A deckbuilder hands you a weak starting deck and then lets you change it between fights. Slay the Spire opens you with ten cards, most of them Strike and Defend; by the end of a good run several of those are gone and the survivors all do something specific. Everything the genre asks of you happens on the screen between encounters, and it comes down to three verbs.
An encounter deals a hand from your deck. You spend a limited resource — energy, mana, actions — to play cards until one side falls over. That tells you whether the last shop screen was right; the shop screen is where the game is played.
Add: take one of three cards on offer, or take none. Remove: pay to delete a card you already own — in Slay the Spire the price rises with every removal you buy in a run, which is the game pricing the strongest move it sells. Upgrade: sharpen a card you kept, usually at a campfire whose alternative is healing, so the two compete for the same stop.
Ask an experienced player which skill takes longest and they will say refusing good cards. That sounds like posturing until you know the shuffle rule.
When the draw pile runs out, the discard pile is shuffled and becomes the new one. Nothing is ever used up. Every card you own comes back, which means a deck is not a collection of options but a probability distribution, and adding a card permanently changes the average hand you will be holding for the rest of the run.
That inverts the instinct every other genre trains. A merely good card makes a deck worse when it already holds a great one, because it stands between you and that card: four brilliant cards in a deck of twenty come up less often than three in a deck of ten. Small and mean beats broad and interesting, and feeling that trade-off is most of the learning curve.
Dominion — Donald X. Vaccarino's 2008 tabletop game, which pioneered building your deck during play rather than the week before — makes the point without mercy. You start with seven Coppers and three Estates, and the Estates are victory points that do nothing while you hold them. One of the strongest cards in its base set is Chapel: no money, no cards, no points — it only lets you bin up to four cards from your hand, and that is enough to bend the game around it.
This is the distinction the storefront keeps blurring. In a collectible card game you build the deck first, from cards acquired elsewhere, and the match tests a plan you already made — Marvel Snap at its shortest: twelve cards, six turns, three locations, deck finished before the first card is played. In a deckbuilder you arrive with almost nothing and the building is the play. Same objects, opposite pleasure, and it belongs in the first line of any pitch.
What made it a modern staple was the fusion with the roguelite: a procedural gauntlet needs rewards worth choosing between, and a deckbuilder needs stakes and a reason to keep drafting. Each solved the other's problem. That instinct — draft an engine, then watch it perform — is shared with auto-battlers, which swap the hand for a board of units, and pushed to its limit it becomes an idle game, where the engine runs without you. The deck need not even be spells: Balatro's is a standard fifty-two-card pack, its engine a row of jokers that rewrite how a poker hand scores, and it is unmistakably the same genre.
Once a community has played enough, most deckbuilders reveal a handful of dominant archetypes, and the interesting question — what can be made of what you were given? — collapses into a duller one: did the key card turn up? A run gets decided by an offer in the first act, and the rest is administration.
The counters all work by keeping questions open. Card pools wide enough that no list is memorisable. Enemies built to invalidate strategies, so a deck that only deals damage, or only blocks, or needs three turns to start, meets something that punishes it. Legible randomness: you should be able to look at a bad hand and trace it to the decision that caused it, because a loss you could not have planned around spends your time rather than testing you. And a ladder — Slay the Spire's is Ascension, twenty levels of stacking handicaps, each reopening a question the level below let you close. None of that is permanent; what a designer controls is how long the solve takes and how much of the pool stays interesting afterwards.
Type: Deckbuilder (built during the run; a collectible game builds it before)
Core loop: Win an encounter with the hand you have, then add, remove or upgrade for the next
Perspective: 2D card UI, front-on — every card gets read, so type and iconography are the art
Win / lose: Clear the run or the opponent; lose when your health or your draw fails you
Session: A run is one sitting; head-to-head matches run in minutes
Progression: In-run deck sculpting; between runs, new cards, characters and difficulty tiers
If you build one, decide early how a player gets rid of a card. Nearly everything the genre is good at descends from that one affordance — and a deckbuilder without pruning is a game in which every reward is also a punishment.