I bought Balatro sometime in early 2025 when everyone on Twitter wouldn't shut up about it. I figured I'd play for an hour, decide it was overrated, and move on. Two hundred hours later I'm still beating my head against Gold Stake on different deck types and I have strong opinions about the meta.
It's been the most surprising entry in my library in years. Worth talking about why.
The pitch, and why it lands
Balatro is poker-themed roguelike deckbuilding. You play poker hands — pairs, flushes, straights — to score points. You buy joker cards that modify how scoring works. Each run is twenty or so rounds against escalating point thresholds. If you fail a threshold, the run ends.
That description undersells it. The genius is that the joker interactions create exponential scoring. A naive flush scores in the dozens of points. A flush plus the right four jokers scores in the hundreds of millions. The whole game is figuring out which combinations of jokers, deck modifications, and hand types compound into absurd scores, and which deck states are dead-end traps.
The first time you discover one of these scoring loops yourself, the game opens up. The first time you discover a second one and realize there are dozens — that's when it stops being a game you play and starts being a game you study.
Why it hooks the developer brain
I wrote a while back about strategy games that think like code, and Balatro fits the pattern. The whole game is system design at runtime. You have a budget of slots and money. You have to decide which jokers compose well, which deck of cards supports your strategy, which booster packs to skip. Every shop is a tradeoff between immediate payoff and long-term scaling.
This is the same kind of thinking you do when you're choosing dependencies for a project. Should I take this library now because it solves my immediate problem, or pass because it locks me into a pattern I'll regret? Should I refactor toward the architecture I want, or duct-tape this run together and survive? The answer in Balatro, like in software, depends on what your current state can support.
The other thing that hooks devs is information density. The game shows you exactly what every card does, exactly how much each will score, exactly what the modifiers will do. There's no hidden information about the system, only hidden information about the future deck state. It's a deterministic engine wrapped in a probabilistic shell, which is also a fair description of every program I've ever written.
The brain-rot specifically
Where Balatro gets dangerous is the run length. A losing run takes maybe ten minutes. A winning run is forty-five. That's an evil span. Long enough that "one more run" feels like a real commitment, short enough that "okay, just one more" is plausible. I have lost entire evenings to this loop in a way I usually associate with shorter games like Slay the Spire.
And it has the same brain-worming property as a good puzzle. I'll be in a meeting and find myself thinking about whether Bootstraps stacks better with Mime or with Brainstorm. I'll wake up at 3am with the realization that I should have run a high card build that one time. The game lives in your head between sessions.
What it does that other roguelikes don't
I've played a lot of Slay the Spire. I've played Monster Train, Across the Obelisk, Inscryption. They're all great. Balatro is in the same category, but it does one thing none of them quite manage: the scoring engine is so transparent that you feel like you're a co-author of every absurd combo, not a discoverer of one.
In Spire, when you find a great deck, the satisfaction is recognition — "ah, this archetype works." In Balatro, the satisfaction is construction — "I built this scoring engine from these components, and the math works out, and now I'm doing seven figures of damage per hand." It's a more engineering satisfaction. Probably why it's been so good at hooking devs.
Where I am now
I've beaten the base game on the highest difficulty with about half the deck variants. The Plasma deck is the one that broke me — it doubles a hand's score after multiplying mult by chips, which sounds fine but completely changes which jokers are good. I've cleared it twice. I have respect.
The 1.0 update last year added a friends list and basic multiplayer challenges, which I assumed I wouldn't care about and now use weekly. Trading run seeds with people who play differently than I do has been a small but consistent source of "huh, I never would have built that." It's more interesting than the screenshot-trading Twitter discourse the game initially generated.
Two hundred hours in, I'm not sure I'll ever fully put it down. It joins my short list of games — Factorio being the other obvious one — that I'll boot up indefinitely because the engine itself is interesting, not because there's content I haven't seen yet. The engine is the content. That's a rare kind of game, and Balatro is the best example of it I've played in a long time.