V2 publishes balance decisions as a player-facing artifact, not only as internal
telemetry. The current major patch is patch-2026.05.s1.
Public Artifacts#
- Official tier list:
V2/balance/public-results/tier-list-2026-05-s1.json - Rank-specific tier views:
V2/balance/public-results/rank-tier-list-2026-05-s1.csv - Auto-generated matchup chart:
V2/balance/public-results/matchup-chart-2026-05-s1.csv - Stage matchup variance:
V2/balance/public-results/stage-matchup-variance-2026-05-s1.csv - Frame-data diff:
V2/balance/public-results/frame-data-diff-2026-05-s1.csv - Public frame-data tool:
apps/v2/web/balance/ - Raw frame-data CSV:
apps/v2/web/balance/frame-data.csv - Companion app surface:
apps/oshun/mobile/v2/balance - Public API surface:
apps/oshun/web/api/v2/frame-data
Cadence#
Monthly major balance patches ship monthly. Minor tuning windows are bi-weekly.
Every change lists rationale, telemetry justification, PTB signal, and public
before/after data under V2/balance/changelog/ or V2/balance/public-results/.
Community comment windows last 14 days after a patch. The next review cannot close until the comment window, opt-in in-game survey, and tournament-specific polls are summarized.
Tier Lists And Matchups#
The official tier list publishes per-fighter tier, rationale, and suggested counters. Community polling is supplementary and never overwrites official telemetry. Bronze, Silver, Gold, Platinum, Diamond, Master, and Grandmaster have separate tier views because execution burden and matchup knowledge shift by rank.
Matchup charts are generated from match-result data and published with per-fighter win rate, rank-tier variation, and optional stage variance notes. Stage variance is advisory; tournament stage bans remain under esports rules.
Frame Data#
Every authored move exposes startup, active, recovery, on-hit, on-block, damage, ruleset, and build branch in the public frame-data tool, companion app, raw CSV, and public API. Patch diffs show exactly which fields changed between versions.
Pro Feedback And Freeze#
sponsored pros and accredited tournament organizers receive pre-patch consultation. A published summary appears after the patch. Major tournaments use a two-week balance freeze; emergency hotfixes are limited to crash, exploit, or game-breaking infinite fixes. Pro-event tier list publication requires pro and tournament organizer sign-off.
Recommender And A/B Testing#
The ML balance recommender trains on patch deltas, match outcomes, community feedback, PTB surveys, and public replay tags. It suggests buff or nerf candidates with confidence scores, but designers review every candidate and no recommendation is auto-applied.
Balance changes can run through PTB A/B tests before live rollout. The framework requires a public result URI, rollback runbook, PTB signal coverage, and minimum sample size before a recommendation can graduate to accepted.