The short answer: a roguelite lets you keep something between runs, and a roguelike does not. That is how the words are used in practice, on store pages and in reviews, in 2026.
It is also a shorthand that the people who wrote the original definition never used — and once you know what they actually wrote, the arguments you see in forum threads make a lot more sense. This piece gives you the real source, the real history, and a test you can apply yourself.
The one-line difference most people mean
| Roguelike | Roguelite | |
|---|---|---|
| Death | Total reset — the run is gone | Run is gone, but permanent upgrades stay |
| Between runs | You keep knowledge, nothing else | You keep currency, unlocks, stat upgrades |
| Turn-based | Almost always | Usually not — typically real-time action |
| Grid-based movement | Almost always | Rarely |
| Layout | Generated fresh per run | Generated fresh per run |
| Typical examples | NetHack, Angband, Caves of Qud | Hades, Dead Cells, Slay the Spire |
If you only wanted the practical distinction, that table is it. The rest of this article is why that table is incomplete.
Where the definition actually comes from
The reference definition is the Berlin Interpretation. It was produced at the International Roguelike Development Conference in 2008 — RogueBasin describes it as "the product of a discussion between all who attended (including Ido Yehieli, Radomir Dopieralski and Jeff Lait)" — and posted to the Usenet group rec.games.roguelike.development by Jeff Lait on 16 September 2008.
Its framing statement matters as much as its criteria:
'Roguelike' refers to a genre, not merely 'like-Rogue'. The genre is represented by its canon. The canon for Roguelikes is ADOM, Angband, Crawl, Nethack, and Rogue.
It then lists nine high-value factors: random environment generation, permadeath, turn-based, grid-based, non-modal, complexity, resource management, hack'n'slash, and exploration and discovery. A second set of low-value factors covers things like a single player character, tactical challenge, ASCII display, and dungeons.
Two things are worth noticing.
First, meta-progression is not in it. The single criterion most people now use to separate the two words does not appear anywhere in the document that defines the genre.
Second, it says outright that it is not a checklist:
This list can be used to determine how roguelike a game is. Missing some points does not mean the game is not a roguelike. Likewise, possessing some points does not mean the game is a roguelike.
So a game is not disqualified by failing a criterion. The Berlin Interpretation describes a centre of gravity, not a boundary.
Where each word came from
"Roguelike" is the older term by roughly two decades. Temple of the Roguelike's history of the term dates its use to around 1993 on Usenet, where it grouped discussion of "single-user, fantasy role-playing computer games, generally set in a dungeon, run with a simple graphic interface." The original reference set was Rogue, Hack, NetHack, Moria, Angband, and to an extent Larn and Omega.
"Roguelite" arrived later, as an answer to a problem. The same history describes it: "Because of these big changes on the game format, there was a growing confusion on the meaning of the 'roguelike' term… A series of terms including roguelite, roguelike-like and action roguelike then appeared to refer to these games."
Who coined it is less certain than the internet suggests. Save or Quit's history of the two words states that the developers of Rogue Legacy were the first to use "rogue-lite" — but we could not find a primary source for that claim: no dated post, interview, or statement from the studio itself. Treat it as one outlet's attribution rather than settled fact. The same article notes that competing labels were tried and abandoned, including "roguelike-like" and "Procedural Death Labyrinths."
The disagreement that never gets resolved
If meta-progression is the popular test, it is not the practitioners' test.
Zeno Rogue — the developer of HyperRogue, a traditional roguelike — argues directly against it. The claim that roguelikes reset you and roguelites let you keep things is, in that developer's words, "a common misconception." The proposed alternative:
Roguelike is a genre primarily defined by its immediate gameplay (just like genres such as platformer or first-person shooter).
Under that reading, the split is about how a turn feels — deliberate, turn-based, strategic, versus fast, real-time, tactical — and a roguelite is any genre at all wrapped around Rogue's run structure. That is why you can have a roguelite platformer, a roguelite shooter and a roguelite deckbuilder, but not really a "roguelike shooter."
