EliF77/brawlstars-ranked

Brawl Stars Ranked Matches — season53

Ranked match telemetry from the public Brawl Stars API, one row per ranked set. A set is up to three games on one fixed map and mode, first team to two wins; record holds the game-by-game outcome.

Collected and published automatically by an open-source ETL pipeline.

At a glance

Ranked sets	1,369,934
Covering	2026-08-20 to 2026-09-14
Distinct brawlers	106
Modes	6
Maps	30
Format	Parquet, one file per season
Files	1
Size	80 MB (6.6x smaller than the source database)

Loading

from datasets import load_dataset

ds = load_dataset("EliF77/brawlstars-ranked", split="train")

Each season is one file, written in time order. Parquet stores the min and max of every row group, so a reader asking for a date range skips the groups that cannot match rather than scanning the season:

import pyarrow.dataset as ds_
table = ds_.dataset("data", partitioning="hive").to_table(
    filter=ds_.field("battle_time") >= "20251110T000000.000Z"
)

Columns

One row per ranked set. Core fields cover time, mode, map, and the record; the six drafted brawlers are flattened into t{team}_b{slot}_* columns.

Column	Meaning
battle_time	UTC timestamp of the last game in the set
mode, map	Fixed for the whole set
record	Game-by-game result, e.g. T1-T1, T2-T1-T1
star_*	The final game's MVP: brawler, power, tag, elo
avg_elo	Mean elo across the six players
skill_ns	avg_elo normalized against the elo distribution local in time
skill_ns_ok	1 when the row's time bin had enough samples to trust skill_ns
t{1,2}_b{0,1,2}_name	The brawler drafted in that slot
t{1,2}_b{0,1,2}_elo	That brawler's ranked rating at match time
t{1,2}_b{0,1,2}_power	That brawler's power level
t{1,2}_b{0,1,2}_tag	Player tag of whoever brought it

Every participant is identified, not only the star player, so a set can be joined to all six players in it. star_player_tag matches one of the six.

Filter on skill_ns_ok = 1 for analyses that depend on the skill feature.

Why skill_ns exists

Ranked elo resets each season and re-stratifies over the following weeks, so an average of 16 in week one describes a very different match from an average of 16 in week three. skill_ns converts each set's avg_elo to a percentile within a fixed 3-day time bin and maps it to a symmetric unbounded scale, making matches comparable across the season.

Modes and maps

Modes: bounty, brawlBall, gemGrab, heist, hotZone, knockout

Maps: Beach Ball, Belle's Rock, Bridge Too Far, Center Stage, Double Swoosh, Dry Season, Dueling Beetles, Flaring Phoenix, Gem Fort, Hard Rock Mine, Hideout, Hot Potato, Kaboom Canyon, Layer Cake, New Horizons, Open Business, Out in the Open, Parallel Plays, Pinball Dreams, Pinhole Punt, Pit Stop, Ring of Fire, Rustic Arcade, Safe Zone, Safe(r) Zone, Shooting Star, Sneaky Fields, Spiraling Out, Triple Dribble, Undermine

Most-picked brawlers

Brawler	Picks
GRIFF	411,486
BROCK	406,927
GUS	304,325
RICO	296,932
EDGAR	268,939
SURGE	258,581
NORI	251,784
BO	205,559
SHADE	201,280
EMZ	187,115

Limitations

• No draft order and no bans. The API returns the six final brawlers with no pick sequence, and the slot index is positional, not draft order.
• Not a uniform sample. Rows are gathered by crawling the player graph breadth-first from seed players, filtered to a target elo band. Popular and higher-rated players are over-represented relative to the whole population.
• One season per dataset. Balance changes shift the meta between seasons, so pooling them mixes distinct regimes.
• Elo is per-brawler, not per-player: a strong player on an unfamiliar brawler carries a low rating into the match.

Provenance

Built from soloRanked battles only. Deduplicated on (battle_time, map, star_player_tag). Sets with an average elo above 23 are dropped as bot or corrupted records. Full pipeline documentation and design notes are in the source repository.

License and attribution

MIT for the pipeline and this dataset compilation. Derived from the public Brawl Stars API. Not affiliated with, endorsed, sponsored, or specifically approved by Supercell; fan content made under Supercell's Fan Content Policy.