How the numbers are made

How personal history counts games and explains gaps in the record.

Elixir describes the histories people ask it to record. A win rate describes those games and their opponents; it is not a prediction or a judgment of player skill. The denominators behind every rate are on The battle model.

Card levels: described, not adjusted for

Every battle side's deck is recorded with each card's level, so the record can say how far a player's cards were above or below the opponent's: the mean_level_gap field on battles_decks and battles_cards rows, players_summary decks and clans_standings members (level_gap_battles beside it on deck rows and members counts the battles it averages), with comparable and a note when two rows were played at gaps half a level apart. That is the whole of it. Elixir does not serve a level-expected win rate or a score of a player against one: in matchmade modes a win rate carries almost no information about the player once the matchmaker has paired them. Read the gap as a fact about the battles in a row, never as a judgment of who played them.

Rival intelligence and coverage

war_rivals aggregates recorded river-race readings: races seen, fame, zero-fame races and seasons spanned. This is recorded history, not a forecast. races_observed counts every sighting, the week in progress included; the fame statistics (mean_fame, median_fame, max_fame, zero_fame_races) pool the finished races only, and finished_races is their count, the denominator to read them over. A rival with no finished shared race has null fame statistics, not zero.

Every response carries an envelope with its computation time. Subject tools expose history and source freshness where applicable. Check elixir_coverage for measured reading intervals. Missing coverage is unknown, not evidence of completeness; no recorded battles is not proof of no play. How often a player is fetched, and why a sitting can still, rarely, roll past the ~30-entry battle log, is on Recording and coverage.