The criticism goes the other way too. RogueBasin records developer Darren Grey's article "Screw the Berlin Interpretation!", which calls the definition "inaccurate, outdated, and unrepresentative of a vibrant and open genre" and singles out criteria like ASCII display as irrelevant. (We could only reach that article through an archive link referenced on RogueBasin, so we are citing RogueBasin's account of the argument rather than the original text.)
Even engaged fans hedge. A PC Gamer forums thread on the terms lands on meta-progression as the working line while conceding that "roguelite just means 'with elements from Rogue'… it's not as precise as people think."
A test you can actually apply
Instead of one question with a contested answer, ask four. Each is independently checkable, and a game can pass some and fail others — which is exactly what the Berlin Interpretation said would happen.
- Does death reset everything? If nothing at all carries over, that points roguelike. If you bank currency, unlocks, or stat upgrades, that points roguelite.
- Is it turn-based, and does it move on a grid? These are two of the nine high-value factors, and they are the ones modern action games almost always drop.
- Is the space generated fresh each run? Random environment generation is on the list too — but note that both categories almost always pass this, so it rarely settles an argument.
- Does the run itself change, or only your character? Games where each run's rules can be rewritten — by relics, by loadout, by what you chose to build — sit in a different place from games where only the numbers go up.
Worked examples, using the four questions rather than one:
| Game | Resets fully? | Turn-based + grid? | Generated per run? | Verdict |
|---|---|---|---|---|
| NetHack | Yes | Yes | Yes | Roguelike, uncontested |
| Caves of Qud | Yes | Yes | Yes | Roguelike |
| Hades | No — Darkness persists | No | Yes | Roguelite |
| Dead Cells | No — permanent unlocks | No | Yes | Roguelite |
| Slay the Spire | Mostly — unlocks expand the card pool only | Turn-based, not grid-based | Yes | Roguelite, but the closest case |
Slay the Spire is the interesting one. It is turn-based, its runs reset, and its persistent progression only widens the pool of cards you might be offered rather than making you stronger. It fails "grid-based" and little else — a good illustration of why a nine-factor spectrum describes reality better than a binary.
Why the argument keeps happening
Because both sides are describing something real.
The people who say "meta-progression" are describing what changed commercially: after Rogue Legacy, Dead Cells and Hades, a very large audience arrived at run-based games through titles that keep something between attempts. For that audience the distinction is the one that matters at purchase time.
The people who say "turn-based, grid-based, non-modal" are describing what the word originally meant, and they are right that the modern usage lost most of the definition on the way.
A useful way to hold both: roguelike is a genre with a canon; roguelite is a structure you can bolt onto any genre. That is close to Zeno Rogue's formulation, and it explains why "roguelite" spread so far so fast — it was never claiming to be a genre in the first place.
What this means when you're choosing what to play
If you want the original thing — deliberate, punishing, deep, where a single misread turn ends a twenty-hour character — look for turn-based and grid-based in the description, and be sceptical of anything that advertises permanent upgrades.
If you want the modern thing — a tight thirty-minute loop where each run builds something and failure still moves you forward — you want a roguelite, and the permanent-upgrade screen is a feature rather than a compromise.
And if what you actually enjoy is the build — the moment a run stops being a set of numbers and starts being a machine you assembled — that is its own axis, and it cuts across both labels. Slay the Spire and Balatro sit there. You can explore that same build-first territory across the games in HowWorks Adventure, where the signature decisions change with each game rather than with the genre label alone.
If you want the wider map rather than the definition, the AI games guide covers what is playable now, and the AI RPG roundup covers the narrative end of the same replay-driven structure.
Sources
- RogueBasin — Berlin Interpretation (primary source, preserved verbatim)
- Jeff Lait's original Usenet post, 16 September 2008
- Temple of the Roguelike — A short history of the Roguelike term
- Zeno Rogue — What is the difference between Roguelike and Roguelite?
- Save or Quit — A History of the words roguelike and rogue-lite
- PC Gamer forums — Roguelike & Roguelite, what do these terms really mean?