la serie/

How it is measured

Nothing is matched by name resemblance and nothing is filled in by eye. A project missing a piece comes out with that piece null and a warning, not with an invented figure.

fact	where it comes from	tool
who each project is	the series' front-page card	`recoge_portada.py`
which directory it lives in	git's `remote`, matched against the card's URL	`recoge_portada.py`
which binary it disassembles	the variable in the repository's Makefile	`recoge_binarios.py`
size and sha256	the file, read again	`recoge_binarios.py`
instructions and comments	the `.asm` files the repository publishes	`recoge_cifras.py`
what the repository declares	its own `make densidad`	`recoge_cifras.py`
the hidden mark	the binary, swept end to end	`recoge_marca.py`
who signs the game	the readable runs in the binary	`recoge_creditos.py`

The same yardstick for all forty-eight

Every repository carries its own copy of densidad.py, written over months, and comparing between games requires knowing they count alike. There are four variants and all four use the same criterion: they differ in line endings, in a special case for Nemesis (a bank with no code), and in Trailblazer's adding up several tape pieces.

So here everything is measured with one canonical copy and, on top of that, each repository's own make densidad is run to check: thirty-four agree exactly and none disagree. That is what makes the figures valid for the fourteen that do not carry that target, and a test watches it.

What this database cannot claim

That a game "has no credits". Only that none appear in ASCII.
That two cartridges share no routine: if it was assembled at different addresses, a byte comparison cannot see it.
Where the "labels" figure published by six front-page cards comes from. It is neither the routine count, nor the README's named-labels count, nor the listing's label count. It is recorded as an open question.