A visual essay on probability, from wardrobe choices to Bayes' theorem
What if every probability problem you've ever seen was really just one idea wearing different costumes?
Let me show you what I mean. Suppose you're getting dressed in the dark. You've got three shirts — red, blue, green — and two pairs of pants — jeans and khakis. You reach in and grab one shirt and one pair of pants at random.
What are the chances you end up wearing your favorite shirt — the blue one?
Don't reach for a formula. Just count. How many total outfits are there? Three shirts times two pants: six outfits. How many of those include the blue shirt? Two — blue with jeans, blue with khakis. So the probability is 2 out of 6, or one-third.
That's it. That's the whole game.
You just did something that has a fancy name in textbooks — you computed a probability by counting favorable outcomes and dividing by total outcomes. And I want to make you a promise: by the end of this page, you will see that even the results that seem intimidating — Bayes' theorem, the binomial distribution, the birthday paradox — are doing exactly what you just did. Counting things, then dividing.
But here's what makes this interesting. If probability is just counting, why does it so reliably make fools of us? Put 23 people in a room and there's a better-than-even chance two of them share a birthday — out of 365 possible days. That feels impossible. Or consider this: a medical test is 90% accurate, and it says you're sick. The actual chance you have the disease? Roughly 8%. Our intuition doesn't just stumble on these problems. It falls flat on its face.
The beautiful thing is that each of those surprises dissolves the moment you count carefully. That's what we're here to do.
The notation gets denser. The numbers get larger. But the idea? The idea fits on a napkin. Favorable outcomes divided by total outcomes. That's the entire subject of elementary probability.
Let's count some things together.
Tap the outfits with the blue shirt.
0 out of 6
The wardrobe grid: 3 shirts × 2 pants = 6 equally likely outfits.
Before we can answer any probability question, we need to answer a simpler one first: what could possibly happen?
Suppose you roll a standard six-sided die. What could happen? You could get a 1, 2, 3, 4, 5, or 6. That's it. Six possible outcomes, and every one of them is equally likely. This complete list — {1, 2, 3, 4, 5, 6} — has a name. It's called the sample space, written S. Nothing more than an exhaustive catalog of every distinct outcome.
Two rules for building a good sample space: list everything, and list nothing twice.1 For the counting approach on this page, we'll work with outcomes that are equally likely — like die faces and card draws. That keeps our formula clean: favorable count divided by total count.
Roll two dice — one red, one white. How many total outcomes are there?
You could try listing them: (1,1), (1,2), (1,3)... but that gets tedious fast. Instead, picture a grid. The red die's value runs along one axis, the white die's along the other. Each cell in the grid is one outcome. Six rows, six columns: thirty-six cells. Thirty-six equally likely outcomes.
This grid is going to be our workhorse. Over the next several sections, we'll keep coming back to it, highlighting different cells to answer different questions. The grid never changes — only the question does. And every time, the answer is the same process: count the highlighted cells, divide by 36.
That's the power of having a clear sample space. It turns a vague question — "how likely is this?" — into a concrete one: "how many of these thirty-six cells satisfy my condition?"
So here's the grid — 36 cells, every outcome accounted for. But stare at it for a moment and a question starts to itch: what happens when you start asking things of it?
36 total outcomes
The two-dice sample space: every cell is one of 36 equally likely outcomes. Hover to inspect.
Here's the one formula. The only one on this page. Everything else we do is a technique for making this formula easier to apply:
$P(A) = \frac{|A|}{|S|}$
In plain English: the probability of event A equals the number of outcomes where A happens, divided by the total number of outcomes. Favorable over total.
Back to the two-dice grid.
What's the probability the sum is 7? Look at the grid and find every cell where the two numbers add to 7. There's (1,6), (2,5), (3,4), (4,3), (5,2), and (6,1). That's 6 cells out of 36. So $P(\text{sum} = 7) = 6/36 = 1/6$. Roughly 16.7%.
What about doubles? Doubles are cells where both dice show the same number: (1,1), (2,2), (3,3), (4,4), (5,5), (6,6). That's 6 out of 36. Same probability — 1/6. Interesting: you're exactly as likely to roll a sum of 7 as you are to roll doubles.
