How the scores work

Where decklists come from, and what is missing

Melee events publish full standings, so every deck in the field is known. MTGO challenges publish only the top 32 decklists (the exact number of entrants is on the page and stored, but the other lists are not). MTGO leagues publish only 5-0 lists. That matters for two different things: shares (how much of the field a deck is) and win rates.

Complete event

Published lists cover the whole field (at least 95% of entrants). Melee events with full standings, small MTGO challenges, and magic.gg dumps count here.

Cut event

An MTGO challenge with more entrants than published lists: what you see is the top 32.

League

5-0 lists only. Never complete, and never used for win rates.

The "Complete events only" switch on the Metagame and Staples pages restricts everything to complete events, so a share means "this much of what people actually played". With it off, shares include top-32 cuts and lean toward decks that win. Mixing the two without saying so is the thing this page exists to avoid.

Swiss records

MTGO challenge records include the top-8 playoff. Before any win-rate math, playoff matches are removed by finishing position (the winner played 3 extra matches and won them all, the finalist 3 with one loss, semifinalists 2, quarterfinalists 1), so every deck is compared on its Swiss record.

Adjusted win rate

A top-32 record is high because the deck had to win to be published. The adjusted win rate models that directly: every deck from a cut event contributes its record conditional on having beaten the cut record (a truncated binomial likelihood), decks sitting exactly on the cut record are left out because that bin is only partly published, and decks from complete events enter with no truncation. One win probability per archetype is fitted to all of them over the last 90 days, with a 95% profile interval. As a sanity check, raw record win rates in Modern average around 76% while adjusted ones average around 55%.

It is still a win rate against the field the deck actually faced, so it says how a deck did, not how it would do against a specific opponent. Use the matchup pages for that.

Cut conversion

Reaching the top 32 often is a strong signal, but a deck that is 20% of the field will reach it often on volume alone. Cut conversion divides the deck's appearances in top-32 cuts by the number expected from its share of complete fields in the same format and window. 1.0× means it makes the cut exactly as often as its popularity predicts, 1.5× means half again as often. It is shrunk toward 1.0× for small samples, needs at least five decks in complete fields to be computed at all, and is capped between 0.33× and 3×. Formats without enough complete-field data (currently Pioneer and Vintage) show no conversion.

Strength

Strength is one number in win-rate points above the field. It takes the adjusted win rate, shrinks it toward 50% by sample size (a 60% deck on 10 lists counts for less than a 60% deck on 300), and blends it 60/40 with the cut conversion, where a 2× conversion is worth about 9 points. A strength of +5 reads as "about five points better than an average deck, on the evidence so far". Sort by the 95% floor when you want to be conservative with small samples.

What it is not: it is not a metagame share, it does not know about matchups, and it cannot see decks that never get published. A deck with no cut appearances and a handful of full-field results will sit near zero until there is evidence either way.

Current coverage (last 90 days)

Coverage and adjusted rates are rebuilt nightly after archetype classification; the underlying scraped data is never modified by them.