ISSUE 001 / OPEN THE NOTEBOOK
Good stories.
Visible workings.
Six findings from official Brawlhalla API responses, with the inputs and limits left in. Observed 29 October 2026.
What we sampled
We requested pages 1–20 of the ALL-region Ranked 1v1 leaderboard, with 50 entries per page and order_by=rating. There were 1,000 unique player IDs. The main profile collection ran from 11:26:40 to 11:28:28 UTC on 29 October 2026; initial probes and repeat checks occurred outside that interval.
The returned rank order was not strictly descending current Elo. Treat this as the first 1,000 returned leaderboard entries, rather than a rigorously defined top 1,000 by Elo or a representative sample of all players. Separate page and profile requests are not an atomic snapshot.
What didn’t line up
- 20 profiles failed with HTTP 500 after retries, clustered at returned positions 813 and 815–833. Missingness cannot be assumed random.
- 241 of the 980 successful profiles had 1–9 fewer summed legend games than player games. We do not know why.
- Some profile rank fields differed from leaderboard positions. Where a position is mentioned, it comes from the leaderboard.
The four featured ranked profiles reconcile exactly. Those profiles and Lopes’s all-mode response were fetched again and were unchanged. Repeatability does not guarantee the API updates instantly.
What the download contains
The JSON includes the selected players’ necessary formula inputs, weapon catalog labels and 1,000 cohort audit rows. Those audit rows record fetch status, returned position, Orion rating and the legends qualifying for the 2,000-Elo/20-game rule. It is an excerpt and audit summary, not a full match archive. The CSV contains the displayed metrics, numerators, denominators, exact values, scope, observation times and source URLs.
Field dictionary & reproduction
- games / wins
- API-reported counters in the named mode. An exact calendar window is not established.
- rating / peak_rating
- Current legend Elo versus reported peak. Roster breadth uses current rating.
- weapon_one / weapon_two
- Frozen catalog metadata maps Sidra’s first slot to Cannon and second to Sword.
- kos / ko_unarmed
- Total and unarmed knockout counters from All modes.
- observedAt
- When BrawlPop received the API response, not a match timestamp.
For percentages, divide the CSV numerator by its denominator and multiply by 100. The breadth row counts legends satisfying both thresholds; its qualifying rows are in the JSON. Source record fields retain the official API names.
Boundaries that stay with the story
Ranked and All modes remain separate. No story establishes yesterday’s games, a consecutive streak, a world legend record, a population legend win rate or a probability of reaching Diamond. A personal comparison uses a later API response with the same formula. It does not become part of this fixed edition.
Unique matches cannot be counted by simply adding player games: two sampled players may have participated in the same match. We do not know opponent strength or match sequence.
Use and cite
Each story offers an illustrated PNG and a copy-ready citation. Keep its date, scope, denominator and qualification when quoting a finding. Attribution links are optional; there are no hidden or required backlinks. AI-generated fan art is separate from the official game data. BrawlPop is not affiliated with Ubisoft or Blue Mammoth Games.
Corrections & editions
This is the 29 October 2026 edition. New observations belong to a new edition; previous numbers should not silently change. There are no recorded corrections to this edition. Operator contact status and general data notes are available separately.


