THE DECK LAB

SCORE ANY DECK AGAINST THE LIVE META

Build a deck and the win-chance model measures it against the live meta, or search single-card swaps that raise its win rate. Below, the decks the model itself rates highest right now.

THE DECK LAB

CRAFT A DECK, MEASURE IT AGAINST THE META

Meta segment

Cards: 0/8
Avg elixir: —
Special: 0/3
Evo: 0/2
Hero: 0/2

Tap a card's EVO or HERO pill to equip a special form — 1 Evo, 1 Hero, plus 1 Wild slot either can take (3 max). The win-chance model scores evolutions; hero forms are shown for the build.

Measure vs meta

Score current deck against live meta.

500

50 · faster6,000 · full panel

Time

~3s est.

Actual coverage

100.0%

ML upgradeExperimental

Compare 144 single-card swaps.

120

24 · faster500 · max

Test all cardsTwo passes: every card vs 24 diverse decks, then 4 finalists per slot vs selected panel.

Time

~29s est.

Actual coverage

100.0%

More decks increase meta coverage and runtime. Coverage comes from current live deck frequencies; times remain rough estimates.

121 cards

THE MODEL'S OWN RANKING

TOP DECKS ACCORDING TO THE WIN-CHANCE MODEL

Ranked Ladder

Scoring the meta with the win-chance model — first build can take a minute…

ANALYST NOTES

The model board is the win-chance model's opinion: each popular deck scored against the meta panel weighted by how often a fair matchmaker would draw each opponent. It knows matchups, not piloting difficulty.
The deck lab and the experimental ML upgrade run on the same model and the same meta panel, and are also exposed as commercial API endpoints under /api/v1.
This is descriptive data. Small samples, mode mix, balance patches, card levels, and selection bias can all move a deck up or down before the broader player base confirms it.

THE DECK LAB

SCORE ANY DECK AGAINST THE LIVE META

Build a deck and the win-chance model measures it against the live meta, or search single-card swaps that raise its win rate. Below, the decks the model itself rates highest right now.

THE DECK LAB

CRAFT A DECK, MEASURE IT AGAINST THE META

Meta segment

Cards: 0/8
Avg elixir: —
Special: 0/3
Evo: 0/2
Hero: 0/2

Tap a card's EVO or HERO pill to equip a special form — 1 Evo, 1 Hero, plus 1 Wild slot either can take (3 max). The win-chance model scores evolutions; hero forms are shown for the build.

Measure vs meta

Score current deck against live meta.

500

50 · faster6,000 · full panel

Time

~3s est.

Actual coverage

100.0%

ML upgradeExperimental

Compare 144 single-card swaps.

120

24 · faster500 · max

Test all cardsTwo passes: every card vs 24 diverse decks, then 4 finalists per slot vs selected panel.

Time

~29s est.

Actual coverage

100.0%

More decks increase meta coverage and runtime. Coverage comes from current live deck frequencies; times remain rough estimates.

121 cards

THE MODEL'S OWN RANKING

TOP DECKS ACCORDING TO THE WIN-CHANCE MODEL

Ranked Ladder

11 most-played decks, each scored against 11 meta decks weighted by play frequency / model v5-e2aa7bfc0c76 / built Jul 07, 01:43 PM

9.1% of scored meta playbest 88%worst 49%volatility 25

65.7%

model WR vs meta

9.1% of scored meta playbest 80%worst 42%volatility 26

63.0%

model WR vs meta

9.1% of scored meta playbest 84%worst 50%volatility 24

62.6%

model WR vs meta

9.1% of scored meta playbest 86%worst 41%volatility 32

61.7%

model WR vs meta

9.1% of scored meta playbest 74%worst 23%volatility 35

50.6%

model WR vs meta

9.1% of scored meta playbest 74%worst 24%volatility 28

50.3%

model WR vs meta

9.1% of scored meta playbest 70%worst 26%volatility 32

50.3%

model WR vs meta

9.1% of scored meta playbest 70%worst 24%volatility 32

41.2%

model WR vs meta

9.1% of scored meta playbest 56%worst 22%volatility 23

40.5%

model WR vs meta

9.1% of scored meta playbest 53%worst 20%volatility 23

33.8%

model WR vs meta

Scoring the meta with the win-chance model — first build can take a minute…

ANALYST NOTES

The model board is the win-chance model's opinion: each popular deck scored against the meta panel weighted by how often a fair matchmaker would draw each opponent. It knows matchups, not piloting difficulty.
The deck lab and the experimental ML upgrade run on the same model and the same meta panel, and are also exposed as commercial API endpoints under /api/v1.
This is descriptive data. Small samples, mode mix, balance patches, card levels, and selection bias can all move a deck up or down before the broader player base confirms it.