Why Roguelikes Went Mainstream and Roguelites Ate Them

In 2008 a group of developers met in Berlin at the International Roguelike Development Conference and tried to write down what a roguelike is. They produced what everyone now calls the Berlin Interpretation: a set of high-value factors — random environment generation, permadeath, turn-based play, a grid, non-modality, complexity, resource management, hack-and-slash, exploration and discovery — plus a handful of lesser ones. It was careful, it was argued over by people who had spent decades in this stuff, and it was obsolete on arrival.

Because 2008 was also the year Derek Yu released Spelunky, which is real-time, runs on a physics simulation, has no grid you’d recognise and no turns at all, and which everyone on earth immediately called a roguelike. The definition lost in the same twelve months it was written. Sixteen years later the word means something the Berlin room would not have voted for, and the interesting question is what got traded away in the meantime.

What the original claim was

Rogue, 1980. Michael Toy and Glenn Wichman, a BSD machine at UC Santa Cruz, later Ken Arnold’s curses library making the whole thing portable and Jon Lane’s work alongside. It shipped with 4.2BSD in 1984, which is how it colonised every university in the world: it came free with the operating system, and every student who found it lost a term to it.

The claim underneath it is small and severe. A run is a complete object. You bring nothing in. You take nothing out. The dungeon is generated fresh, your character starts with a mace and a bit of food, and when you die the whole apparatus is gone — the levels, the amulet, the pet dog, all of it. The only thing that crosses from run to run lives in your head.

Everything the genre did for the next twenty-five years honoured that. Hack (Jay Fenlason, 1982) into NetHack (1987 and still going). Moria (1983) into Angband (1990). ADOM (1994). Dungeon Crawl Stone Soup (2006). These are enormously complicated games with the same tiny contract at the centre, and they are healthier now than they have ever been — free, deep, under active development, played by people who will cheerfully explain why your Gnomish Wizard died. The actual roguelikes never went anywhere. They just stopped owning the word.

What made the word portable

Two things happened at once around 2008–2012.

Spelunky proved the contract survives a genre transplant. Take the run-is-the- only-copy rule, put it in a platformer with a shopkeeper and a physics engine, and the thing that made Rogue matter turns out to be the contract, with the turn-based grid as an implementation detail of 1980 hardware. That’s a real discovery and Yu deserves the credit for it.

Then The Binding of Isaac (Edmund McMillen and Florian Himsl, 2011, a Flash game made in about three months) and FTL (Subset Games, 2012, funded on Kickstarter) proved the contract sells. Both are cheap to produce, enormous in replay value, and structured so a run fits in a lunch break. Steam had just become the place everyone bought games, and a genre whose content cost is a generator rather than a level designer is exactly what a small team can ship into it. The economics and the aesthetics agreed for once.

Then Rogue Legacy moved the money

Cellar Door Games shipped Rogue Legacy in 2013 and did the thing that ate everything.

When you die, you don’t restart. Your heir does — a new character with a randomly generated trait, sometimes useful, sometimes colour-blindness, sometimes vertigo that flips the screen. You keep the gold. You spend it at the manor on permanent upgrades: more health, more classes, better armour, forever. And here’s the piece of design that shows Cellar Door understood exactly what they’d done — Charon stands at the castle gate and takes every remaining coin before you go in. You cannot hoard. The gold is worthless as savings and only exists as an upgrade, so the game forces the spending it built its retention loop on.

The word for it settled as roguelite, and what it does is subtle. It doesn’t make the game easier in any given moment. It relocates where progress is stored. In a roguelike, progress lives in the player’s head — you get better, the game doesn’t move. In a roguelite, progress lives in a database, and the run becomes the mechanism by which you extract it.

That relocation is why the genre went from a Unix curiosity to a Steam category with thousands of entries. It solved the recruitment problem. A roguelike asks you to lose a hundred times and be grateful for the education. A roguelite lets you lose a hundred times and shows you a number going up. One of those is a harder sell than the other, and the market has been extremely clear about which.

The honest version

Which brings me to Hades, and to why I’ve no interest in playing purist about this.

Supergiant shipped it in 2020 after two years in early access and it took a BAFTA and, in the year the category existed at all, a Hugo. What it does with meta- progression is the only fully honest use of the mechanic I’ve seen: it spends it on narrative. You die, you wash up in the House of Hades, and there are new lines waiting — from your father, from Nyx, from Achilles, from whoever you’ve been giving nectar to. The dying is how the story is delivered. Every other game in the genre treats death as an interruption to be compensated for with a permanent stat; Hades treats it as the page turn.

That’s meta-progression paying its own way, and it’s rare. Hades II went into early access earlier this month, which means the studio gets to argue with its own solution in public, and I’d rather watch that than another manor full of upgrade nodes.

Dead Cells (Motion Twin, 2018) is the other one worth respecting, because it keeps the skill claim alive. The permanent unlocks mostly widen the pool of weapons that can drop rather than raising your numbers, and the Boss Stem Cells put difficulty up as you progress — so the reward for mastery is a harder game. The roguevania blueprint works because Motion Twin resisted the obvious move.

What happens when the contract meets a full-price game

Returnal (Housemarque, 2021) is the best natural experiment the genre has run, because it took the roguelike contract and put it in a seventy-quid PS5 exclusive with a motion-captured cast and a Hollywood budget.

The contract came under immediate strain, and the argument was public. Runs lasted hours. There was no way to suspend one — close the game, lose the progress — and a console has a queue of other things wanting your telly. Players who had happily lost a hundred FTL runs at fifteen minutes each found that losing a three-hour run at full price felt like being robbed, and said so loudly enough that Housemarque patched in a suspend feature within months.

That’s the contract meeting economics and losing. A run is the only copy, which is a fine bargain when a run costs twenty minutes and a fiver of your money. Scale the run to an evening and the price to a night out, and the player starts demanding the thing the whole genre exists to withhold. Nothing about the design changed. The stake changed, and the stake was doing all the work.

What got traded

Here’s the cost, plainly. When progress lives in a database, a run stops being an argument about your play and becomes an argument about your hours. Lose enough times and the game hands you the win anyway, because the upgrade tree is a guarantee dressed as a challenge. The loop stops being the argument and becomes a payment plan.

You can feel it in how these games end. A roguelike win is a statement about the player: you did the thing, with the same tools you always had, and the reason it worked this time is you. A roguelite win frequently arrives on a Tuesday because your health bar is 40% longer than it was in March. Both produce a victory screen. Only one of them means anything about you, and the games that know this — Dead Cells scaling the difficulty, Charon confiscating the gold, Balatro handing you stakes that get worse as you get better — spend real design effort building the escape hatch back to the original contract.

Where this leaves the word

Nowhere useful, which is fine. Genre names are marketing that escaped, and demanding precision from them is a hobby for people who enjoy losing. Vampire Survivors (2022) is filed under roguelite and it’s an idle game with a scythe. Nobody is confused in practice.

The thing worth keeping is the contract, because it’s the part that was actually worth something. A run is the only copy. Everything good about this entire family of games descends from that sentence, including the ones that broke it, and you can tell which designers understand it by whether they built something to push back against their own upgrade tree. Rogue got there in 1980 by deleting a file. It’s been downhill and uphill ever since, usually at the same time.