
Slay the Spire 2
- PC
The deckbuilder that defined the genre is back in Early Access, and the bones are already excellent.
No tested score yet
Buy below: Under $26.99 in Early Access; wait for 1.0 if you want the full run.

The deckbuilder that defined the genre is back in Early Access, and the bones are already excellent.
No tested score yet
Buy below: Under $26.99 in Early Access; wait for 1.0 if you want the full run.
A turn-based Unity build on modern hardware is a stress test the game cannot lose: the 240 fps cap exists because the engine could go higher. Battery life on Steam Deck matches the original's reputation, and sleep-resume mid-run works exactly the way a roguelike demands.
Stability for an Early Access title is quietly excellent. Monthly patches with readable changelogs, rare crash reports, and no save corruption across our test runs.
Early Access patch cadence has run roughly monthly with hotfixes in between, and the save format has survived every balance swing without corruption, including two rewrites that invalidated mid-run builds; carry-over logic is documented in the patch notes, which is more than most Early Access titles manage.
The interface scales cleanly at 4K and on handheld: card text stays legible at Deck resolution, and the latency between a card click and its animation sits under one frame — a small number that deck game veterans feel immediately.
Art and sound continue the first game's economy of scope: no voice acting, no cutscenes, and every asset spends its budget on readability. A run is a hundred small decisions, and the presentation never buries one of them.
The bones are the first game's bones, tightened: card pools, relics, and elite paths with two new characters whose mechanics genuinely change how a run is planned rather than reskinning old ones. Mod support, the first game's long tail, is on the roadmap.
The new characters are structural additions rather than reskins, each pulling the card pool in a direction the first game never attempted. The ascension ladder is back, and the balance argument among Early Access players has already restarted, which is how you know the design landed.
The modifier system on map nodes extends the first game's elite-path decisions into a run-defining layer, and the community's tier-list arguments restarted within hours of each patch, which is the healthiest sign of depth a deckbuilder can show.
The hold verdict is arithmetic, not criticism. The EA build is a fraction of the 1.0 content, and balance patches swing hard enough that a beloved build can break overnight. If you need the complete thought, wait; if you want to grow with it, the bones are already excellent.
Buy-to-play with no store, no season pass, and no DLC pressure: the model is the old one, and Mega Crit's track record says the 1.0 price will not punish early buyers.
The EA price undercuts the expected 1.0 sticker, which makes joining now a fair deal for deckbuilder veterans. If the first game is anything to go by, 1.0 will also be the last time you pay for it.
Everyone else: wishlist, wait for 1.0, and enjoy the definitive version of the genre this studio created.
Unity build runs cool and fast; 240 fps is the cap doing its job, not a demand. Turn-based load means any modern laptop is overkill.
Early Access patches land roughly monthly with changelogs that respect the player. Crash reports are rare for an EA title.
Buy-to-play, no store, no DLC roadmap pressure. Mega Crit's model is the old one: you buy the game, you get the game.
Steam client only, no Denuvo, offline play works.
Full gamepad support with card-targeting that feels deliberate, carried over from the first game's console ports.
The original was the definitive Deck game and the sequel keeps the crown: 60 fps, hours of battery, perfect sleep-resume.
The EA build is roughly the first game's core with new characters; the price is fair, but 1.0 is the complete thought.
The score is weighted from re-runnable dimensions, with the evidence recorded row by row above. Games we have not tested get no score and show "Not tested". Thresholds and weights live in src/lib/; the content layer holds facts only.
Full method in Methodology.