About Title Screen
Title Screen is a reference of facts about video games: who published a game, when and where it came out, how many can play, what kind of game it is, which platform it appeared on first, and the identifiers (serials, set names, checksums) that let software recognise a copy. It covers the home consoles and handhelds of generations 2 to 7, the arcade, and part of generation 8.
It is built in the spirit of CDDB: a listing of key facts, searchable by people, and shipped as a machine-readable archive any program can use.
What is in it, and what is not
Facts and identifiers only. No descriptions, synopses, reviews, scores, biographies, artwork or screenshots. Every value on a page is a title, a name, a date, a number, a code or a checksum.
Three levels of subject:
- a game (a work) across every platform and market;
- a release: the game on one platform in one market, with its own date, publisher and serial;
- an identified copy (a dump): an entry of No-Intro, Redump, MAME, FBNeo or the MAME software lists, with its files and their checksums.
Where the facts come from
Every fact is resolved from statements made by sources, and every statement is kept, with where it came from and how it was joined to the game. The release page shows them under Evidence. Sources are ranked in tiers: 1 the media itself (headers the publisher wrote into the cartridge or disc), 2 the curated dump lists and MAME, 3 the press of the day, 4 the web databases. The highest tier wins; within a tier, agreement wins; a disagreement that matters is flagged rather than guessed.
The sources today: No-Intro and Redump (through their DATs and libretro-database), MAME and its software lists, FinalBurn Neo, Wikidata, Wikipedia's lists of games, GameTDB, the LaunchBox games database and catver.ini. Only facts are taken from them. How a statement is joined to a copy, how a fact is decided, what cannot be checked and how much the sources agree with each other, measured each build: Methodology. Each release's evidence has a page of its own, linked from the marks beside its facts.
Identifying a copy
A game can be identified from its bytes, however the file is named and whatever it is packed in: see Identify and the two specifications, the canonical form and the disc fingerprint. They are written so that anyone can implement them without our code.
When a publisher sold identical bytes in two markets, no checksum can tell the two apart, and Title Screen says so: a lookup returns every release the copy was sold as.
Ids
Games, releases and companies have short stable ids (w, r and c followed by seven letters). An id never changes meaning; when two games turn out to be one, the retired id redirects to the one that absorbed it.
For programs
Everything here is available as JSON: see the API. The whole database is downloadable, one archive per build (SQLite, JSON Lines and CSV, with the schema, the sources and the evidence rows): Downloads.
Visits
Visits are counted on the server, one row per request, to see what is used and where readers come from: the page, the referring site, the country and region, the kind of device and browser. No script runs in your browser, no cookie is set, and your address is not kept: a hash of it with your browser's name, salted and changed every day, counts visitors for that day and cannot be linked across days. Nothing is shared with anyone.
Licence
The data is published under CC BY 4.0: use it for anything, with credit ("Data: Title Screen (https://titlescreen.org), CC BY 4.0"). The code is under the MIT licence. The archive is a compilation of facts; the sources it is compiled from are acknowledged in each archive's SOURCES.md, and their terms do not pass to it.
Coverage
How complete the facts are, platform by platform: Coverage. The database is built from public lists and grows as more of the primary sources (the media's own headers, the title screens, the magazines of the day) are read.