I've been thinking about a taxonomy of Roguelikes that should help us speak more clearly about this genre - or group of genres - that we love. I'd rather do this than just call things "roguelites", which basically doesn't mean anything. So here we go!
True Rogues: you're alone in a dangerous, randomly-generated dungeon, moving one turn at a time (except for speed-altering mechanics), with the possibility of permadeath always looming. Less objectively, these games tend to be more dangerous up-front, and require the player to master the mechanics in early levels - while still ramping up the threat for players who survive to later floors. Rogue, Brogue, Nethack, Jupiter Hell, and DCSS all fit here.
Bandlikes: inspired by Angband. Distinguished from True Rogues by the presence of one or more "towns" - places of safety that allow you to recover or improve outside of danger - with the attendant "town portal" abilities to get you there & back easily. This results, quite deliberately, in a longer "run". Also they tend to ease the player in - early floors have a lot of weak monsters designed to pad the player's early experience levels. I'd put Caves of Qud and Tangledeep (on hardcore mode) here.
Mystery Dungeons: think Shiren, or basically any console Roguelike. Take the mechanics of a True Rogue, but add some degree of meta-progression which can lead to an all-but-guaranteed win over time. Outside of official Chunsoft-made Mystery Dungeon games, I'd also put Nippon Ichi's ZHP and Guided Fate Paradox here.
Action Rogues: you still get random dungeons and permadeath, but now in real time! For whatever reason, these games tend to have "variety" meta-progression - you can unlock new features that don't objectively make things easier, but add more variation to future runs instead. Spelunky, Gungeon, 20XX, Streets of Rogue, and Necrodancer fit here.
Coffee-break Rogues: seemed to be all the rage a while back, but I haven't heard about them recently. These are one-floor dungeons with still enemies, where figuring out the ideal way to have your character approach each encounter is the key to success.
Cardlikes: focused on card-based battles, with dungeons generally (but not always) abstracted into icons for fast traversal. Slay the Spire is the most famous example, and I'm enjoying Dicey Dungeon here too.
Darkest Dungeon clones: basically Darkest Dungeon and the games which clearly want to be regarded as like DD. Vambrace: Cold Soul and Warsaw come to mind, since they're in my library.
Grinders: having only random dungeons, and no permadeath - or at least the ability to reload a save in case of defeat - I sometimes see these discussed in RL communities. Dragon Quest Monsters 1, Pokemon Mystery Dungeon (on its main quest anyway) and Lufia: The Legend Returns are the best examples. I'd also put Rogue Legacy here since the grinding basically obliterates any concept of loss from death.
I think in some cases a game can fit multiple terms - Rogue Legacy is an Action Rogue and a Grinder, Diablo (on Hardcore mode) is a Bandlike and an Action Rogue, Tainted Grail is a DD clone and a Spirelike, and One Step From Eden is a Spirelike and an Action Rogue. Most Mystery Dungeon games have True Rogue modes or bonus dungeons outside of the main experience, too.
There's a few games that I can't quite classify yet - Into the Breach and Dwarf Fortress, mainly - but there's always room for improvement.
I think this could help us when presenting new games to the community. Any thoughts?
Word of warning: systematically classifying video games is HARD. It’s a bit like classifying any form of creative media: music, cinema, visual arts, etc. It’s hit-or-miss. RPG forums routinely fall into that rut and the infamous corollary: [insert game here] is (or is not) an RPG.
If you’re dead set on this endeavour, I’d suggest identifying main features and tagging games with a number of them. Try and pick required ones if possible. Or don’t, because gate keeping sucks. If you know how to code, this is sort of the Composition over inheritance mindset.
I agree with this methodology, and it's reminiscent of how traditional roguelikes are defined here. I've used a similar approach in my own endeavor of defining incremental games - define a canon, find the qualities they share, and indicate which ones seem most important to have.