Difficulty Is a Design Choice, Not a Moral One

Every eighteen months or so the same argument comes round, and every time it arrives dressed as a question about difficulty when it is actually a question about deserving. Should this game have an easy mode. Would an easy mode ruin it. Did you really finish it if you turned something down. The argument has a remarkable ability to generate heat while producing no design insight whatsoever, and the reason is that both sides have agreed on a premise that is wrong. They both treat the number of times a game kills you as a statement about the player.

It is a statement about the game. Difficulty is the rate at which a game hands you information, and the shape of the punishment when you fail to use it. That is all it has ever been. Once you look at it that way, the interesting question stops being whether a slider is allowed and starts being what any given slider actually moves — because most of them move the wrong thing, and the ones that work are doing something far stranger than making the numbers smaller.

The coin slot is still in there

Start with where the furniture came from. Lives, continues, the three-strikes structure that still shows up in games nobody would call arcade games: that is revenue architecture. A coin-op cabinet in 1982 needed to end your session inside about ninety seconds to two minutes, because the cabinet’s economics were measured in plays per hour. The difficulty curve was a pricing mechanism with a sprite on top.

Home computers inherited the shape without inheriting the reason. I loaded these things off tape as a kid, and the games that came down the wire were frequently arcade conversions that kept the lives, kept the punishing collision boxes, and threw away the coin slot that made them make sense. What was left was a game that would happily end after two minutes and then ask you to wait four more for it to load again. Some of that was cruelty. Most of it was inertia — a design convention copied because it was the convention, from a machine with a different business model in a different room. The long patience of the tape load did strange things to what we were willing to accept.

That inheritance matters because it explains why so much difficulty in games is vestigial. It is there because it was there. When a modern action game gives you Easy, Normal, Hard and Very Hard, and those settings adjust enemy health and player damage taken, it is performing a ritual whose origin nobody in the room remembers. Nothing about the design is different across those four settings. The same encounter happens, in the same order, with the same information, and the only variable is how many times you press the button before the corpse falls over. That is a length adjuster wearing difficulty’s coat.

What Doom understood in 1993

Doom shipped with five skill levels and gave them names instead of numbers. I’m Too Young to Die. Hey, Not Too Rough. Hurt Me Plenty. Ultra-Violence. Nightmare!, with the exclamation mark, added late in development, and guarded by a confirmation prompt admitting the setting isn’t remotely fair.

Look at what actually changes. The lower skills reduce damage and thin out the monster placement. Ultra-Violence is the intended composition — the level as designed, every monster the designer put there. And Nightmare respawns the dead, speeds up enemy movement and attacks, and doubles ammo pickups to keep you armed while it does it. That last detail is the whole essay in miniature. Nightmare rewrites what the level is. On Ultra-Violence a room is a problem you solve and then own. On Nightmare a room you cleared is a room that will refill behind you, so the level stops being a sequence of fights and becomes a route — a question about movement and about what you can afford to leave alive. The level design that still teaches teaches different lessons at different settings, because the settings are different readings of the same architecture.

That is a difficulty option worth having. It is a genuine second design, authored, with its own logic, achieved through one respawn flag and one ammo multiplier. Cheap to build, and it produces a game rather than a chore.

The slider that lies

Compare that with the standard modern implementation and the emptiness is embarrassing. Damage multipliers give you the same encounter, slower or faster. They do not change what you must know, when you must know it, or what the space is for. They are a duration control mislabelled as a difficulty control, and players can feel the mislabelling even when they can’t name it — it’s why “Hard” in so many action games feels like Normal with a sponge, and why the highest setting so often feels less like a challenge and more like an insult delivered in arithmetic.

