Skip to main content

The Stealth Letters, Part Two

A Good Hiding

The fraternity of stealth impresarios turn their brains towards the question of stealth multiplayer, with contributions from Andy 'Monaco' Schatz, Raf 'Dishonored" Colantonio, Patrick 'Splinter Cell: Blacklist' Redding, and questions directed by the handsome Nels 'Mark Of The Ninja' Anderson. The first part of this developer round-table article is here, and definitely worth a read

Nels Anderson, Lead Designer of Mark of the Ninja

Hi again all,

Wow, quickly, I just want to say this is ending up even more interesting that I had hoped. Thank you all again for taking the time to share your thoughts. I'm still humbled you all agreed, but I'd say it seems to be bearing fruit!

Okay there seem to be a couple interesting threads we've touched on. (Well, far more than that, but I don't want to turn this into a tome) Raph, you mention ensuring that level spaces in Dishonored were flexible in terms of the player's approach. That was an explicit goal for us in Ninja as well. sidescrolling 2D means we kind of have "less" space to work with anyway (at least not without creating giant, sprawling maps that are not only absurd to build, but are tremendously easy to get lost in), so we basically had to make any particular space very rich in terms of how players could interact with it. Any particular object or opportunity could be manipulated a number of equally valid ways depending on the player's inclinations. It sounds like that's the approach you all took with Dishonored as well.

But what I find tremendously rewarding as a designer, and I'd be curious if you all feel the same way, is when a player is able to combine various rules and interactions in ways that you never predicted but are totally valid. Not in like a "They found a crazy bug" way (although that can be great too), but simply, "Wow, I never thought of doing that." Creating a series of interconnected systems that results in unpredicted but valid outputs, and outputs that support the way players be expressive, couldn't be more satisfying for me. I'm not sure if "honest" is the right word for it, but the experiences created in this way feel more genuine or organic somehow. It's not a scripted, fixed moment that will play out the same way for every player ever time. The experiences created are more fundamentally theirs and almost more real, in a way.

The thing is, it seems hard to be really deliberate about creating that kind of experience. I mean, it's a design goal, of course, but it doesn't feel like there are exact, known construction steps that can be taken to arrive at that player experience. Do you all find that to be the case? It seems that it's just a matter of 1) providing a lot of richness in what your objects/environment can "do" and 2) being open to adding more on top of that. Saying "yes" to what the player wants to do, if you will. And figuring out those "yes"es by just playtesting the everliving hell out of the game as its being built. I think Andy's observation of "complex but predictable" is tremendously important there as well. If the systems aren't consistent, it's very difficult for players to have any confidence in being able to manipulate them.

But maybe you're all far more regimented than I? I'm curious as to what your processes are to actually get at the kind of flexible, player-centric experience it seems we all wanted to created.

And I guess we'd be neglectful to not talk about multiplayer in stealth games! I find the multiplayer in the recent Assassin's Creed games to be fascinating and tremendously enjoyable because it's about subtle behaviour, rather than being invisible. It still manifests a satisfaction very similar to that provided by "hiding" stealth gameplay, but in a very different way. It ends up feeling even more cat & mouse than I imagine actual hiding/finding multiplayer stealth gameplay would. Obvious Chris Hecker is really focusing on this in Spy Party, but via asymmetrical player roles and 1v1 dueling gameplay. Do you think there's still fertile ground to be tilled here?

Andy, the barely-controlled chaos of the multiplayer Monaco sessions I've played are so definitive, it almost feels like playing that game solo would be like playing a different game. Do you think the experiences are actually that different? Was considering single player vs. multiplayer a large consideration in your design process?

And Pat, if you'll forgive me, I'd definitely say that the co-op Splinter Cell: Conviction, which you lead, was the strongest part of that game. And I remember your excellent GDC talk that year about designing for co-op. Have your thoughts changed any since? Or more simply, was there anything you all learned doing the SC:C co-op that was counter-intuitive or surprising in terms of providing rich stealth gameplay? Do you think the presence of another player is a great source of rich interaction and potential? Does it also add extra informational complexity, since both players now need to be aware of both their own state and their partner or partner's? Perhaps the presence of another player is indeed a source of complexity and potential chaos, but that chaos can turn into some very interesting moments, almost a la FC2's chaos leading to improvisation.

This will be my last letter, so thank you all again for participating and I really look forward to reading all your thoughts! Best,


Patrick Redding, Game Director on Splinter Cell: Blacklist

Nels, it's true that the other player is a source of noise in the system and makes it harder to read (and easier to perturb) the stealth conditions. But if you look at two analogous situations, one in single-player where your AI "buddy" messes up and causes the enemy to detect you, and one in co-op where your best friend spectacularly miscalculates and the two of you have to beat a hasty escape… Is there any doubt which one feels frustrating and which one feels like amusing hijinks?

I also remain convinced that co-operative games have an easier time maintaining suspension of disbelief because of the social component. For example on Conviction we were free to use presentation layer elements like the teammate silhouette that would have been declared too gamey for the single-player mode. Other players can be a source of distraction, but that also means they can distract you from the game's inconsistencies and can entertain you in a way that is utterly personal and unique.

