Duel of the Arcana · Deck design report · I of III

Ashka's Inferno

A linear Fire deck built to win as fast as possible with no help from the opponent — designed by reading all 200 cards, then validated by goldfishing hundreds of games through the project's own rules engine.

§1How the deck was found

I read the rulebook, all 200 cards, and the eight characters, then wrote a goldfish simulator on top of the real engine in engine/ — a scripted pilot plays the deck against an inert dummy, and every line is executed by the actual step() reducer, so all the stack, speed, mana, and status rules are enforced for real. The clock counts the pilot's primary actions (the game has no turns). I iterated deck and pilot together: each bad line the traces exposed — a wasted Stoke, a channeled-away payoff card, a hand kept when it should mulligan — became a policy fix, and the statistics below come from the final version.

§2Why Fire, and why Ashka

The key fact about speed in this game: casting mid-chain is free. Only primary actions cost "time," and one Cast opens a chain in which you may keep adding spells as long as you can pay. So the fastest deck is the one that converts a single primary action into the most damage — and Fire is the only arcana that generates extra mana inside the chain while its damage rates are already the benchmark.

Alternatives I considered and rejected: Earth ramps efficiently into The Mountain Walks + Body Slam, but armor halves every Rest and the pieces are clunky in sequence. Celestial has the biggest single hit (Supernova, 27) but pays seven mana for one spell's worth of a Fire chain. Shadow's Deathmark is a six-action clock that ignores health entirely — a fine card, but it can't be accelerated. Ashka is the character: her passive adds free damage to every chain's second spell, and her ability counts as casting a spell, which matters in §3.

§3The chain engine

Three cards make the mana: Burning Ritual 0 · ⚡0 is free and, being speed 0, sits at the bottom of the stack resolving last — so the Fire source it creates lives for the entire chain. Set Your Heart Alight F · ⚡3 nets two sources while it's on the stack. Stoke 1F · ⚡5 doubles every Fire source for the rest of the chain — but it's a spell, so you must let it resolve mid-chain before spending anything, or the doubling is wasted. The payoff is Apotheosis of Fire 3FF · ⚡6: Storm copies it once per spell cast this chain, by anyone.

In practice the deck doesn't need the full ceremony every time. Its floor is honest burn — Immolate ×2 in one chain is 17 with the passive (each sees the other as "another Fire spell cast this chain"), and Overload FFF · ⚡6 is 9 damage as a chain-capper. The engine pieces just turn good chains into lethal ones.

§4The list — 50 cards, mono-Fire

§5Goldfish results

200 games, deck going first, no opposing interaction. The x-axis is the pilot's primary actions taken when the cumulative damage crossed each character's starting health — so each row reads "how fast this deck kills that character if they never interact."

A real 10-action kill from the logs (seed 31676) — no storm needed:

§6Caveats, and what's next

These numbers are a floor, not a ceiling. The pilot is a scripted policy; it still occasionally sequences a chain wrong or digs when it should dump. A human playing perfectly is probably one to two actions faster on the median. Nothing here accounts for interaction — though it's worth noting the deck's better lines are naturally resilient: Overload as a capper gives counterspells only one good target, and a countered spell still raises the storm count.

On Deep Study: I tested a Mind splash (3 Deep Study, 2 Glimpse of Omniscience) and it was slower here — the Mind source dilutes what Stoke doubles, and this deck would rather spend the chain's mana on damage. That's evidence about this shell, not about the card: draw-three-for-two at speed 4 is absurd rate, and it wants a deck built around it. That's exactly where Deck II is headed.

§7Head-to-head: Iggy's "fuck you"

Iggy's list — Mortessa splashing Fire, hand attack plus burn — through the same harness, same dummy, same seeds. Because the standard dummy Thinks every action, its hand balloons and inflates Flame Frenzy (with Mortessa, every card the opponent discards is 1 health); the "resting dummy" row is the fairer read for that deck.

It's fast — faster on average than the Inferno at the top end, and its dream draws are absurd. The recorded 3-action kill: double Set Your Heart Alight (Burned stacks, so the second channels four fleeting sources), Flame Frenzy into a fat hand for 10 health via Mortessa, Dredge Up Darkness refueling from the fresh discard pile, then Zzap, Immolate, and an Ambush Knife in the Dark — 25 damage on the third primary action. Against the biggest characters it's clearly ahead (vs 34 HP: mean 16.2 against the Inferno's 17.2) because health loss and hand punishment don't care about the target's size. The costs: consistency (86% vs 96% — two arcana plus a splash bricks more draws) and blood — the pilot routinely finishes at 12–17 of Mortessa's 26 health, so in a live race every card it casts helps its opponent's clock too.

Two things the goldfish can't see, both in its favor: health loss ignores armor and prevention entirely (Earth and Divine defense blank half my deck, and none of theirs), and Let the Gates Rattle — which my pilot doesn't even attempt — is a real doomsday wincon against interactive opponents. One caveat the other way: this pilot got one round of tuning against the Inferno's five, so its true numbers are a little better than measured. Verdict: as a pure goldfish the two are near-even with the Inferno more reliable; as a deck for real games, "fuck you" trades that reliability for reach my deck simply doesn't have.

Deck I of III · Simulated on the project's rules engine (289 tests green) with a scripted pilot; 200 seeded games per configuration. Deck II — the Deep Study deck — and Deck III to follow.