Mr. P's observed win rate rose: Brawl Stars sample report, 2026-09-29

What we compared

We compared the recent window, 2026-09-21 to 2026-09-28, with the preceding window, 2026-09-14 to 2026-09-21 (UTC). Both are inside the recorded 2026-08 balance window. The data was computed at 2026-09-28T20:16:15.324Z. These dates describe this saved report; the live meta board will move on as more logs arrive.

Mr. P: read the movement alongside its uncertainty

The earlier raw rate was 50.6%, with a Wilson interval of 47.9%–53.4%; the recent rate is 60.7%, with an interval of 57.9%–63.3%. The recent sample covers 470 distinct players, compared with 694 previously. Repeated-player concentration reduces the effective sample to 470 recently and 694 previously. The intervals do not overlap, but our selection compares many brawlers and is exploratory, so this is not a formal significance claim.

Mr. POpen Mr. P's live mode and matchup data before applying this overall movement to a specific game.→

Darryl: read the movement alongside its uncertainty

The earlier raw rate was 40.0%, with a Wilson interval of 38.2%–41.8%; the recent rate is 47.0%, with an interval of 45.4%–48.7%. The recent sample covers 1,667 distinct players, compared with 1,689 previously. Repeated-player concentration reduces the effective sample to 1,667 recently and 1,689 previously. The intervals do not overlap, but our selection compares many brawlers and is exploratory, so this is not a formal significance claim.

DarrylOpen Darryl's live mode and matchup data before applying this overall movement to a specific game.→

Charlie: read the movement alongside its uncertainty

The earlier raw rate was 52.1%, with a Wilson interval of 49.4%–54.9%; the recent rate is 57.9%, with an interval of 55.2%–60.6%. The recent sample covers 507 distinct players, compared with 646 previously. Repeated-player concentration reduces the effective sample to 507 recently and 646 previously. The intervals do not overlap, but our selection compares many brawlers and is exploratory, so this is not a formal significance claim.

CharlieOpen Charlie's live mode and matchup data before applying this overall movement to a specific game.→

How to use this in your next decision

Use Mr. P's improvement as a reason to inspect the current evidence, rather than as a reason to spend resources immediately. First match the mode and map you intend to play. Then check whether that smaller sample supports a recommendation. A broad improvement can coexist with weak results in one mode, so an overall number should not overrule a poor fit for the objective.

Next, compare the recommended loadout with your unlocked equipment and the roles your teammates already fill. The battle data behind this report does not establish which gadget or star power caused an outcome. Our build guidance is a separate editorial layer. Treat experience with the character, objective coverage, and available equipment as decisions the aggregate cannot make for you.

Limits of this snapshot

A brawler appearance is not a unique battle: one team battle contributes results for several participants. Rates therefore must not be added together and described as independent matches. These logs come from a collected set of players, not a random sample of every Brawl Stars player. Changes in who is collected, trophy levels, team composition, and map rotation can move the rates. Showdown placements are excluded from this win-rate comparison.

The table uses raw wins divided by decided results for a transparent comparison. Rankings elsewhere on Brawlest may use a shrunken rate and a different selected window, so they need not have the same order. The sample gate adjusts for repeated players; the displayed Wilson intervals do not fully model every dependency or collection bias. No causal explanation is inferred from a change in these numbers.

This report was assembled automatically from saved aggregate evidence. It was not a human playtest, and its publication date is not an assertion that a new patch launched that day. The evidence download below preserves the inputs so a correction can be checked against the same snapshot.