What about the sum being greater than 9? Now we need sums of 10, 11, or 12. Count the cells: (4,6), (5,5), (5,6), (6,4), (6,5), (6,6). That's 6 cells. Wait — 1/6 again? Let's double check. Sums of 10: (4,6), (5,5), (6,4) — 3 cells. Sums of 11: (5,6), (6,5) — 2 cells. Sums of 12: (6,6) — 1 cell. Total: 6 cells. $P(\text{sum} > 9) = 6/36 = 1/6$.
Notice what we never needed. We never needed a special formula for sums, or a separate formula for doubles. Every time, we did exactly the same thing: count the cells that qualify, divide by 36. The hard part was never the division. The hard part — the only hard part — was counting correctly.
One more thing worth noting. Probability always lands between 0 and 1. Why? Because the count of favorable outcomes can't be negative (you can't have fewer than zero cells) and can't exceed the total (every favorable outcome is, by definition, one of the total outcomes). Zero means impossible. One means certain. Everything else is somewhere on that line.
But here's a question that should nag at you: what happens when the sample space is too big to draw? A deck of 52 cards has over 2.5 million possible 5-card hands. You can't draw that grid. So how do you count things you can't list?
Select an event, then click the cells you think qualify.
The event highlighter: same grid, different questions. Every probability is "count the highlighted cells, divide by 36."
You're playing poker. Someone deals you five cards, and you glance down at five hearts. A flush. Nice hand — but how nice, exactly? What were the chances?
To answer that, you'd need to know the total number of possible 5-card hands you could have been dealt. That number turns out to be 2,598,960. We are obviously not going to list all 2.5 million of them. Our two-dice grid had 36 outcomes, and we could draw every one. But 2.5 million? We need a way to count without listing.
So how do you count 2.5 million things without writing any of them down? Instead of reaching for a formula, let's just think about what's going on.
Remember the wardrobe from the introduction? Three shirts, two pairs of pants, six outfits. We didn't list all six to know there were six. We just multiplied: 3 times 2.
This is the multiplication principle: if you make one choice from a options, then a second choice from b options (regardless of which first option you picked), the total number of combined outcomes is $a \times b$. Three shirts times two pants, six outfits. Six red-die faces times six white-die faces, thirty-six outcomes.
It extends to any number of sequential choices. Take a 4-digit PIN: 10 choices for the first digit, 10 for the second, 10 for the third, 10 for the fourth. Total: $10 \times 10 \times 10 \times 10 = 10{,}000$. That's ten thousand PINs, counted without listing a single one.
And here's a quick probability payoff. What's the probability someone guesses your 4-digit PIN on the first try? The sample space is 10,000 equally likely PINs. Exactly one is yours. So $P(\text{correct guess}) = 1/10{,}000 = 0.01\%$. Same formula as the dice grid — just a bigger denominator.
A tree diagram makes this vivid. Start at a root node. Branch into 3 shirt options. From each shirt, branch into 2 pants options. Count the leaves at the bottom: 6. The multiplication principle is just saying "the number of leaves equals the product of the branching factors." It's still counting — you're just doing it structurally instead of one-by-one.
Now here's a puzzle that leads somewhere important. How many ways can you arrange 4 books on a shelf?
For the first position, you have 4 choices. Once you place a book, you have 3 remaining for the second position. Then 2 for the third. Then 1 for the last. By the multiplication principle: $4 \times 3 \times 2 \times 1 = 24$ arrangements.
This pattern — $n \times (n-1) \times (n-2) \times \ldots \times 1$ — shows up so often that it gets its own notation: $n!$, read "n factorial." So $4! = 24$. It counts the number of ways to arrange n distinct objects in a line.
What about a partial arrangement? Suppose you have 10 books but only 3 shelf positions. You choose an ordered sequence of 3 from 10: that's $10 \times 9 \times 8 = 720$ arrangements. We call this $P(10, 3)$ — the number of permutations of 3 items from 10. The formula is $P(n,k) = n! / (n-k)!$, but that's just a compact way of writing "multiply the top k falling numbers starting from n."
But there's a subtle question lurking here. In those book arrangements, order mattered — "ABC" and "BCA" were different. What happens when order doesn't matter?
