Unbrewed Balance Dashboard

Deck Balance Tracker

… games logged

Pilots

About these numbers

What is the win-rate bar & the 95% CI band?

The pill is the deck's win-rate point estimate (green ≥55%, red ≤45%, cream in the neutral 45–55% zone); the faint vertical line is 50% (even). The translucent band around it is the 95% confidence interval — the range the deck's true win rate is likely in, given its number of games. Fewer games → wider band → less certain. The axis runs 25–75%; values past the ends sit at the edge. Hover any bar for the exact figures.

How is the confidence interval calculated?

It's a Wilson score interval at 95% (z = 1.96), chosen because it stays within 0–100% and stays accurate for small samples and lopsided win rates. With p = wins / games and n = games: center = (p + z²/2n) / (1 + z²/n), margin = z·√(p(1−p)/n + z²/4n²) / (1 + z²/n). Example: 28 wins in 39 games ≈ 71.8% → 95% CI ≈ 56%–83%.

How are balance flags decided?

A deck is flagged OVER only when its entire 95% CI is above 55%, and UNDER only when the entire CI is below 45% — and only with at least 30 games. Using the whole interval (not just the win rate) prevents small, noisy samples from looking falsely extreme.

What is pick rate?

How often a deck appears, as a share of the games in the current view. The format and pilot filters change it.

What is the deck profile / play mix?

How a deck's cards split across Attack / Defense / Versatile / Scheme. When the deck registry has the real card list, these are exact card counts (with Σ values on the deck page); otherwise it's a play mix derived from card telemetry, labeled as such.

What does the 2v2 synergy Δ mean?

Δ = the pair's actual 2v2 win rate minus its expected win rate (the average of each deck's solo 2v2 win rate). A big positive Δ means the duo is more than the sum of its parts. Click a pair to see the opponents it beats and loses to.

How do I read the matchup matrix & pilot filters?

Matrix: each cell is the row deck's 1v1 win rate vs the column deck (green favors the row), needing at least 3 games — “·” means not enough data. Pilots: the chips restrict to games where every seat is one of the selected pilot types (human or a specific bot); with all on, there's no filter.

Aggregate statistics only — no per-player or raw game data is shown.