The co-op designer's goal is to emphasize the subset of game features that are magnified by the presence and contribution of the other player(s). The lesson there is that when the core systems are quite deep, multiple players will naturally seek out mechanical ways of supporting each other; but if the game is relatively shallow, there won't be much potential for interesting complementary or asymmetric play and you'll be left with a "double-solo" or "triple-solo" dynamic.

The main conclusion I took from watching people play Conviction co-op was that we should try to design co-op mechanics that were intrinsically asymmetric that players could use expressively anywhere in the map. That encourages them to develop collaborative strategies and viable playstyles that the designer didn't necessarily anticipate.

On Conviction we gave players the ability to buff each other's combat effectiveness and survivability, but stealth was left as largely an individual player's problem. Players could share marks and piggyback on executing those marks. But there was nothing one player could do to actively make the other stealthier -- except deliberately get detected to trick the AI into investigating their Last Known Position while the teammate slipped away. That's a valid strategy, but it occupies a pretty shaky definition of stealth. In hindsight I would have wanted to open that possibility space up more.

Raph, you touched on the challenges of letting the player float between the extremes of the stealth-action spectrum, and not simply baking the stealth path and/or combat path into the level design. On Blacklist we've been wrestling with those kinds of level design considerations and also how to position the systemic AI as an equally effective foil to both the stealth player and combat player.

In classic SC, the relative vulnerability of the player meant that even a small number of patrollers were a source of tension and a serious threat if the player was detected. Once we expanded the player abilities to include faster movement and accessible weapon play, we needed to reconcile the older AI model with something more action-friendly.

When the player enters a new area of the game, they are by default undetected. As Nels referenced in his first letter, the world and everyone in it is oblivious to this trespasser; and at least in our case, unaware of what the player intends to do to them. By default, the initial state of the world and its AI needs to provide challenge and tension for the traditional "ghost" player, who wants to complete their game objectives while leaving no trace. But from that position, players can reorient tactically towards the complete elimination of the enemy while still remaining undetected. That means the AI behaviour needs to dynamically create windows of opportunity for the player to strike from their hiding place (shadows/cover/concealment) and vanish leaving bodies in their wake. The AI has to flow back and forth between these two roles without seeming predictable.

On Dishonoured, it seems like you would have even more complex requirements for the AI because of the addition of supernatural powers that are specifically designed to manipulate the enemies and the immediate environment. Raph, did you find that you wanted to lean towards simpler models of physical stealth as a result so that players would have an easier time reading the output of their powers?

Raphael Colantonio, co-creative director of Dishonored

Hey Patrick,

We went through many iterations with our stealth model, trying different approaches, both from the perspective of the AI and from the perspective of the player's visibility conditions.

Eventually what worked best for us was to go with a model which relies on a mix of occlusion and shadow. But here is the interesting bit: from the alpha testing, we've noticed that players understand hiding in Shadows at a distance, but they don't get it at short distance: there is something that breaks their affordance. We kept on hearing: "why is the AI blind when I'm 5 feet away from it?". Eventually we realized they had a point, most of all given our lighting model which is fairly soft. As soon as we added a rule to account for the shadow only at a distance, people started to feel the stealth was more intuitive.

To your point about the powers: We see powers as an added layer of gameplay and context modifiers, and our general purpose AIs were just following the rules of simulation, so they kind worked out "magically". For example, when the player uses Blink (the power that teleports the player forward), the AI reacts in a consistent way based on the fact that they briefly saw something.

Thanks guys for that conversation, lots of fun to discuss this fascinating topic and its challenges.


Andy Schatz, creator of Monaco

So I guess it's left to me to wrap up the conversation in some coherent way... but I would honestly prefer to leave the subject with a question rather than an answer. What is a stealth game? By the same token, what is a game genre?

One of the most consistent pieces of feedback I've encountered while running the beta for Monaco is that Monaco is "not really a stealth game" because most levels are not intended to be ghosted. Levels are intended to have a rise and fall of sneak and chase that swings back and forth 5-10 times over the course of 10 minutes. So is Monaco not really a stealth game? Certainly in Single Player we find that people tend to play more stealthily, while in multiplayer there tends to be a lot more mayhem -- probably because mayhem always trumps stealth, and so thieves tend to relax their behavior and accept more risk, knowing that if they don't, someone else will.

But I would argue that what truly defines stealth are not the mechanics established a decade or two ago by Metal Gear and Thief, but the emotions evoked by playground games like Hide and Seek. Stealth games are the perfect combination of emotion and strategy in that you are primarily rewarded for outsmarting your enemies, but emotional tension rises and falls to much greater heights than pure strategy games. Skill plays a smaller role than most FPS games, while your observational powers are tested at a much higher level.

The challenge of designers of stealth games is in rewarding smart play above all else, without the decision making ever losing touch with the reptilian center of our brains. Fear, Survival, the chase, and the hunt all tap into our animal sides, but in the quiet moments, we are asking the player think, to strategize, to exercise their intellect.

And with that, I return to fixing bugs! Thanks to all of you, especially Nels, for making this happen! I'll be seeing you around industry events I'm sure... as long as you aren't standing still in that shadow in the corner with your hood pulled up acting like you are totally not suspicious because why would anyone look twice at a guy in a cloak standing in the shadows?

And that concludes The Stealth Letters. Thankyou to everyone who contributed.

Read this next