METHODOLOGY
How we count.
Every statistic on TrophyCoach names its cohort, window, and definition — this page is where those live. Values are re-measured from the production database quarterly; the figures below were last measured on July 2, 2026. Rounding is always DOWN to the shown precision (791 renders as 750+, never 800+).
The homepage numbers
Players coached
500+Distinct player tags with at least one completed coaching analysis. "Coached" means an analysis actually ran — this is deliberately NOT the (larger) registered-accounts count, which includes accounts that never ran one.
Source: COUNT(DISTINCT player_tag) over completed rows in the analyses table (513 measured 2026-07-28).
Battles analyzed
18,000+The sum of battles read across all completed coaching analyses. Each analysis reads up to ~30 recent battles from the official battle log at run time.
Source: SUM(battles_analyzed) over completed rows in the analyses table.
Trophies gained
90,000+The sum of positive trophy movement measured inside analysis windows (the 25-battle trophy delta at analysis time, counting only gains). It is a measured total across the coached population — not a promise about any individual player.
Source: SUM(GREATEST(trophy_delta_25, 0)) over completed rows in the analyses table.
Report and meta windows
Coaching reports read your most recent battles from the official battle log at the moment you run them — typically the last ~30 battles for Clash Royale, the last 25 for the Brawl Stars diagnostic, and the last ~50 attacks and defenses for the Clash of Clans Battle Report. Every rendered metric names its window (“last 28 battles”, “next 10 ladder matches”); a number without a window doesn't ship. Meta tables are computed from continuously collected ranked battles and always show the date window and sample size they were built from.
The numbers inside your reports
Clash Royale. Win rate is wins over the battles in the stated window — no smoothing, no weighting. Elixir leaked and tower hit points are per-battle fields from the official battle log, so “2.4 leaked/match in losses” is the average of the leaked values across your losses in that window. Card and evolution levels come from the same log; matchup and meta context comes from our continuously collected ranked-battle corpus, and every meta table shows its date window and sample size.
Brawl Stars. The map-fit diagnostic computes your win rate per brawler-and-map pair from your own logged battles (~25 at run time). Small samples are shown as small samples — a 2-battle pair is labelled as such, never blended into a confident-looking rate. Ranked progress is graded from rank snapshots between sessions, because the game does not expose per-match trophy change in Ranked.
Clash of Clans. Three-star rate is three-star attacks over total attacks in your log (~50 entries), grouped by army archetype decoded from the battle entries. Town Hall matchup rows use the same arithmetic per opponent TH level. Army Meta tables are computed from ranked battles only — home-village logs blend farming and event armies, which corrupts the popularity and success numbers.
How focus challenges are graded
A focus challenge freezes its target when you accept it: your baseline is the same metric measured over the battles immediately before you accepted, and the grade compares it against your next qualifying ladder battles (usually 10, extending up to 25 when a metric needs more readings). If we can't vouch for the full window — a capture gap, too few readable battles — the challenge is voided with an explanation rather than graded on partial evidence. Grades are pass, partial, or fail: measured, never estimated.
What we never claim
The official APIs expose match outcomes, not match interiors. There is no replay or event stream — so TrophyCoach never claims to know card placement, cycle order, timing, or “misplays”. Brawl Stars Ranked omits per-match trophy change, so ranked progress is graded from rank snapshots between sessions. Clash of Clans war detail requires the clan's war log to be public.
Questions about a specific number? Ask us — if we can't show the query behind a stat, we take the stat down.