Player Statistics

Each finalized match emits a normalized event into the analytics pipeline, where a worker fans it out into per-player AnalyticsRollup documents bucketed by day, week, and month. The upserts are idempotent and keyed on player_id + bucket + dimension, so re-running a closure (for example after a result correction) deterministically rewrites the same row instead of double-counting. Hourly cron jobs roll up the current day, a daily_close job seals yesterday at midnight tenant-time, and a monthly_close job locks the prior month for trend charts.

For today's bucket a thin in-memory live-delta layer adds events that arrived since the last hourly tick, so the public profile and the player's own dashboard always look up-to-date without waiting for the next cron tick. Match statistics (F10.01.01) and win rate by format (F10.01.02) are read straight from these rollups; performance by competition level (F10.01.03), season-by-season comparison (F10.01.05), and casual vs official breakdown (F10.01.14) just slice the same rollups by an extra dimension. Head-to-head records (F10.01.04) and partnership statistics (F10.01.15) are stored in pair-keyed rollups (player_id, opponent_id) and (player_id, partner_id) so the comparison view can be answered by a single indexed read.

Position effectiveness (F10.01.06) reads per-position rollups (pointeur/tireur/milieu) populated from the position field on each lineup. ELO and Federation Points history charts (F10.01.11, F10.01.12) and the dual ranking comparison (F10.01.13) are rendered from append-only rating snapshots, with the dual chart aligning two series on the same time axis. Career statistics, milestones, and best results (F10.01.07, F10.01.08) come from a separate aggregate that watches for thresholds (100 wins, first podium, longest streak).

Activity metrics (F10.01.09) and the public player stats page (F10.01.10) are gated by privacy settings, so a player can publish, restrict to federation, or hide their profile entirely.