Ask most players whether they enjoy statistics and you will get a flat no. Watch the same people spend forty minutes deciding whether to remove a Strike from a Slay the Spire deck and you are watching someone perform hypergeometric reasoning for fun. The roguelike deckbuilder genre pulled off something educators have attempted for a century. It made the arithmetic the interesting part, not the obstacle, by attaching it to a decision that matters immediately and punishes you for getting it wrong.
Deck Thinning Is the Cleanest Lesson in Games
Start with the single most counterintuitive idea in deckbuilders, and the one every new player resists.
Removing a card makes your deck better. Not a bad card specifically. Any card. The maths is unforgiving, and once it lands, it reorganizes how you play.
Here is the formula doing the work. If your deck has N cards and exactly one copy of the card you want, your chance of drawing it in an opening hand of five is precisely 5/N. It falls straight out of the hypergeometric distribution, and nothing more complicated than that is required.
- A 20-card deck gives you 25%.
- A 15-card deck gives you 33.3%.
- A 10-card deck gives you 50%.
So removing five cards from a twenty-card deck improves your odds of seeing a key card by a third, without adding anything. The cards you already own simply arrive more often.
With two copies, the numbers climb steeply: 44.7% at twenty cards, 57.1% at fifteen, 77.8% at ten. Which is why experienced players treat card removal as more valuable than most card additions, and why the shop’s remove service is frequently the best purchase available.
New players almost universally get this backward, because acquiring things feels like progress and discarding them feels like loss. The game teaches otherwise through about thirty failed runs.
Expected Value, Taught by Being Punished
The second concept the genre drills is the one that matters most outside it.
Every run presents the same decision shape. Take the reliable option or the volatile one? Ten damage guaranteed, or a coin flip for twenty-five?
The arithmetic is trivial. A 50% chance of 25 is worth 12.5, which beats a guaranteed 10. A 40% chance of 25 is worth exactly 10, so the two are identical and the choice comes down to something else entirely.
What the genre adds, and what makes it a genuinely good teacher, is that expected value is not always the right answer. A run has a finite number of combats, which means you cannot rely on things averaging out. If a single loss ends the attempt, the guaranteed 10 is frequently correct even when the coin flip has higher expectation.
That distinction between expected value and survivability is a genuinely advanced idea, and thousands of people have internalized it without ever encountering the term. It is also why the same player will take the volatile option in an early combat and refuse it against a boss, which is correct reasoning rather than inconsistency.
Balatro Made Compounding Visible
Compound growth is famously hard to intuit. Humans are built for linear estimates and consistently underestimate what repeated multiplication does.
Balatro solved that by making it the whole game. Scores do not add; they multiply, and the difference between a run that ends at eight thousand points and one that ends in the billions is a handful of decisions about how multipliers interact.
Players learn, rapidly and without instruction, that increasing a multiplier is worth more than increasing a base value, and that two multipliers together are worth far more than the sum of their parts. That is an exponential intuition arriving through play rather than through a textbook, and it sticks considerably better.
One developer built it, it sold in enormous numbers, and its central appeal is arithmetic. That is a genuinely strange sentence and it is accurate.
Conditional Probability, Without the Name
The third concept is the one players use constantly without ever calling it that.
Mid-combat, you have drawn seven cards from a fifteen-card deck. The odds on the remaining eight are no longer the odds you started with. Every card seen changes the composition of what is left.
Experienced players track this instinctively. They know roughly what is still in the draw pile, they play around what could appear, and they adjust when a shuffle resets the state. That is conditional probability in operation, and nobody who does it fluently would describe it that way.
The genre also teaches its inverse. Systems with a pity mechanic, where a guaranteed outcome arrives after enough failures, actively break the independence assumption. Knowing which systems have memory and which do not is the difference between correct play and the gambler’s fallacy, and games make that distinction concrete in a way abstract explanation never does.
Casino Games Are the Memoryless Benchmark
Casino games sit at the opposite end of that spectrum from a pity system, and deliberately so.
A licensed slot or table game is memoryless by design: the previous spin has no bearing on the next one, and independence is exactly what an independent test house certifies before the title reaches a regulated floor. The published return figure means something because nothing in the machine is quietly adjusting toward it.
Which makes casino mechanics a useful reference point for anyone trying to work out which kind of system a game is running. Online casino comparisons in states with no domestic licensing frame it the same way, and much of what Florida users prefer comes down to the same instinct a deckbuilder trains: a known distribution beats a hidden one, even when the known one is worse.
A gacha pull with an undisclosed rate and a soft pity curve is the harder system to reason about, not the easier one, precisely because you cannot tell where you are in it.
Why This Works When Classrooms Struggle
Three reasons, and they generalize to anything anyone is trying to teach.
- The feedback is immediate and personal. Get the deck maths wrong and you lose a run you were invested in. Nothing in a problem set carries that.
- The reasoning is instrumental rather than assigned. Nobody is calculating draw odds because they were told to. They are doing it because it determines whether they win, which is a completely different motivational structure.
- Repetition is built into the form. A roguelike is designed to be replayed, so the same class of decision arrives hundreds of times with varying inputs. That is spaced practice, delivered by a system that people voluntarily return to.
It is also why the genre produces such argumentative communities. A disagreement about whether to take a card is a disagreement about a calculation, which is resolvable, and our gaming opinion pieces archive is full of the same instinct applied to less tractable questions.
The genre did not set out to teach anyone anything. It just happened to build systems where probabilistic reasoning is the skill that separates a good run from a bad one, and then made losing enjoyable enough that people kept practicing.
Which is more than most educational software has managed, and it was entirely a side effect.
