Leagues replace the global leaderboard

Leagues iteration

This release builds on the original Leagues launch with the backend work that makes small-group competition actually hold up at scale. The headline addition is leaderboard analytics: every score write now feeds an aggregation layer that lets us measure how players distribute across brackets, where promotion thresholds bunch up, and which league sizes keep sessions competitive rather than lopsided.

On the admin side we shipped dedicated leaderboard tooling and username search. Operators can resolve a player by username, inspect their current league placement and score history, and audit a bracket end to end without running raw SQL against the scores table. That tooling is what lets us tune the system in production instead of guessing from the outside.

Scoring and scoping

The scoring pipeline got tightened so a leaderboard entry is scoped to the right league instance, season, and time window rather than collapsing into one global ladder. A global ladder only rewards the top fraction of a percent of players. Scoping scores to a player's own bracket means the ranks above and below you are real competitors at your level, and a promotion is a measurable event the system records rather than a rounding error against the entire user base.

We also refined how ties, re-entries, and late score submissions resolve so a player cannot leapfrog a bracket through ordering quirks. Each submission is reconciled against the authoritative league standing before placement updates, which keeps the displayed rank consistent with the stored data.

The competition-first framing is the whole reason this exists. Free-to-play economies tend to extract from the bottom of the ladder while only the top sees the leaderboard as fun. Leagues invert that: the analytics and admin tooling here are aimed at keeping the middle of the distribution engaged, because that is where most players actually live.

Promotion thresholds and league sizes remain tunable, and the analytics layer now gives us the data to move them deliberately rather than by feel. Expect continued adjustments as bracket distributions settle.