Devil May Cry 3 produced the most instructive accident here. Capcom’s western release shipped with the difficulty labels shifted: what the box called Normal was, in effect, the Japanese version’s Hard. Western players met a game that punished them for reasons they had no framework to understand, and a substantial number concluded the game was broken, or that they were. Capcom fixed it in the Special Edition. Nothing about the encounter design was wrong. The label was wrong, and the label is a promise about what the game thinks you’re doing. Break the promise and the same content reads as hostility.

Celeste opens the box and shows you the wiring

Celeste in 2018 did the thing everyone had been arguing about, and did it so plainly that the argument should have ended. It has an Assist Mode, buried one menu deep, guarded by a text screen that says — in substance — that the game is designed as a challenge, and that if the default is inaccessible, this exists so you can have the game anyway.

The contents matter more than the gesture. Assist Mode lets you slow the game down to as low as half speed, grant yourself infinite stamina, add a second air dash, turn on invincibility, and skip a chapter outright. Those are five separate answers to five separate reasons a player might be stuck, and that granularity is the design argument. A player with a tremor needs game speed. A player who can execute but can’t read the room in time needs speed too, for different reasons. A player who understands the room and cannot land a two-dash sequence needs the extra dash. Lumping all of that under Easy would tell you nothing about any of them.

And Assist Mode is off by default, with the fiction fully intact: Madeline’s climb is about doing a hard thing badly and repeatedly, and the game never pretends otherwise. Matt Thorson’s team let those two facts sit next to each other without apparent difficulty, which suggests the tension the argument insists on may be imaginary.

Hades turns the slider into a dial that talks back

Supergiant went further and made the assist itself a system. God Mode in Hades grants 20% damage resistance, then adds 2% for every death, up to a ceiling of 80%. It is an adaptive difficulty that reads your failures and answers them gradually — you keep dying, the game keeps quietly conceding, and at some point you get through. It never announces that you’ve been carried, and the number stops moving at 80%, so you are never simply invulnerable.

The elegant part is the symmetry. The same game ships the Pact of Punishment, where you spend Heat to make runs harder — more boss health, tighter timers, fewer healing options — in exchange for better rewards. One game, two dials, both running through the same escalation machine. Hades treats difficulty as a negotiable property of the run, which is exactly what a roguelike difficulty is, and the death loop absorbs it without a seam because the loop was already the argument. The design that solved narrative repetition also quietly solved the easy-mode war, and almost nobody noticed.

The FromSoftware exception that isn’t one

Which brings us to the game the argument is always secretly about. FromSoftware ships no difficulty menu, and this is treated as a principled refusal. It is a design choice with a real justification — a single shared difficulty is what makes the messages, the bloodstains and the co-op mean anything, because everyone is describing the same wall.

But notice that Dark Souls has a difficulty slider. It’s called summoning. Bring in two phantoms and a boss gains health while the incoming damage divides across three bodies and the attack pattern loses coherence; the fight becomes materially easier. Add a shield build, a hundred spare levels from grinding, or a weapon upgrade path the boss wasn’t balanced for, and you have adjusted the numbers as surely as any menu would. The level design that folds back on itself never cared how you got through. It cared that the world was one shared object.

So the refusal is really about a shared frame of reference, which is a different thing with a different justification. Those two are separable, and the argument would be shorter if more people said so.

The actual question

Here is what I want from a difficulty option, after forty years of them. Tell me what changes. Change something real — placement, respawn, information, tempo, what the space demands. If all you’re moving is a damage multiplier, say so honestly and call it what it is, because a player who knows they’re adjusting length will make a better choice than one who thinks they’re adjusting worth.

Cuphead is worth the last word, because it did the honest version of the dishonest thing. Simple mode exists; it cuts boss phases. It also declines to give you the contract, so the game tells you plainly that you have played a shorter version of the fight and the progression won’t accept it. You can dislike that. You cannot call it unclear. Studio MDHR decided what its game was for — the full pattern, learned — and then built the door and labelled it accurately.

Difficulty is a question of what the game is for. Answer that, and the slider designs itself. Leave it unanswered, and no number of settings will save you.