The metagame

Which decks win, which cards get played, and how the archetypes match up. Every number says where it came from, and a dash means there are not enough games to say.

Only simulated bot-vs-bot games on published decks. · 0 simulated · last 90 days

About these numbers (2 notes)
  • Simulated on-play / on-draw splits are hidden: calibration could not be decided yet.
  • Simulated mulligan rates are hidden: calibration could not be decided yet.

How these numbers are made

Games counted

0

0 simulated

Average length

—

per player, per finished game

Going first wins

—

needs more games, or a calibrated simulator

Archetypes tracked

0

30+ games each

The decks winning most

No deck has played enough games to rank yet.

Rating is the win rate a record can defend, so volume beats a hot streak. A deck needs 15 games to be ranked. Browse every community deck

Most played cards

Every card ›
No card has been played in this window yet.

Deck matchups

Full matrix ›
Not enough games yet to compare archetypes. Each needs at least 30 games, and each pairing at least 12.

Archetype board

Every archetype ›
1

Blue/Purple Midrange

24 published decks · no new decks

—

needs 30+ games

Share
43% of published decks
Rating
—
Length
—
Mulligan
—
On the play
—
On the draw
—
Games
0

How it wins

—
2

Blue/Red Midrange

20 published decks · 6 new decks (was 2)

—

needs 30+ games

Share
36% of published decks
Rating
—
Length
—
Mulligan
—
On the play
—
On the draw
—
Games
0

How it wins

—
3

Blue/Green Midrange

4 published decks · 2 new decks (was 0)

—

needs 30+ games

Share
7% of published decks
Rating
—
Length
—
Mulligan
—
On the play
—
On the draw
—
Games
0

How it wins

—
4

Purple/Red Midrange

4 published decks · no new decks

—

needs 30+ games

Share
7% of published decks
Rating
—
Length
—
Mulligan
—
On the play
—
On the draw
—
Games
0

How it wins

—
5

Purple/Red Aggro

3 published decks · 0 new decks (was 1)

—

needs 30+ games

Share
5% of published decks
Rating
—
Length
—
Mulligan
—
On the play
—
On the draw
—
Games
0

How it wins

—
6

Green/Purple Midrange

1 published deck · no new decks

—

needs 30+ games

Share
2% of published decks
Rating
—
Length
—
Mulligan
—
On the play
—
On the draw
—
Games
0

How it wins

—

How these numbers are made

Two kinds of games. Player games are finished online, friendly, private and tournament matches between people. Simulated games are bot-versus-bot games on published decks, played on our own machines with the same rules engine as the game, by bots at Hard or higher (the nightly run plays Hard; deeper checks of a single deck can play Elite or Grandmaster). The Combined view counts a simulated game as 0.25 of a player game; Players and Simulated show each on its own. Practice games against a bot are not counted as player games.

Share and trend. Which pairings get simulated is decided by our scheduler, not by players, so simulated games say nothing about popularity. Share of play therefore uses player games only. Where there are none, we show the share of published decks instead and say so; its trend is how many decks were created in the last 14 days against the 14 before.

Rating. The win rate a record can defend: the lower end of a 95% confidence interval, so a long record counts for more than a lucky streak. Draws count as half a win.

Small samples are hidden, not rounded. An archetype shows a rate from 30 games, a pairing from 12, a deck rating from 15, a card win rate from 30. On-play and on-draw need 15 games each way. A dash means not enough games. Simulated on-play, on-draw and mulligan figures also need the simulator to pass its calibration check, which is run at Hard (games at a stronger level count toward win rates only), and say why when they are hidden.

Archetypes are read from the list. Nobody writes their plan on the tin, so the label comes from the 50 cards: the colour pair, then a role. Ramp: 6+ cards that add souls or discount gear. Control: 7+ that remove, sweep or stun Pals. Combo: 14+ that search the deck or recur from the graveyard. Aggro: an average cost of 3.9 or less. Anything else is Midrange. It is a heuristic, and it can disagree with what a deck's author calls it.

Length is turns per player, not counting the opponent's. Simulated bots never concede or run out the clock, so a simulated “how it wins” can only split damage from decking. Simulated win rates are against the decks we schedule, which favours well-known decks.