How the numbers work
What is collected
Brawlest reads the official Brawl Stars API on a schedule. Each run pulls the brawler list, the event rotation, and the recent battle log of a set of players — seeded from global rankings and then expanded to players seen in those battles. Battles are stored once, keyed by a fingerprint derived from the match itself, so the same battle appearing in several players’ logs counts once.
Brawler statistics, skins and map renders are not in the API. Those come from the Brawl Stars wiki and are re-hosted here with the source page and revision recorded per record. The brawler catalog was last read on 20 Jul 2026 and the skin catalog on 20 Jul 2026.
How win rate is calculated
The API reports one result per battle: the outcome for the player whose log was read. That result applies to their whole team, and the opposing team gets the inverse. Each battle therefore contributes one outcome per participant.
The raw win rate is wins over decided matches — draws are counted and reported, but they are not in the denominator. Every rate is then shrunk toward the global mean in proportion to how thin its sample is, so a brawler with forty matches does not outrank one with four thousand on noise alone. The published figure is that adjusted rate; a 95% Wilson interval is computed alongside it and drives what gets published at all.
How pick rate is calculated
Pick rate is a brawler’s share of all brawler slots in the same window. A 3v3 battle contributes six slots. Under this definition every brawler’s pick rate in a window sums to exactly 100%, which is not true of the other plausible denominators.
Windows
The same pipeline runs over several slices of the same battles: everything, one mode, one map, one trophy range, and the last seven days against the seven before them.
Map windows use the map identity resolved when the battle was first stored, not a lookup performed later. Event ids are reused as rotation changes, so a later lookup would attribute old battles to whatever map now answers to that id.
Trophy ranges
A trophy range is the count a player actually held on that brawler when the match was played — the one field in a battle log that says anything about who was playing. Picking a range changes which matches are counted. It never changes how much evidence a number needs before it is published, so a range with few matches shows fewer brawlers rather than weaker ones.
- Under 500 — Below the first Star Power unlock threshold for most accounts, so loadouts are incomplete and results say more about progression than about the brawler.
- 500–999 — Full loadouts are common and matchmaking is dense here; this is the ordinary player the site is mostly describing.
- 1000+ — Sustained high-trophy play on a single brawler. Small population, so this band clears publication thresholds slowly and honestly reports when it has not.
Trophy readings were only added to the pipeline recently, and a match is stored once, when it is first seen. Matches recorded before that carry no reading and belong to no range, so the ranges cover a smaller slice of the data than the overall figures do and will keep filling with each run.
Trends
An arrow compares the last seven days against the seven before them, with a one-point dead band so it does not flip on ordinary noise. It is drawn only when both weeks are genuinely comparable: equal in length, and each carrying enough matches to be published on its own. A week that would not be shown as a number cannot be subtracted from one that would.
Both conditions currently fail. The earlier week is cut short at the July 8 maintenance boundary, and it holds a small fraction of the matches the recent week does, because the game only returns a player’s most recent battles and this site has not been collecting them for long. So no trends are published yet, and the pages that would carry one say so.
When a number is not shown
- Under 100 decided matches, a row is not published at all.
- A ranked recommendation needs 500 decided matches and a 95% interval no wider than ±3 points.
- Pairings and team compositions need 20 decided matches before they are stored, and the same publication gate before they are shown.
- Where a map has too few matches, the mode’s numbers are shown instead and labelled as the mode’s.
How the player leaderboard is ordered
By trophies, exactly as the official API reports them for an account. Brawl Stars publishes no other score for a player, so that is the only thing there is to sort by — there is no rating, grade or percentile anywhere on this site, because inventing one and putting it beside somebody’s name would be a claim nobody could check.
The one choice Brawlest makes is which accounts appear. It starts from the global ranking and widens to the people those players meet in battle, so the board is a slice of the top of the game rather than a census of it.
A trophy change is the difference between two readings of the same profile, taken at least twelve hours apart within the last week, and every one is shown with the period it covers. A player whose profile has only been read once shows no change at all: a zero would say they went nowhere, which is a different statement from not having been measured twice.
A player’s page shows what the API returns for that tag and nothing else. Those pages are not indexed by search engines.
Update frequency
Ingestion runs roughly twice a day and aggregates are recomputed after each run. Pages show when the numbers they display were last recomputed.
Limits worth knowing
- Players are reached by starting from country rankings and following the opponents and team-mates found in their battles. Nothing records which route a player arrived by, so there is no “top players” group to report on and this site does not claim one. A trophy range is the honest version of that question.
- Showdown is excluded from win rates. The API reports a placement for it, not a win, and a placement is not a win rate.
- The battle log does not record which Gadget, Star Power or Gears a player had equipped. Builds on this site are editorial recommendations with a review date, not measured per-build win rates, and they are labelled that way everywhere they appear.
- Skin prices are the availability line as it read when the catalog was last checked. They are not a live price feed.