Honest first: this game does not exist yet. No build, no Steam page, no trailer — just a small pile of design notes and a question I can't stop chewing on. So treat this less as "here's a thing I made" and more as "here's a thing I'm circling, and what circling it has taught me."
It starts, like a lot of things do for me lately, with Balatro. I went in expecting a fun little poker thing and came out rattled — it felt less like a game I was enjoying and more like a near-perfect object somebody had carved. The math and the dopamine were the same thing: you optimize, and optimizing is the fun, and that almost never happens. Most games I love, I love despite a part where the smart play and the good time quietly come apart. Balatro just doesn't have that seam — and the moment I noticed it, I couldn't un-notice it.
The nerd-snipe
So I did the thing I always do: asked a question that'll cost me a month of evenings. What's the actual magic here, and does it survive being moved?
Balatro is a deckbuilder — you shape a deck, you draw, the cards do what cards do, quietly and one at a time. Dice are a different animal. Dice are loud. Dice are the roll happening to you, live, with the whole table watching. The tension isn't in your hand, it's in the air for one second while the cubes are still bouncing. My honest hunch is that this is a better substrate for the build-a-scoring-engine feeling, not a worse one — the randomness is more visceral, more "oh no." But a hunch isn't a design, and the interesting part is finding out whether that loudness is a gift or a curse.
The Farkle skeleton is the obvious place to hang it. Push-your-luck is already a tiny perfect engine — roll, bank or risk, bust and lose it all. It's the dice version of "one more hand." Bolt Balatro's roguelite progression onto that and you've got the shape of the thing. Then I looked at who's already here.
- Dice with Death, Dicealot, Dice Gambit — Farkle roguelites already exist, so the idea isn't novel and I should stop pretending it is.
- My honest read, at a glance: the execution is rough, mostly the graphics — which is the only encouraging part. A crowded idea executed badly is a polish gap, and polish I think I can win.
- That's the whole edge I'm claiming. Not "nobody thought of this." Just "I might out-feel the versions that exist."
The design question I'm actually stuck on
Here's the principle this whole thing lives or dies by — the lens I bring to every game I take apart, and the reason Balatro lodged in my head:
"Any time the best strategy isn't the most fun strategy, I believe, is a failure of the game's design."— me, on what I'm trying not to break
Dice make that fragile. With a deck, you can tune the math until optimal is fun, because you control the deck. With dice, the player is staring down raw probability, and the danger is obvious: the math says bank now, every time, it's the higher expected value — and "always make the boring safe choice" is the most fun-killing optimal strategy there is. Build the rewards wrong and the correct play is also the dullest, and the whole thing collapses into a spreadsheet you watch. So most of my notes aren't features. They're me poking at one fault line: how do you make pushing your luck the optimal play often enough that greed and good play point the same way?
The lead I'm chasing is to make the dice themselves the answer. I've been borrowing the weighted-dice system from Kingdom Come: Deliverance 2 — which I love because it's roleplay-through-mechanics, not menus — where every die has a real, lopsided distribution. The Weighted die comes up one two-thirds of the time. Saint Antiochus' die rolls a three. Every time. I pulled the full table of forty-odd KCD2 dice and their exact probabilities, because here the expected value is the game — choosing which lopsided cube to risk on which roll is the decision, and a Balatro engine on top turns that choice into a build. If the dice are interesting enough, maybe the greedy roll stays the fun one. Maybe. That's the experiment.
The tech plan is deliberately small, because I don't want to fall in love with an engine before I've proven the loop is fun: a quick three.js web prototype for the rolling, basic AI, a progression spine for unlocking dice. If the loop sings in an ugly browser build, then it earns more.
One piece of pure flavor I won't let go of, because it makes me laugh: Henry, the KCD2 farmhand, getting visibly filthier the deeper you ascend — so the final boss is the grubbiest possible Henry, and all he does is sit there and play dice. No idea if that survives a real design. I just want it to.
Where this honestly is
Nowhere, in the way that matters. It's notes, a probability table, a hunch, and a principle I'm scared of breaking. My own one-line verdict — "actually great idea, easy to execute and polish" — is the kind of thing past-me writes that future-me finds adorable, because "easy to execute" has never once been true. But I keep coming back to it, and the coming-back is the signal I trust most.
If you want the wider context — the taste behind all of this, or the other system I keep tinkering with instead of finishing this one — those are the next rooms in the rabbit hole. I'll update this page when there's a build worth touching. Until then: me, the dice, and one stubborn question.
Keep going Why Balatro is a perfect game · The games I love, and why