94 releases 94 tags

RSS feed

• v1.20.0 54d021a934

  Compare

  v1.20.0 Stable

  joeyr released this 2026-09-13 16:17:16 -05:00 | 53 commits to main since this release

  Changed

  • Remade games no longer count toward prediction accuracy. A game that
    ends in the first five minutes is a remake: someone failed to connect and
    the team voted it dead. Riot still records it as a loss for all ten players,
    so every remake previously landed in the accuracy record as a real outcome,
    scoring a prediction against a game nobody played.

    The end-of-game embed now says 🔄 Remake with the duration and nothing
    else: no KDA, no CS/min, no vision, no prediction verdict, no running
    record. The prediction row is kept and flagged remake, which leaves it
    visible in the history while actual_winner stays NULL, and NULL is what
    keeps it out of /aufhocker lol predictions and the per-game record line.

    Detection prefers Riot's own gameEndedInEarlySurrender flag and falls back
    to the five-minute duration threshold when it is absent. A zero duration is
    Riot saying "no data", not a zero-second game, so it is never treated as a
    remake.

    The same filter applies to the recent champion W/L shown on the pre-game
    embed and fed to the prediction prompt, which was counting remakes as
    losses and dragging the record it reported on both.

    Migration 009 adds the flag and backfills it: predictions already scored
    against a sub-five-minute game (identifiable because lol_active_games
    keeps the match result as JSONB and is never pruned) are un-scored, so the
    displayed accuracy corrects itself on deploy.

  Downloads

  • Source code (ZIP)
    0 downloads
  • Source code (TAR.GZ)
    0 downloads