Subpoenaing the matchmaker...
Subpoenaing the matchmaker...
SUBJECT UNDER INVESTIGATION
REFINING WITH WIN-CHANCE MODEL…
Scores below update live as each matchup prediction lands.
Barely perceptible. You're fine. Probably.
High confidence - the sample size holds up in court. Set by sample size: low under 8 battles, medium 8–14, high 15+. Drops a level if too many cards aren't in our database. Thin samples pull the score toward 50.Read the methodology → Analyzed 200 of 200 stored battles for all battles.
Latest API window: 500ranked/ladder battles. Each new analysis merges unseen ranked/ladder battles into this player's history; a report uses up to the newest 200.
Recommended: turn on Tracking (top of this file). A scheduled job then keeps collecting this player's battles beyond the live API window — a larger sample over time means higher confidence and a more precise score, instead of one pulled toward 50 on thin evidence.
This is an observational anomaly score, not proof that matchmaking selected an opponent from your deck or win rate. Factor scores are measured against a fair baseline, then weighted by available evidence. Read the public methodology.
SKILL RATING — PILOT vs DECK Separate from the rigged score. Adds up a predicted win chance for every game (expected wins), compares it to your actual wins, and shows the gap. 50 = exactly as the matchup, level, and trophy gaps predicted; higher = you outplayed the math. The ± is a 95% confidence range.Read the methodology →
85
Expected wins
44
Actual wins
-21±7
Pts vs expected · 95% CI
Throwing Winnable Games
Favorable matchups keep ending in defeat. Something in the play breaks.
The win-chance model expected 85 wins across 200 battles; you took 44 — 21 points below expectation (±7 pts at 95%).
Separate from the rigged score: this measures how your results compare to what the matchup, level, and trophy model predicts — a proxy for piloting, not a read on individual in-match decisions.
50 is neutral. Below 50 means this factor made your recent games harder than a fair draw and pushes the final score toward rigged; above 50 means easier games and pushes it toward lucky.
Win-chance model unavailable — matchup verdict pending for these 200 games. The score reflects bad-matchup timing and card levels until the model returns and fills the matchup read in.
14/20 unfavorable across growing post-win-streak windows (70% vs 56% overall); 367/654 unfavorable across inverse post-loss windows (56% vs 56% overall); 33/61 unfavorable right after a deck change (54% vs 56% overall); longest unfavorable matchup run: 14; streak windows averaged +0 direction-adjusted win-chance points.
Card levels were roughly even.
BAD-MATCHUP TIMING Your average win chance in games played on a win streak, or just after a deck change, vs your overall win chance (the dashed line). Lower while streaking = the draw tightened as you climbed. Reads when the hard games hit, not how many.Read the methodology →
Win chance held steady whether or not you were on a streak.
Each bar averages matchups following completed win streak. Dashed line is overall average. Lower bars after wins, or higher bars after losses, push timing signal toward rigged.
Overall win chance
44%
Post-win gap
+1 pts
Click any row to filter the battle audit to those games.
Each row shows the result, opponent, pure deck-vs-deck win chance, and matchup label. Expand a row to see why the deck was favored, even, or in trouble.
Cards not yet in our counter database (treated as neutral): Ronin. Data patch: June 2026 balance (win-condition role + counter pass).
VERDICT OF THE JURY