04 — Grow
Strata isn't a one-time dataset that goes stale. New games are pulled, classified, and added automatically every week — no manual research, no manual scoring.
The problem
A spreadsheet of manually-scored games is only as current as the last time someone sat down and updated it. That doesn't scale, and it doesn't survive a founder getting busy, sick, or distracted for a month. Any data product that depends on continuous manual labor has a built-in expiration date.
The mechanism
The pipeline pulls candidate titles, checks the database for existing entries, and classifies anything new — entirely without a person in the loop.
01
Pull candidates
Each run queries a game database API for titles meeting a minimum critical score, sorted by rating, walking forward through results pages.
02
Skip duplicates
Every candidate title is checked against the existing database — including fuzzy matching that ignores case and punctuation — before anything is reclassified.
03
Classify new titles
Anything genuinely new gets scored across all 21 traits by the same fixed methodology used for every other game in the database.
04
Store permanently
Results are written to a persistent database that never resets — this week's additions sit alongside every game classified since the project began.
Proof
The database has grown from zero to roughly 295 games through this automated process alone, with no manually-entered classification at any point.
Because the same fixed methodology classifies every game — the first one and the three-hundredth one — the dataset's internal consistency doesn't degrade as it scales. A game classified today is scored exactly the same way a game classified a year from now will be.
Continue exploring