|The last moment of my last Invisible Inc run on "Expert Plus"; don't read the game text if you don't want spoilers|
The last time I was this engaged by a stealth game, it was the first time I played Thief 1 (1998) in 2002... a pirated version I downloaded off Kazaa, with all the cutscenes, music, and voice-over removed to save on file size. What was left was the most avant-garde game I had ever played, a world of footsteps and silence. Between then and now -- Splinter Cells were okay but not my bag, Thief 3 was a sea of mediocrity with a single shining jewel, Dishonored was okay I guess, those bits in The Last of Us quickly wear out their welcome, and Thief 4 was rather unfortunate -- "stealth" has felt a little dead for the last 13 years.
If we look back at the systems design theory behind Thief 1, we can figure out who murdered stealth.
Levels are places, moments, and stories. We can also think of some levels as systems. (For more on this, see "Dark Past, pt. 3") In his PhD dissertation, Dear Esther director Dan Pinchbeck theorized that the function of game narrative was partly to rationalize how single player action games get less complex over time. As you clear a level of enemies or settle into your favored equipment loadout, the game destabilizes into an empty "solved" state.
In stealth games, it is common to clear a level in this way. KO'd or killed enemies (there isn't really a difference) never wake up, so once player poke a hole in a guard patrol then that hole stays there forever. Because stealth games often support some form of body carrying action, many stealth players celebrate level clearing by posing a "family portrait" full of accumulated corpses.
So, the first thing a stealth game can do to feel fresh is to figure out a good way to repopulate and replenish a level -- but that's a problem almost every 3D action game has, so let's look a bit closer at some design patterns specific to stealth:
At GDC 2006, Thief designer Randy Smith gave a really insightful presentation on stealth game systems. (For more on this, see "Dark Past, pt 4") In it, he observed that stealth games have narrow thresholds between remaining undetected and being detected. Because being detected and pursued was an undesirable failure state, their big design goal was to help players recover from failure pretty quickly. Flash bombs, magic powers, parkouring, blinking, "swooping"... these are all designed to re-stabilize the system to help the player escape.
In balancing and tuning a stealth system, Smith argues that a designer's goal is to try to expand that really narrow partial failure margin as much as possible. This is the basis behind guards running slower than you, having rather limited vision cones and slow reaction times, and quickly shrugging, "I guess it was just a rat." Smith even quotes a level design heuristic about keeping guard patrols short and fairly isolated, to help protect the player from "jumping out of the pan and into the fire."
Being detected means failure, and failure is undesirable -- this claim is the petrified baggage weighing down the modern stealth genre. So it's refreshing when Invisible Inc takes the opposite approach, constantly teasing you about your impending death. In doing so, it redefines "failure" in profound ways.
Invisible Inc. is the murderous serial killer cyborg that does not run after you. Instead, it calmly walks toward you, which is somehow even scarier. There is no pretense of "ghosting" being the optimal play style here... in many cases, you will have no choice but to be detected, to KO a guard, or to set off an alarm. Some characters become more powerful only when shit is hitting the fan. Getting detected is not failure here -- what's worse is getting detected at the wrong time.
Eventually, guards wake up and loot becomes harder to steal. The mission constantly pushes back against you, escalating the stakes -- a puzzle that is constantly threatening to unsolve itself to become unsolvable. At the same time, this game doesn't pretend the guards have sophisticated AI (guards are very predictable) because a trash compactor does not need "human-like AI" to profoundly destroy you. The floorplan itself is the AI, as late-game teams of alerted guards flood hallways like poisonous gas. Contrast this interconnectedness and guard-to-guard interaction with Thief's stated design goal of isolating patrols.
In a way, Invisible Inc. is one of the few video games about global warming. Here, failure is not a state, because that would be too easy. Instead, failure is the slow glacial process of watching your loved ones drown. You can always lose more. Unlike every other stealth game, slow and patient observation usually means slowly suffocating death here.
And this apocalyptic relentlessness is possible ONLY if you coach the player to have a different relationship to failure than most stealth games, and decouple detection from failure. One of the most beautiful things that Invisible Inc does is that halfway through a mission, you can always retreat early without actually accomplishing the scouted objective. Retreating is a common mechanic in everyone-dies tactics games like XCOM, but it is totally unheard of in most stealth games, which usually handcuff themselves to a binary mission success or failure.
Back to that last mission of my Invisible Inc playthrough, the only mission without an exit teleporter. The culmination of my 30 hours of play have all led to this moment. I have to get two VIPs through this lobby full of a dozen alerted guards, to the control room at the bottom of the map. If I don't act now, the guards will spread out and become difficult to manage AND the other guards behind me will wake up.
Here's my really clever plan:
I'm going to send my other two characters to run into the lobby. Their jobs are to (a) get caught, (b) die loudly, and (c) in the process of dying loudly, get the guards to look the wrong way and in the wrong place, so my two VIPs can slip past them undetected. The last detail to figure out is the wimpy level 1 grunt standing outside the control room, but he'll be easy to deal with -- I give VIP 1 a taser to KO him next turn.
I hold my breath and begin... and everything goes to plan. A dozen guards look the wrong way. My VIPs salute the agents' noble sacrifice, turn on their cloaking devices, and slip through. A million gears and cogs are working perfectly in concert, like one of those really expensive Swiss clocks...
... and then, at the very last second, a Swiss ant gets lodged in the last gear.
That level 1 grunt? He heard the commotion in the other room, ran out of the range of VIP 1's taser, and then stopped right in the middle of the doorway. Now VIP 2 can't get through the door, and he can't wait any longer because his cloaking device will run out in the next turn and the guard will see him. So just like that, my playthrough ends with a whimper of an overlooked detail. All the pieces were there in front of me, I just failed to connect that one last dot and predict what was really going to happen.
Technically, I haven't failed at all, and both VIPs are alive and undetected and that last guard doesn't even know I'm there. No one is screaming after me or chasing me, because this is a game that knows there's no need for it to run. This is a stealth game where the AI doesn't even know how hugely I've fucked up -- failure itself has been made stealthy.
Real failure isn't about getting spotted; no, real failure is realizing you weren't so clever after all, and maybe you were never really that clever -- and now you've just royally fucked yourself. This is the importance of disconnecting detection from failure, it allows us to have a freer and more profound relationship to failure.
Sure, fear is pretty... but existential dread is beautiful.