Leader win rates measured from real tournament matches, plus deck analysis and the ban list, as they stood on one day.
Every figure below was frozen on the date above, so the corpus counts and the status list have both moved on. It is rendered from a blob of JSON pasted into the file — it fetches nothing, which is why it cannot go stale quietly. It goes stale visibly, against its own printed generation date.
The current numbers are on the home page and the card browser, both of which measure themselves at load.
Everything here is computed from ingested data: tournament matches from the Limitless play API, the official card corpus, and the deck files in the maintainer's sim install. Nothing is estimated or hand-written except the status list at the bottom, which says so.
Row leader's win rate against the column leader — the columns are the same leaders as the rows, in the same order.
Sample size n is under every figure, and anything under n=30 is grayed and slanted — too few games to read. Shown here: the leaders with the most decided matches, not the whole field.
The figure in each cell is the Beta-Binomial posterior mean, shrunk toward each leader's own overall win rate. A raw percentage from four games is noise, and shrinkage says so out loud: the fewer games behind a cell, the closer the printed figure sits to that leader's baseline.
Mirrors are 50% by construction, not by measurement — a leader cannot have an edge on itself, so those cells carry the match count and nothing else. On a mouse, hovering a cell also shows the raw percentage and the W–L split behind it.
Ordered by decided matches.
Win rate here is raw across all opponents. The shrinkage above applies to individual matchups, not to these totals.
The deck files this report was run against — whatever happened to be in that sim install, not a recommended list.
Each file was read off disk, checked against the Tournament Rules Manual §2.4 and the ban list current at the time, then analyzed. Counter power is the sum of the printed Counter value (CR 2-10) over every copy in the deck; average power is over Character copies that print one (CR 2-6-2), and any copy missing a printed Power is left out of it rather than counted as zero.
Not every one of these is read the same way, and the weakest one is worth knowing:
Fetched from Bandai and resolved against the card corpus.
A ban lands on the card number, so it covers every printing of that card — base art, alternate art and reprints alike. The deck checker above resolves each entry against the card corpus, so a ban list naming a card the corpus does not have would fail loudly here rather than pass quietly.
Hand-written, and frozen with the rest of the page.
actionV3 → Effect IR translation — no coverage report yetEffect coverage is now measured and printed on the card browser, the IR translation has a coverage report, and the unified CLI exists. The list above is left as it was written rather than quietly edited, because a status list you can silently rewrite is not a status list.