Back in the mists of time — okay, the early 2000s — I put my nascent banking career on hold and briefly became a music journalist, working for a lifestyle and clubbing website as music editor. It was a wonderful time, until it wasn’t anymore, but 20 years later I naturally only remember the good bits. In my head, it was one 18-month-long party.
I think the taxonomy trouble you're running into is that most genre names tell you something about the core verbs available to the player in moment-to-moment gameplay. In platformers you run and jump; in shooters you shoot; in strategy games you issue orders, etc. But the modern usage of 'roguelike' functions differently, in that it describes how the game is structured, and can be combined with just about any set of player verbs, leading to your confusion.
It didn't always function like that. The Berlin Interpretation that you mentioned has a bunch of factors that describe a specific subtype of RPG (single-character, grid-based, turn-based, overhead view, inventory management, etc.). Lots of great games in that mold have been and continue to be made. I generally call them classical roguelikes for clarity.
But the modern indie roguelike explosion throws most of the Berlin Interpretation factors out the window in favor of focusing on two main principles.
First is no takebacks -- the idea that whatever happens, you live (or die) with the consequences rather than having the safety net of being able to reload a save or checkpoint and try the same exact situation over again until you get it right. This is often referred to as "permadeath", but I think that's misleading because there are so many other possible consequences besides death.
Second is procedural generation of each run -- that whenever you start, you'll gain access to a unique subset of possible resources that you can wield against a unique arrangement of obstacles and enemies.
Together, these provide an effective way of making single-player games tense and unpredictable, and making the player learn systems and interactions rather than memorizing sequences of actions to repeat.
My recollection is that Spelunky showed everyone how those principles could be effectively applied to the core verbs/gameplay of the platformer genre, Desktop Dungeons and Binding of Isaac did the same for puzzle games and twin-stick shooters, and then everyone was off to the races.
Dying a lot is a bit of a red herring in this discussion. It's often found together with a roguelike structure because keeping the length of each run relatively short plays to the strength of the structure (shuffling components into unique and hopefully interesting configurations to engage with). But plenty of very hard games are fully authored rather than procedurally generated, like your example of Dark Souls. And something like Pokemon Mystery Dungeon actually sticks pretty close to the Berlin Interpretation of roguelikes, while still being pretty easy. So high death rate/difficulty is neither necessary nor sufficient.
tl;dr: 'roguelike' is best understood as a game structure that can apply across many genres, not as a genre in and of itself.
I think the taxonomy trouble you're running into is that most genre names tell you something about the core verbs available to the player in moment-to-moment gameplay. In platformers you run and jump; in shooters you shoot; in strategy games you issue orders, etc. But the modern usage of 'roguelike' functions differently, in that it describes how the game is structured, and can be combined with just about any set of player verbs, leading to your confusion.
It didn't always function like that. The Berlin Interpretation that you mentioned has a bunch of factors that describe a specific subtype of RPG (single-character, grid-based, turn-based, overhead view, inventory management, etc.). Lots of great games in that mold have been and continue to be made. I generally call them classical roguelikes for clarity.
But the modern indie roguelike explosion throws most of the Berlin Interpretation factors out the window in favor of focusing on two main principles.
First is no takebacks -- the idea that whatever happens, you live (or die) with the consequences rather than having the safety net of being able to reload a save or checkpoint and try the same exact situation over again until you get it right. This is often referred to as "permadeath", but I think that's misleading because there are so many other possible consequences besides death.
Second is procedural generation of each run -- that whenever you start, you'll gain access to a unique subset of possible resources that you can wield against a unique arrangement of obstacles and enemies.
Together, these provide an effective way of making single-player games tense and unpredictable, and making the player learn systems and interactions rather than memorizing sequences of actions to repeat.
My recollection is that Spelunky showed everyone how those principles could be effectively applied to the core verbs/gameplay of the platformer genre, Desktop Dungeons and Binding of Isaac did the same for puzzle games and twin-stick shooters, and then everyone was off to the races.
Dying a lot is a bit of a red herring in this discussion. It's often found together with a roguelike structure because keeping the length of each run relatively short plays to the strength of the structure (shuffling components into unique and hopefully interesting configurations to engage with). But plenty of very hard games are fully authored rather than procedurally generated, like your example of Dark Souls. And something like Pokemon Mystery Dungeon actually sticks pretty close to the Berlin Interpretation of roguelikes, while still being pretty easy. So high death rate/difficulty is neither necessary nor sufficient.
tl;dr: 'roguelike' is best understood as a game structure that can apply across many genres, not as a genre in and of itself.