The Commander cards
we've watched people play.

Players using Playgroup Live import their decklists before a game and log card actions as they play. Each cast, zone change, and resolution is recorded as a real game event. Stats here are computed from those events, not from decklist scrapes or theorycrafting.

Card stats are recomputed every night at 03:00 UTC from the previous day's game events. Deep-dive narrative pages refresh when a card's underlying numbers drift by more than 25 percent, or when a card first crosses the threshold for a dedicated page.

There are two bars. A card shows its basic play counts once it has appeared in at least 100 tracked Commander games. It earns a full deep-dive page only after a much larger and more diverse sample: at least 300 tracked games and 100 recorded appearances, played by at least 10 different pilots, with no single pilot accounting for more than 40 percent of the copies. Below those bars the sample is too small or too lopsided to make honest claims, so the card still resolves at this URL with whatever counts we already have.

For every card we track how many distinct pilots brought it and the share held by its single most prolific pilot. A card is kept off the ranked strips and deep-dive pages if one player accounts for more than 40 percent of its copies, and a deep-dive page additionally requires at least 10 different pilots. That is the line between real-table data and a couple of friends jamming the same pet card.

EDHREC ranks cards by how often they appear in public decklists. Playgroup.gg ranks by what actually happens when those decks get played. That means we can tell you things a decklist cannot: when cards get cast, how often they stick on the battlefield, and how win rate shifts when they resolve versus when they never leave the library.

"Win rate when cast" is the share of games a player won when this card resolved at least once. "Cast-vs-library delta" is the gap between that and the win rate of games where the same card sat in the library all game, with both numbers normalized to a 25 percent pod baseline so different player counts compare fairly. Read it as a correlation, not proof of impact. Expensive payoffs tend to get cast in games where you have already developed a board and survived deep into the game, so part of any positive delta just reflects the game states in which casting was even possible. We only rank a card on this stat once both groups have at least 30 games and the lift stays positive at the low end of its 95 percent confidence interval.

Yes, the aggregate per-card stats are published under CC-BY 4.0. Link back to the relevant card page and you're good to go. The raw game events themselves are not public; the aggregates are what we publish.