How the Meta Lab works

Ladder numbers Showdown ladder — not official ranked

Usage ranks, the move/item/ability/nature/spread/teammate distributions and the 2-Pokémon pair cores come from Smogon's publicly posted Pokémon Showdown statistics for the Champions ladder formats (the monthly "chaos" files, 1760-rating cutoff). Each import writes one snapshot per format and month, keeps the exact source file URL, and records the battle count as the sample size shown next to every table.

• Usage % is Smogon's rating-weighted usage for the month, converted to a percentage.
• Distribution percentages are each entry's share of that species' recorded total for the month.
• Pair-core share is a pair's fraction of all teammate pairings in the snapshot — a co-occurrence measure, not a win rate.
• It is simulator-ladder data: a different player population than the official in-game ranked ladder. Every surface that renders it says so.

Tournament numbers Limitless tournaments

Tournament rows, standings, pairings and team lists come from the Limitless tournament platform's API — recent VGC events, fetched respectfully (sequential requests, capped pages). Team lists are stored only when the organizer published them; the 3- and 4-Pokémon cores are computed from those public lists, and their share is the percentage of sampled team lists containing the combination.

• Community and organizer events — self-selected entrants, not a ranked-ladder population and not official championship data.
• Species names are matched to our Pokédex with the same normalizer used across Champions pages; unmatched names are kept as text and reported by the import's validation, never guessed.
• Limitless publishes no redistribution terms, so raw payloads stay internal — pages render only derived rows with attribution (see the licenses ledger in the repo, docs/DATA-LICENSES.md).

Snapshots, deltas and versioning

• Every import is a snapshot: one row per (format, sample window) with source URL, sample size and a methodology version (chaos-v1 / limitless-v1). New snapshots supersede older ones per format; history is kept.
• The Δ column and the risers/fallers section compare the selected snapshot against the previous one for the same format. Species without a previous sample show "new" — never a fabricated 0%.
• Imports run through the same validate-first, atomic-swap harness as every dataset on this site: a failed validation writes nothing, and the previous snapshot keeps serving.
• Snapshot dates shown on pages are import provenance — never a page-cache timestamp.

What we won't do

• No percentages without a named, linkable source and a sample size next to them.
• No mixing ladder and tournament populations into one unlabeled number.
• No estimates dressed as measurements — empty sections say what import would fill them.

The wider Champions data contract — what's curated, what's sourced, what's an editorial projection — lives at /champions/data. Back to the Meta Lab.