What Makes a Roguelike? The Definition, the Argument, and a Test That Settles It

Ask what a roguelike is and you will get two answers that both sound authoritative and contradict each other. One describes a turn-based dungeon crawler with a grid and a text display. The other describes a real-time action game with automatic weapons and a level-up screen.

Neither person is wrong. The word is doing two jobs — naming a specific game from 1980 and naming everything descended from it — and almost every argument about the term is really an argument about which job it is doing at that moment.

The game the word points at

Rogue was written around 1980 by Michael Toy and Glenn Wichman, with Ken Arnold later contributing the display work, and it spread through university Unix systems. It presented a dungeon drawn in text characters, generated fresh each time you played, explored one turn at a time. Items were unidentified until you used them. And when your character died, that was the end — no reloading, no saved state to return to.

Those four properties were not conceived as a genre manifesto. They were sensible answers to the constraints of the machine and the sensibilities of the people writing it. Random generation meant the authors did not have to hand-design content and could not spoil their own game. Permadeath meant the save file could not be abused, which on a shared timesharing system was a practical concern as much as an artistic one.

What nobody anticipated is that the combination would turn out to be one of the most durable structures in game design, and that it would still be generating new commercial genres four decades later.

Why the definition became a fight

Through the 1980s and 1990s a family of games grew directly out of Rogue — NetHack from 1987, and later Angband, ADOM and their descendants — each vastly more complex than the original and each recognisably the same kind of thing.

The label itself came later than the games. It settled during the early 1990s out of Usenet discussions among players trying to name what they were all playing, and the term stuck largely for lack of a better one rather than because anyone was pleased with it.

That is the origin of the problem. A name chosen by committee, retroactively, for a family defined by resemblance to one specific game, was never going to hold up when the family got large.

By the mid-2000s, developers were building games that clearly inherited the structure while discarding the surface — real-time instead of turn-based, graphical instead of text, and often much shorter. Traditional players objected that these were not roguelikes. Action players objected that the objection was pedantry. In 2008 a group of developers met in Berlin and tried to write it down.

The Berlin Interpretation, in full

The result lists nine high-value factors and six low-value factors. The high-value ones carry more weight; the low-value ones are common to canonical roguelikes but not decisive.

What the list is actually for

The most important thing about the Berlin Interpretation is stated by its own authors and ignored by almost everyone who cites it: it is a scale, not a border. A game is measured by how many factors it exhibits and how strongly, producing a judgement about how roguelike it is rather than a verdict on whether it is one.

Read that way, most of the arguing evaporates. A real-time action game with procedural levels and permadeath scores on some factors and not others. That is a description, not a rejection.

Read the other way — as a checklist you must pass — the definition immediately excludes most of the games that made the word popular, which is a good sign you are using it wrong.

It is also worth noticing what the list quietly reveals. Six of the nine high-value factors describe decision quality: complexity, resource management, exploration, non-modality, turn-based play, tactical challenge from the low-value list. Only two describe structure. The definition is mostly an argument that a roguelike is a game about making informed choices with incomplete information, and the dungeon is a delivery mechanism.

Roguelike, roguelite, and what a loss costs

The modern split is cleaner than the historical argument, because it hangs on a single question: when you lose, does anything survive?

A strict roguelike sends you back to exactly where you started. Your knowledge improves; nothing else does. The next run is materially identical in difficulty to the last one, and every gain is in your head.

A roguelite keeps something. Currency banked between runs, characters unlocked, upgrades that persist, new items added to the pool for future attempts. Losing still ends the run, but it leaves a deposit. The term spread through developers describing their own work rather than through critics — Rogue Legacy notably billed itself as a genealogical rogue-lite — and it caught on because it named something players could feel.

Some writers proposed less loaded alternatives, "procedural death labyrinth" among them, on the reasonable grounds that naming an entire genre after one 1980 game is odd. None of them stuck. Roguelike and roguelite won, and the language is not going to be tidied up now.

The trade between them is real and worth understanding as a player:

• Strict keeps every run honest. Nothing between you and the game except what you have learned. It also means a bad losing streak feels like pure loss.
• Lite guarantees forward motion, which makes a losing run tolerable and keeps people playing. It also risks turning difficulty into a waiting game, where the answer to a wall is not "learn it" but "grind past it."

Neither is better. But the second one has a failure mode worth watching for, because a game where persistence solves everything is not really asking you to make decisions any more.