Here's a question that sounds simple but has a trap in it.
You have 4 colored balls — red, blue, green, yellow. You're going to choose 2 of them. How many different choices can you make?
If you reach for the multiplication principle, you'd say: 4 options for the first ball, 3 for the second, that's 12. And you'd be half right. There are indeed 12 ways to pick an ordered pair. But did you want {red, blue} and {blue, red} to count as different choices? If you're just grabbing two balls from a bag, those are the same pair.
This is the fork in the road for all counting problems: does order matter?
When it does, you're counting permutations. When it doesn't, you're counting combinations. And the relationship between them is wonderfully clean.
Those 12 ordered pairs group naturally into 6 unordered pairs. Every unordered pair — like {red, blue} — appears exactly twice in the ordered list (as red-then-blue and blue-then-red). So the combination count is 12 / 2 = 6. We divided by 2 because that's $2!$ — the number of ways to rearrange 2 items.
That's the general formula. To count combinations of k items from n (recall that $n!$ means $n \times (n-1) \times \ldots \times 1$):
$C(n, k) = \frac{n!}{k!(n-k)!}$
Start with the permutation count $(n! / (n-k)!)$, then divide by $k!$ to erase the redundant orderings. You're not learning a new concept — you're taking the permutation count and removing duplicates. It's still counting. You're just being careful not to count the same group twice.
This number $C(n,k)$ — also written as "n choose k" — shows up everywhere. How many 5-card poker hands can you deal from a 52-card deck? $C(52, 5) = 2{,}598{,}960$. You would never list them all, but you can count them precisely.
Here's something lovely. If you arrange the values of $C(n,k)$ in a triangle — n going down the rows, k going across — you get Pascal's triangle. Each entry is the sum of the two entries above it. This isn't a coincidence: choosing k items from n+1 either includes the new item ($C(n, k-1)$ ways) or doesn't ($C(n, k)$ ways). The triangle is a counting table — every single number in it counts something concrete.
Pascal's triangle: each entry is $C(n,k)$ — the number of ways to choose $k$ items from $n$. Every number is the sum of the two above it.
We've now assembled the full counting toolkit — permutations, combinations, enormous sample spaces sized up without listing a single outcome. But here's the thing: real probability questions are rarely about one event in isolation. They're about events that overlap. And when events share outcomes, counting gets treacherous.
The outfit tree: each path from root to leaf is one outfit. 3 shirts × 2 pants = 6 leaves.
The factorial cascade: each position has fewer choices. 4 × 3 × 2 × 1 = 4! = 24.
12 ordered pairs collapse into 6 unordered pairs when you divide out the 2! duplicate orderings.
Here's a question that sounds straightforward but has a trap in it: what's the probability of rolling a sum of 7 or at least one 6?
Your first instinct might be to count separately. "Sum of 7" lights up 6 cells. "At least one 6" lights up 11 cells. So... 17 out of 36? But wait. Look carefully at the grid. Some cells are in both groups. The outcomes (1,6) and (6,1) have a sum of 7 and include a 6. If we just add 6 + 11, we've counted those cells twice.
This is the inclusion-exclusion principle, and it's not a formula to memorize — it's a correction for a counting error. When you count A's outcomes and B's outcomes and add them up, you've double-counted anything in the overlap. So you subtract it:
$P(A \cup B) = P(A) + P(B) - P(A \cap B)$
Reading the symbols: $A \cup B$ ("A union B") means all outcomes in either event, and $A \cap B$ ("A intersect B") means the outcomes in both. In counts: $|A \cup B| = |A| + |B| - |A \cap B|$. For our example: 6 + 11 - 2 = 15 cells, so $P = 15/36$.
Sometimes two events can't happen simultaneously — like rolling a sum of 7 and rolling a sum of 11 on the same throw. These are mutually exclusive events: they share no outcomes. When A and B are mutually exclusive, the overlap is zero, so $P(A \cup B) = P(A) + P(B)$. No correction needed, because there's nothing to double-count.
Now let me show you a trick that often makes hard problems easy. Instead of counting what you want, count what you don't want and subtract.
