Who is counting the games?

A new game appears somewhere in the world roughly every three minutes. Somebody must be keeping track.

Several people are, it turns out, and they have been at it for twenty-six years. What follows is the map: who counts games, when they started, what they are actually good at, and the one thing none of them do.

The archivists, 1999 to 2011

MobyGames opened on 1 March 1999, built by Jim Leonard and Brian Hirt, with David Berk joining eighteen months later. It remains the most careful games archive in existence. Every entry is vetted by contributors, credits are tracked down to individual staff, and coverage of the pre-1990s era is better than anywhere else because actual humans went and checked.

That care is also the constraint. A database that verifies things cannot absorb four hundred games a day. MobyGames was designed for a medium that produced a few thousand titles a year, and it does that job beautifully.

Since 2022 it has been owned by Atari SA. Hold that thought.

Giant Bomb followed in 2008, taking the wiki approach: games, characters, concepts and people all as linked entities, written by an editorial staff with a personality. It was always as much a publication as a database, and its data layer never grew into serious infrastructure.

HowLongToBeat arrived in 2011 and did something narrower and smarter than everyone else: it answered one question. How long is this game? Main story, extras, completionist, speedrun. The data comes from players reporting their own playthroughs, which makes it the first of these to run on a genuine feedback loop from the people actually playing.

The API era, 2015 to 2021

The second wave was not built for readers. It was built for other software.

IGDB, the Internet Game Database, was founded in Sweden in 2015 and did the thing the older archives never quite managed: a clean, well-documented commercial API with properly linked entities, so anyone could build a game app without assembling a catalogue first. On 17 September 2019 it was acquired by Twitch, which is to say by Amazon.

RAWG took the other route, going wide rather than deep. Over 350,000 titles across more than fifty platforms, generous free tier at 20,000 requests a month, attribution required. If you need a lot of games quickly and can live with rougher metadata, this is the pragmatic pick.

Backloggd launched on 15 July 2021 and is best understood as Letterboxd for games: not an archive at all, but a place where players log what they played and rate it. It runs on IGDB's data, which is the tell. The catalogue had become a commodity; the interesting layer had moved to what players do with it.

Alongside all of this sit the two Steam observatories. SteamDB produces the best release statistics anywhere, built by watching Valve's own systems, and offers no official API while actively turning away automated traffic. SteamSpy estimates ownership and sales, needs no key, and comes with no guarantees at all.

The shape of the hole

Line them all up and one property is shared by every single one.

They are organised around store entries.

MobyGames, IGDB, RAWG, Backloggd, SteamDB, SteamSpy: each of them assumes a game is a thing that got published somewhere, with a page, an identifier and a release date. That assumption was correct for the entire history of the medium, right up until roughly two years ago.

It is not correct any more. The 1,170 games submitted to the 2025 Vibe Coding Game Jam were required to run in a browser, with no login, free to play. Most of them never touched a storefront. For every database listed above, they do not exist. Not "poorly covered". Absent.

That is the fastest-growing part of the medium, and the entire formal record of games is blind to it.

There is a second thing worth noticing while everything is lined up. The single most complete archive of games history is owned by Atari. The best commercial API is owned by Amazon. The memory of this medium is corporate property, and it was acquired quietly, while everyone was arguing about something else.

The new wave, 2025 to 2026

So who is covering the new field? Three attempts so far, and the honest summary is that all of them are thin.

vibegames.io is a directory of vibe-coded games sorted into thirteen categories: simulators, open world, puzzle, escape room, shooters, flight sim, farming, sailing and so on. Games are playable straight from the listing, which is the right instinct. What it has no trace of is competition: no play counts, no scores, no ranking, no way to tell whether anyone enjoyed any of it. The footer identifies it as a project from a "ProfitSwarm AI Business Challenge", which tells you roughly how much weight to put on it as long-term infrastructure.

vibecoding.app compares more than 150 AI coding tools, agents and apps, with games as one category among many. Listings carry screenshots, a builder description, the stack used and a vote count, with the top projects clearing three thousand votes. It is the closest thing the scene has to a ranked list, but it ranks tools, not play.

vybeguide.ai launched on 30 January 2026 with a daily-updated database of 50,760 open source projects. Impressive as a crawl, aimed at developers looking for libraries, not players looking for a game.

And then there are the jams themselves. vibej.am carries the 2026 edition with 50,000 dollars in prizes, and the jam sites are, ironically, the best data source in the field: every entry is a live URL with an author attached, verified by the submission rules. They are also the most perishable, because a jam page is a snapshot of one fortnight and nobody maintains it afterwards.

Who is actually moving

Sorting by momentum rather than by size gives a different picture.

Genuinely moving: the jams. From one organiser's tweet to 1,170 games in a fortnight, then to a 50,000 dollar prize pool the following year. Nothing else in this space has that slope.

Moving quietly: Backloggd and HowLongToBeat, because both run on player-contributed data rather than catalogue ingestion. That is the only model here that gets better as the number of games explodes, instead of drowning in it.

Stable and consolidating: IGDB and MobyGames. Both are excellent, both are owned, and neither is going to change what it fundamentally is.

Thin: the vibe directories. Every one of them is a catalogue assembled quickly, and a catalogue is the easiest thing to build and the least defensible thing to own. There is no moat in a list.

What nobody is doing

Put the whole map together and the gaps are specific.

Nobody counts the new field. There is no running measurement of how many games are being made outside the stores, by whom, at what rate.

Nobody aggregates play. Every database records that a game exists. None of them records that anyone played it and how it went. For store games, Steam covers this. For the browser field, it is a void.

Nobody ranks across games. Leaderboard infrastructure exists, globalstats.io and various open source kits, but all of it is for a developer to install into one game. There is no table that says who is best across the field, and therefore no reason for a player to travel between games.

Nobody serves the player. Almost everything above was built for archivists, for developers, or for other software. HowLongToBeat and Backloggd are the exceptions, and both are about games you already know you want.

Directories without a contest are catalogues. Contests without a directory are invisible. The two have never been joined, and joining them is the thing worth building.

---

Sources and dates

• MobyGames: founded 1 March 1999 by Jim Leonard and Brian Hirt, David Berk joining eighteen months later; owned by Atari SA since 2022
• Giant Bomb: founded 2008
• HowLongToBeat: created 2011
• IGDB: founded in Sweden 2015, acquired by Twitch/Amazon 17 September 2019
• Backloggd: launched 15 July 2021, built on IGDB data
• RAWG: 350,000-plus titles, 50-plus platforms, 20,000 free API requests per month with attribution required
• SteamDB and SteamSpy: unofficial, no supported public API; SteamDB refused automated access when checked on 4 August 2026
• vibegames.io, vibecoding.app; vybeguide.ai launched 30 January 2026 with 50,760 projects
• Vibe Code Game Jam 2025: 1,170 entries; vibej.am for the 2026 edition and its prize pool

RAWG's founding year is deliberately absent: we could not verify it from a source we trust, so it is not stated.