Many people have surmised that showing statistics aggregated across all 17Lands users doesn't quite tell the whole story in some cases. For example, if inexperienced players play a card a lot that good players only utilize in a small number of "correct" situations, the inexperienced players' habits will drag down a card's performance unnecessarily. To help combat this, for some high-volume formats, we've broken out a few performance groups. The exact criteria for these may change over time, but this general premise will still hold:
A few things to note here:
Additionally, please note that this is not meant to praise or shame any individual user; we have no intention of revealing which group a user falls into, neither directly nor to the public.
As our understanding of the Arena logs has evolved, and as the ways people have tried to use our statistics has grown, we have gone through iterations of how our various card-level statistics are calculated.
In an effort to provide more transparency, we attempt to document clear definitions here. We also document their changes over time, as we expect them to continue to evolve as we add or replace metrics, improve our compute capabilities, and respond to feedback from the community.
These statistics apply to the draft portion of an event. While in theory we could deduce some information about packs from multiple 17Lands users in the same draft pod, we treat each person’s draft as entirely independent of any other information we have about the same draft pod.
Description: The number of packs in which a card was seen. Cards that come back around when we see them again "on the wheel" are only counted once. When we have information about what was picked but not the other cards that are available (as has been the case for P1P1 for most of the duration of human drafts and also happens occasionally due to connection errors), we do count that one picked card as being "seen".
Caveats: At this time, Arena does not have any duplicates of cards in a pack (unlike paper or MTGO packs that can have duplicates due to foils), but if that were to change, two of a card in a single pack would count twice; we do not do any deduplication.
Description: The average pick number where this card was last seen in packs. When a card comes back around on the wheel, only the second time around counts toward the average.
Caveats: When P1P1 contents are missing, ALSA is slightly elevated for cards that often do not wheel because cards are missing the data from pick 1. Cards that are picked later are less affected because their missing pick 1 data is no longer relevant when pick 9 comes around.
Description: The number of instances of this card picked by 17Lands drafters.
Description: The average pick number at which this card was taken by 17Lands drafters.
These statistics apply to the gameplay portion of an event. They provide data at the granularity of a game (not a match, even for Bo3) because players can and do change their decks between games in the same match or event. If for some reason we find that any of the data for a game is not consistent (e.g. had a card in hand that wasn’t in the deck / sideboard, no deck was linked to the game, etc.), that game doesn’t contribute to any of the in-game statistics.
Description: The number of games played with this card in the maindeck, multiplied by the number of copies. Playing ten copies of a card in a deck has ten times the weight as playing one copy.
Description: The rate at which this card was included in the deck (when available in the card pool). This is weighted by number of games and copies of the card. For example, a deck that played eight games with three maindeck copies of a card and another deck playing four games with one copy of the card in the maindeck and one in the sideboard would result in an 87.5% MD.
Caveats: This also includes games 2 and 3 for best-of-3 matches. Accordingly, you may see bigger
differences in play rate for highly conditional cards like
Description: The win rate of decks with at least one copy of this card in the maindeck, weighted by the number of copies in the deck. Playing ten copies of a card in a deck has ten times the weight as playing one copy.
Description: The number of games where an instance of this card was in the opening hand. Each instance of a card is counted, so a game having two copies of a card has twice as much weight as an opening hand with only one copy.
Caveats: Note that an "opening hand" is the final hand that was kept after any mulligan actions, so a single mulligan will only have 6 cards count towards opening-hand data. Cards put on the bottom are ignored, as are the hands that were mulliganed away. Additionally, this metric is biased by the fact that some cards - usually expensive or otherwise hard-to-cast ones - are more likely to contribute to a hand being mulliganed or to the card being put on the bottom after a mulligan.
Description: The win rate of games where an instance of this card was in the opening hand. Each instance of a card is counted, so a game having two copies of a card has twice as much weight as an opening hand with only one copy.
Caveats: See the note above about the definition of an "opening hand".
