Three fields, three different authors
Every store page tells you what kind of game you are looking at, and it tells you three times, in three fields, written by different people for different reasons. One of them is a form the developer filled in. One is a list of features. One is a vote. They disagree constantly, nothing reconciles them, and none of the three arrives with the chart data this site is built from.
Three fields, and who writes each one
They sit within a few centimetres of each other on the same page, which is why almost nobody notices they are different things.
- Genre — the developer picks from a fixed list. Chosen at submission from a short set of options, and the set is the giveaway: alongside Action, Adventure, RPG, Strategy and Simulation it contains Indie and Early Access. One of those is a budget and one is a release state. Neither describes what you do when the game starts.
- Category — the developer ticks features. Single-player, multi-player, co-op, achievements, cloud saves, controller support. This field is honest about itself: it never claimed to be a genre. It is also the most reliable of the three, because every item in it is a fact that can be checked by launching the game.
- Tags — players vote. The long list of words under the title, ordered by how many people applied each one. This is the field everybody actually reads, it is the only one written by the people who played the thing, and it is the least governed. Developers can seed the first ones themselves, and after that it is a public vote with no quorum.
Three authors, three purposes, one visual block. A page can carry “Indie” as its genre, “Single-player” as its category and “Roguelike Deckbuilder” as its top tag, and only the third one tells you anything about the next hour of your life.
Why the tag list is a popularity contest
Tags are ordered by votes. That one design decision produces every odd thing about them.
- The order measures agreement, not accuracy. A word that four hundred people applied outranks a word that forty people applied, even when the second word is the precise one. Precise words are precise because fewer people reach for them, so the ranking is biased against exactly the tags that would tell you the most.
- The big tags swallow everything. Action, Adventure, Indie, Singleplayer, Casual. These sit near the top of an enormous number of pages, because they are true of an enormous number of games. A field that is true of everything sorts nothing.
- Discovery money rides on it. Tags feed the store’s recommendation surfaces, so a tag is not only a description — it is a shelf. That gives everyone involved a reason to push the list toward whichever shelf is busiest, and the description is a side-effect.
- Joke tags are a genuine feature of the data. Communities organise to apply words that are funny rather than true, and the store has had to moderate the vocabulary for exactly that reason. Most of the time this is harmless noise. It is a useful reminder of what the field is: a vote, with everything a vote implies.
Six words that do not mean what the field means
Each of these is used daily by players to mean something specific, and sits in a store field meaning something else.
| Word | What a player means by it | Where it actually lives |
|---|---|---|
| RPG | Numbers go up, choices stick, there is a build | Genre list — covers everything from a forty-hour epic to a phone idler |
| Roguelike | You die, you lose the run, you keep the knowledge | Player tag — and the community has argued over its boundary for two decades |
| Souls-like | Deliberate combat, punishing, recover your losses where you fell | Player tag named after one series — no store field can hold this |
| Survival | Meters that drain, crafting, the first night is the test | Player tag, and also a genre word, meaning different things in each |
| Indie | Small team, personal, not made by a publisher | Genre field — a claim about who made it, standing where the kind of game should be |
| Early Access | Unfinished, and sold that way on purpose | Genre field — a release state, standing where the kind of game should be |
The last two rows are the ones worth keeping. They are not imprecise genres; they are not genres at all, and they sit in the genre field on a very large number of pages.
What actually predicts the number on this board
Two games with identical tags can sit a hundred thousand players apart, and two games with nothing in common can sit next to each other. The things that decide it are not in any of the three fields.
- How long one session is. A game built around a twelve-minute match holds a different number of bodies at any instant than a game built around a three-hour raid, even with the same number of people playing it that day. A concurrent count is a measure of overlap, and session length is what decides overlap.
- Whether there is a reason to open it today. A daily reset, a weekly quest, a season that ends on a date. These are scheduling devices, not genres, and they are the single most visible difference between the games that stay near the top and the games that visit.
- Whether there is a price at the door. A gate cuts the population before any of this starts, which is a large enough effect to deserve its own page on this site.
- Where the peak is. A premium game peaks in its launch week and settles. A live game peaks whenever it ships something. Reading a single afternoon’s concurrent count tells you where in that cycle you happened to look, and nothing about the kind of game.
Why this site prints no genre column
- It is not in the data. The chart response carries a rank, an id, a peak and last week’s rank. Even the game’s name has to be fetched separately. A genre column would mean a second source joined on the id.
- A second source is a second version of the truth. The moment this site prints a genre that the store shows differently, the whole table becomes something to argue with. That trade is not worth a column nobody asked for.
- It would not sort anything. A column where most rows read Action, Indie or Singleplayer groups the hundred into one large group and a tail. The two rankings already on the charts page disagree with each other by fourteen places on average, and that disagreement is information. A genre column would be decoration.
There is a version of this site that guesses a genre for every row and looks more complete. It would be more complete in the way a filled-in blank is more complete than an honest one.