other interesting thing is how the level design multiplies the AI director. The director only has like 10 gambits or so, but the portion of a level they appear in creates scenario. E.G. a smoker spawns and grabs the rearmost player: In a tunnel, this feels like, “Don’t stray behind!” and in a field this feels like, “They are trying to separate us!” and on a rooftop it feels like, “It’s gonna pull you off the edge!”

    I think this is an interesting example of created content combining with procedural content for great results.

    • A large part of what my early research focused on was this exact point, and I found that you surprisingly need very little at all to get this effect. Basically, any variation at all, even unintentional, if considered will be grasped onto by the brain as intentional intelligence. I wouldn’t expect Valve needed to tell location stories at all to make the AI Director work – like you point out, just putting them in a tunnel and using base AI is enough. Or around a corner, or over near a lamp post, or even right in front of you.

      We’re amazingly manipulatible creatures. It’s the thing that originally convinced me procedural narrative would actually work. Because even when I got it wrong, at the worst players would ignore it and at the best players would do their darnedest to perceive you as getting it right. Having worked on character AI for so long, where the slightest mistake means harsh criticism, I found it rather refreshing.

  10. Pingback: Dodo Blog » Blog Archive » IAnim - Valve

  11. Hi, I think your blog might be having browser compatibility issues.
    When I look at your blog in Opera, it looks fine but when opening in Internet Explorer, it has
    some overlapping. I just wanted to give you a quick heads up!
    Other then that, fantastic blog!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s