Permadeath Horror Game: The Designer’s Trick That’s Harder Than It Looks

The first time a horror game permanently killed a character I liked and didn’t bring them back, I was playing Until Dawn for the first time and I lost Matt about three hours in. He fell from a radio tower. I was tired. I made a bad QTE. The game saved over my save and Matt was gone for the rest of the playthrough.

What surprised me wasn’t the death. The death was telegraphed. What surprised me was how the game kept going. Other characters mentioned him. There were scenes that should have had him in them and didn’t. The story bent around the empty space. It didn’t acknowledge the empty space, exactly — it just moved on.

That’s the trick. That’s the thing permadeath in horror games is supposed to do. And it is much, much harder to design than it looks.

The naive version

Here’s how most people imagine permadeath horror games work: there’s a script, you fail a moment, the script removes that character, you continue. Simple.

It is not simple.

If the character has lines in later scenes, those lines have to be cut. If they have to interact with another character in act three, that interaction has to be rewritten or removed. If they’re load-bearing for a plot point — they hold the key, they explain the backstory, they kill the killer — you need contingencies. You need other characters who could plausibly do those things. You need scene branches that work whether or not they’re present.

The amount of writing that goes into making a single playable death “just keep going” is enormous. Until Dawn has eight characters. Eight playable deaths. Combinatorial explosion if you try to handle every possible subset.

How Until Dawn actually does it

Reverse-engineered from playing it three times: Until Dawn keeps a list of “required” characters for each major scene, with fallbacks if the required character is dead. The fallbacks are themselves rank-ordered. Mike is required for the cabin sequence; if Mike is dead, it’s Chris; if Chris is dead, it’s Jess; and so on. The game has a deep enough cast that there’s almost always someone available to take the load.

Each character also has a “permadeath threshold” before which their death changes the story significantly and after which their death barely matters. Hannah and Beth’s prologue deaths are baked-in. Matt’s death in act one bends the story noticeably. A character’s death in act four hardly registers — the third act is already in motion.

This is craft. This is the part no one writes about.

How The Quarry does it

The Quarry uses a similar approach but with nine playable characters and a tighter run-time. The result is that the late-game can feel a little perfunctory if you’ve lost two or three early — there aren’t quite enough remaining characters to support the third act, and the game leans on the survivors more heavily than the writing wants to support.

This is the bound. Permadeath horror games scale with cast size, and at some point you run out of warm bodies to carry the story.

How The Walking Dead Season One does it (badly, on purpose)

Telltale’s seminal entry mostly cheats. Choices that look like they’ll change the cast usually don’t. The deaths that stick are usually scripted. The “I made this choice and the game respected it” feeling is real for maybe five moments across the season. The rest is theater.

This isn’t a criticism. The Walking Dead’s emotional weight comes from the five moments that stick. Telltale knew the audience couldn’t tell the difference between five real branches and forty fake ones, and the game runs on the five.

How Detroit: Become Human does it (most ambitiously)

Three playable protagonists, all of whom can die at multiple points. The game accommodates and shows you the flowchart of what you missed. The full-permadeath playthrough — where all three die — is a real ending the game supports. Most players never see it.

Detroit is the most ambitious permadeath horror-adjacent game ever made. The flowchart at chapter end is the design statement: here is everything that could have happened, here is what did, here is what you missed.

The contract

Permadeath horror games require honoring three rules:

1. No reload. The death has to stick. If the game lets you reload to before the death, the design collapses. Until Dawn auto-saves over your previous save. You have to live with it. That’s not a UX problem. It’s the genre.

2. The cast must be deep enough. If you can’t lose three characters and still have a story, your cast is too small for permadeath. Six is the minimum. Eight is comfortable. Three protagonists (Detroit) only works because the protagonists are doing different storylines.

3. The deaths must be earnable, not random. A QTE that’s impossible isn’t permadeath. A QTE that punishes inattention is. The game has to respect the player’s understanding of why the death happened, or the death feels like the game cheating. Until Dawn’s deaths are mostly earnable. The Quarry’s are mostly earnable. House of Ashes has at least one death that I think isn’t, and it’s the entry I argue with most.

What I’m trying to build

I’m a solo developer working on The Bone Game, the first title under TSWT Presents. Six teenagers, a Friday night, a story told from a treehouse. The permadeath structure is meant to honor all three rules above — the cast is six (workable), the deaths are earnable through clear input failures, and the saves write through.

The added wrinkle: the narrator can die too. In most cinematic horror games, the framing character is safe — Until Dawn doesn’t let the lodge itself fall into the snow. In The Bone Game, the narrator can be lost mid-story, and the story has to be passed to one of the listeners. Or it can be abandoned. Or it can keep being told to a smaller and smaller audience.

The structural challenge of letting the narrator die is what I think is interesting about the form. It’s also what I’m spending most of my development time figuring out.

If anything here resonated, you can put The Bone Game on your wishlist. The companion post is Everyone Can Die Horror Game — which is more about the player experience of total-loss endings. This post was about the craft. That one is about the feeling.