Description: The number of times this card was drawn from the deck into hand, not counting the opening hand. This includes cards that were drawn at the start of a turn or due to card drawing abilities. It does not include cards returned to hand from the battlefield or graveyard, tutored, or played from exile. Each copy is counted, so a game having two copies of a card drawn has twice as much weight as a game with only one copy drawn.
Note that "tutoring" is defined as any effect where the player chooses from cards in their library. This includes
effects like those of
Caveats: It is possible for the same instance of a card to be counted if it is put back into the library
and drawn again, so cards that put themselves back in (e.g.
Description: The win rate of games where an instance of this card was drawn from the deck into hand, not counting cards from the opening hand. Each instance of a card is counted, so a game drawing two copies of a card has twice as much weight as a game drawing only one copy.
Caveats: See the caveats above about the same instance of a card being drawn multiple times.
Description: The number of times this card was drawn into hand, either in the opening hand or later. Each instance of a card is counted, so a game having a copy of a card in the opening hand and then two more copies drawn later has three times as much weight as an opening hand with only one copy.
Caveats: See the caveats above about the same instance of a card being drawn multiple times. Note that there is a slight possibility that a card is drawn into hand more times than it is in the deck (e.g. if it is shuffled back in).
Description: The win rate of games where an instance of this card was drawn into hand, either in the opening hand or later. Each instance of a card is counted, so a game having a copy of a card in the opening hand and then two more copies drawn later has three times as much weight as an opening hand with only one copy.
Caveats: See the caveats above about the same instance of a card being drawn multiple times.
Description: The copies of a card that were in the maindeck, minus the number of copies that were drawn or tutored ("seen"). If more copies are seen in a game than are in the maindeck, this value is set to 0.
Description: The win rate in games where one or more copies of the card was in the maindeck but not seen at any stage of the game, weighted by the number of copies seen.
Caveats: See the notes above for Number of Games Not Seen.
Description: The difference between Games in Hand Win Rate and Games Not Seen Win Rate.
Caveats: This metric is a simple difference and does not weight by the number of games in each situation, which may overvalue powerful late-game cards.
The arrows indicate that the card is an outlier for the metric. We define an outlier as any value that is at least 1.5 standard deviations away from the mean. You can hover over the value to learn more.
When you hover over values in these columns, you can learn:
22 of 230)z-score (how many standard deviations its value is from the mean)When we use the terms "maindeck" or "sideboard" throughout the site, we are referring to the placement of cards at the start of each game, not match.
Unless otherwise indicated, when we change the computation of a statistic, it only affects data for that date moving forward.
2023-09-21: We added a definition for Play Rate and clarifying note about "maindeck" and "sideboard".
2022-10-16: We updated the calculation for games-drawn and games-in-hand. Prior to this change, tutored cards were unintentionally included in these calculations. Additionally, the games-not-drawn metric was renamed to games-not-seen to account for the fact that it also excludes cards tutored to hand. These changes were backfilled to 2022-09-01, when DMU was released on Arena.
2022-02-27: A major overhaul to in-game statistics outlined here. These changes were backfilled to 2022-02-10, when NEO was released on Arena. Prior to this date GD and GIH included cards bounced to hand, not just drawn from the deck. Additionally, prior to this, GP and GND only counted at most once per game, not per copy of the card.
2021-04-16: Prior to this date, we would only throw out some of the data - Opening Hand, Games Drawn, Games in Hand, Games Not Drawn - if there was a discrepancy between gameplay and deck data, but we would still count the Games Played data. Because of this, we likely included some wins and losses in Games Played data with decks that did not completely match the actual deck played.
2021-03-29: Until this point, we only counted at most one instance of a card for any given game for the Games in Hand data.
2021-03-29: Until this date, we occasionally counted gameplay statistics for multiple deck colors if a user played decks with different color designations in the same event. This most often happened if someone splashed a card in some games but not others.
2021-01-24: Added the ability to filter on time ranges and on a per-maindeck-color basis.