Brawl Stars Ranked

Active now: the wiki’s season-50 table, joined by name to the anonymous cache’s season-50 pool rows. “Pool battles” is how many battles the cache saw on the map — never a pick or win rate. Every pick cell is being measured.

Beach Ball has no measured season-50 pool row (unseen, or below the 50-battle floor), so it is labelled unmeasured. 5 further maps (Pit Stop, Rustic Arcade, Goldarm Gulch, Flowing Springs, Safe(r) Zone) were seen in the cache but not in the wiki list; they are kept separate and never counted as pool members.

The cache records the map of every battle, but our map table mixes queues. A Ranked-only per-map pass has not been run, so no map shows a pick or win rate here. It would produce this column; until then every row prints “Being measured — per-map Ranked picks”.

An “edge” row is too thin, or covers only the season’s first six hours, and never becomes a rate. Seasons 48 and 49 are history: they cannot be read as the current pool.

* Gold and Masters carry the two source conflicts: the official table prints Gold 1,499 and Masters 8,500, while the wiki prints Gold I 1,500 and Masters I 8,250. Both values are kept; see “Sources disagree”.

Supercell requires 9+ Power for a brawler to be eligible; one Ranked season is 1 month; 4 Ranked seasons fit one Pro Pass.

Masters I sits at 8,250 here because this is the wiki’s minor-rank table; Supercell’s group table starts Masters at 8,500. The calculator uses these floors and says so whenever the two disagree for your Elo.

Free pick — Bronze I to Gold III

22s to pick; a timeout means “match cancelled”.

Bans — from Diamond I

22.5s per ban; a timeout picks a random ban; a team may not ban the same brawler twice.

Best of 3 — from Mythic I

First to 2 rounds; a third-round draw is “winner decided automatically”. A coin flip decides first pick. Picks 22s, loadout 17s; a swap needs Power 11.

Pick order when blue picks first: Blue 3 → Red 3 → Red 2 → Blue 2 → Blue 1 → Red 1.

Mythic I and above: highest Elo or room host, last pick

Roster gates inside the draft

• Diamond I: 9 brawlers at Power 9
• Mythic I: 12 brawlers at Power 11

Reset: 6 minor ranks down, floor Silver I. The ranked boost pays 10% extra; trophies are not used; Ranked unlocks at 1000 trophies.

Masters is solo-queue only; team sizes are 1, 2, 3.

• Silver — never demotes to Bronze (the reset only goes down to Silver I)
• Gold, Diamond — cannot drop below the full rank during the season; within the rank a drop is possible
• Mythic, Legendary, Masters — drops below the full rank only by losing a match at exactly the threshold Elo
• Pro — no floor protection (can fall to Masters III)

Season end rule: third Wednesday ends; third Thursday starts.

Ranked (solo) battles of the global top 49601 trophy players · 2026-09-16–2026-10-16 · n = 197495 battles · Preview sample. 197,495 matches · 1,175,694 slots. Pick share sums to 100% over all 108 brawlers; win rate is side-balanced, mirrors excluded, and carries its Wilson 95% margin.

“Players” is how many different top players’ logs stand behind the row. A rate below the 100-decided-game floor has no number, so a blank cell is a floor, not a zero.

Ordered by the win-rate lower bound (win rate − Wilson margin) on the Ranked ladder. It answers “what survives its own margin”, not “what covers the pool”.

A per-map Ranked pass would replace the last column with the number of season-50 maps a brawler actually covers. Until it runs, no coverage figure is printed.

Share is the brawler’s slots ÷ all slots in that window and ladder, so a roster sums to 100%. Win rate is side-balanced ½·(w0/n0 + w1/n1), mirrors excluded, and is printed only when both sides clear the 100-decided-game floor.

• Penny56.8% → 48.8% · -8.0 pts (margin 4.8%)n = 9,645 vs 415
• Belle42.1% → 48.3% · +6.2 pts (margin 6.1%)n = 6,719 vs 259
• Stu55.5% → 49.8% · -5.7 pts (margin 4.4%)n = 12,705 vs 506
• Shade52.5% → 57.5% · +5.0 pts (margin 3.1%)n = 25,964 vs 1,027
• Nori47.8% → 52.3% · +4.5 pts (margin 4.1%)n = 15,074 vs 609
• Brock54.2% → 51.1% · -3.1 pts (margin 2.4%)n = 38,921 vs 1,751

Only changes bigger than the combined Wilson margin are listed; everything else is “no clear change”, not zero.

Being measured — per-rank pick and win rates

The v1/v2 caches do not store brawler.trophies (the Ranked index) for soloRanked rows — recordOf keeps ids, power and star player only. A re-collection with an `rk` column is required before bands can be counted.

Profiles of 2000 sampled global top trophy players · 2026-10-19 · Preview sample

The highest Elo in the sample is 8,282 — above the wiki’s Masters I floor of 8,250 and below Supercell’s Masters start of 8,500 — exactly the gap between the two sources, not a third threshold.

Whose games these are

Ranked (solo) battles of the global top 49601 trophy players · 2026-09-16–2026-10-16 · n = 197495 battles · Preview sample. The sample is the battle logs of 49,601 top-ranked players (the global top 200 and the top 200 of each country list). It is high-level play and a preview — never “all players”.

How a win rate is counted

A match is in the sample only because a top player is in it, so every rate is counted on the log owner’s side and on the opposing side separately and the two are averaged: ½·(w0/n0 + w1/n1). A rate is printed only when both sides clear the floor; the margin is the Wilson 95% half-width. Mirrors are left out of win rates but count in pick share.

What the ladder is

Two sources are kept side by side. The official table (Supercell Support, “Intro to Ranked”) gives each group’s score range, format, picking style, roster requirement and reset. The wiki’s table gives each minor rank’s Elo, victory/draw/defeat values and minimum gain and loss. Where they print different first scores, both are on the page.

What the map pool is

The 26 season-50 maps are the wiki’s active-maps table. Each is joined by name to the anonymous cache’s season-50 pool rows, which add a battle count and the first time the map was seen. A wiki map with no row, or fewer than 50 battles, is labelled unmeasured rather than given a number.

What is not measured

Per-rank pick and win rates: the v1/v2 caches do not store the rank index of a soloRanked row, so the rank-band table has zero rows. Per-map Ranked picks: our map table mixes queues and no Ranked-only per-map pass has been run. Both print “Being measured” with no figure.

Which numbers are inputs

The readiness meter, the Elo calculator and the session journal use only what the reader types on this device. They are labelled input, stored in localStorage, and never sent anywhere or presented as observed matchmaking data.

A store update is not a balance patch: only the entries whose wiki section lists balance changes move the sample’s patch boundary.