2022-05-04 edit: credit where credit is due: it seems like Steam has updated some of this data, so “The Problem” section is a little bit fixed. The ideas explored in “A Linguistic Solution” are still relevant, though. Original post follows.
I was super excited to see Steam’s most recent themed sale around Roguelikes. I’m a big fan of the genre and was hoping this would be an opportunity to signal boost some great, but lesser-known games like Cogmind, Jupiter Hell, or Golden Krone Hotel. Heck, they even put up a blog post of their own trying desperately to define the difference between those more niche games, and more popular ones like Hades, Slay the Spire, or Noita.
The origin of the term “Rogue-like” and how it evolved prior to 2011 is probably worth a whole separate post. For the moment we’ll focus on how Rogue-likes are defined today.
According to Steam’s sale blog post, Rogue-like games that follow traditional inspiration from the 1980 hit Rogue and include these elements:
- Random map generation
- Perma death
- Turn-based combat
- Grid-based movement
- Complexity to allow multiple solutions
- Non-modal (all actions can be performed at any time)
- Resource management
- Hack-n-slash combat
When we flip into the “Roguelike” section of the sale we see games that use that list more of an optional “or” criteria as opposed to a strict “and”.
Note: this comparison image has since been patched, but there still exists a problem of thematic overlap addressed in the next section.
The left and right hand sides here are (in theory) separate pages for the “Roguelike” and “Roguelite” genres. But if we abide by the blog post’s own definitions, we have some issues. Slay the Spire does not use grid-based movement. Noita is not turn-based. Gunfire Reborn is neither of those! The venn-diagram overlap between Dead Cells and Rogue are almost mutually exclusive in their play style, and presentation. Not every game with resource management is therefore a Rogue-like. The tab for Rogue-lite would fit this looseness better, which is what we see when we flip over there and see a 76% overlap. Why have two separate categories at all?
The failing here isn’t so much Steam’s fault, or the wild west that are user tags, but probably more of the English language. What is the difference between a -like and a -lite, really? In theory, the -lite version would be the more casual one and more on the scale of X-com instead of Nethack. The missing element from the front page steam sale is the, maybe deliberate?, exclusion of Traditional Roguelike-tagged games.
These are the type of games I would associate with a 1980 mainframe freeware game with keyboard only controls. Only, these are better. Yeah the complexity of the graphics aren’t as detailed, but there’s an unstated element of deep appreciation for a game that makes me use my imagination more than usual when gaming. These are the games that should be in the “-like” tab of the sale.
A Linguistic Solution
So what’s the solution here? We need to up our language game. A “Traditional Rogue” versus an “Action Rogue”. This does away with the confusing similarity of -like vs -lite. A game that is a “Traditional Rogue” is what it says on the tin: follows the guidelines set by a 1980s mainframe game with grid-based movement and deeper gameplay mechanics than something like an Atari game.
2,740 games on steam have a tag of “-like”. 1,620 (59%) of those have the tag of “-lite”.
2,574 games on steam have a tag of “-lite”. 1,620 (63%!) of those having the “-like” tag.
Of 173 “Traditional Rogue-like” games, 154 (89%) have a tag containing “Rogue-like” and only 54 (31%) have a tag of “Rogue-lite”.
There’s only about a 4% difference between what a “-like” and “-lite” is by just pure tags, whereas there’s a massive 58% difference between those and a “Traditional Rogue” game.
I mean we even have the grandpappy Rogue itself up on steam, and the similar games are an ok list, but not as close as I would have expected.
I mean how similar is Neon Abyss to Rogue anyway? One way to do that is with cosine similarity. We have a distribution of tag data for game A, and compare it to game B. The more similar, the closer the result is to 1.
Co-signing the Category
This was a fun evening coding project to add this to my list of functions of Steam_Data_Science. Using this definition we can then come back to the games in the “-like” tab of the sale and really calculate how close they are, indeed, to the OG Rogue:
At the end of the day, Steam should replace the “Rogue-Like” tab in the sale with the “Traditional Roguelike” tag instead. This would generate a lot more variability between the two lists, be more accurate since the “-Lite” tag is almost moot, and help some lesser-known games get into the spotlight.
The take home message here, I suppose, is that like Metroid, Doom, Castlevania, Dark Souls, and Slay the Spire, if your game is ground breaking enough and popular enough, you can get your own category of inspired games dedicated to it. Let’s just hope the linguistics majors can give us a much needed helping hand to classify them correctly.
Post title art via Zyalin on Steam