What's the probability that at least one die shows a 6? You could hunt for all those cells — there are 11 of them, scattered around the edges of the grid. Or you could flip the question: what's the probability that neither die shows a 6? That's a tidy 5 × 5 block in the corner: 25 cells. So the probability of not getting any 6 is 25/36, and therefore:
$P(\text{at least one 6}) = 1 - 25/36 = 11/36$
This is the complement rule: $P(\bar{A}) = 1 - P(A)$. It works because "A" and "not A" together account for every outcome — because everything is either in A or not in A. Those two groups, together, account for every cell in the grid. In counting terms: the number of outcomes where A doesn't happen is the total minus the number where it does: $|\bar{A}| = |S| - |A|$.
Complements, unions, intersections — they sound like set theory jargon. And they are. But at the level of our grid, they're just different ways of highlighting cells. Different highlighter colors, same grid, same counting.
What changes when we learn new information?
Venn regions rendered on the actual grid. Every "set operation" is just a different way of coloring cells.
Let me set up a scene. You're playing a dice game. Someone at the table just rolled two dice, and you didn't see the result — but a friend peeks and tells you: "The sum is at least 9."
Now you want to know: what's the probability they rolled doubles?
This is a conditional probability question, and here's why it's different from what we've done so far. Normally, you'd count doubles across all 36 outcomes — there are 6 of them, giving $6/36 = 1/6$. But you've just learned something. You know the sum is at least 9. That rules out a bunch of outcomes. In fact, it rules out most of them.
Go to the grid. Which cells have a sum of at least 9? Count them: (3,6), (4,5), (4,6), (5,4), (5,5), (5,6), (6,3), (6,4), (6,5), (6,6). That's 10 cells. This is your new world — the only outcomes still in play. Everything else has been eliminated by the information your friend gave you.
Within these 10 surviving cells, how many are doubles? Two: (5,5) and (6,6). So the probability of doubles, given a sum of at least 9, is 2/10 = 1/5.
Notice what happened: the unconditional probability of doubles was $1/6 \approx 16.7\%$. After conditioning on the sum being at least 9, it rose to $1/5 = 20\%$. Learning that the sum is high made doubles slightly more likely — which makes intuitive sense, since high sums tend to involve numbers close to each other.
Look what happened. We didn't use a new technique. We used the same old formula — favorable over total. But "total" shrank. When you condition on event B, you're saying "I know B happened, so the only outcomes I care about are the ones where B is true." The denominator changes from $|S|$ to $|B|$. The numerator becomes $|A \cap B|$ — the outcomes where both A and B happen:
$P(A \mid B) = \frac{|A \cap B|}{|B|}$
Or equivalently, $P(A \mid B) = P(A \cap B) / P(B)$. Same fraction, same counting — just in a smaller room.
Here's a question worth pausing on: does learning B always change the probability of A? Not necessarily. If knowing B doesn't change the ratio of favorable-to-total for A, then A and B are independent. Formally, A and B are independent when $P(A \mid B) = P(A)$ — conditioning on B doesn't shrink or grow A's share.
For example, consider "Red die is 4" and "White die is 3." Knowing the white die is 3 doesn't tell you anything about the red die. The two events are independent. But "Sum is at least 9" and "Red die is 6" are not independent — knowing the sum is large makes a high red-die value more likely.
One more thing. By rearranging the conditional probability formula, we get the multiplication rule: $P(A \cap B) = P(A \mid B) \times P(B)$. This is useful for chain reasoning — computing the probability of A-and-B by first asking "what's the probability of B?" and then "given B happened, what's the probability of A?"
And that chain reasoning leads somewhere powerful. We can condition on B to find $P(A \mid B)$. But what if we know $P(B \mid A)$ and need $P(A \mid B)$?
Think about it for a moment. A dog is very likely to have four legs. But does that mean a four-legged animal is very likely to be a dog? Of course not — cats, horses, and tables also have four legs. $P(\text{four legs} \mid \text{dog})$ is high, but $P(\text{dog} \mid \text{four legs})$ is much lower. These two conditional probabilities can be wildly different, and confusing them is one of the most common errors in probabilistic reasoning.
The tool for reversing a conditional probability is called Bayes' theorem. And it's — you guessed it — just counting.
The grid shrinks: conditioning removes outcomes from the world. The formula stays the same — only the denominator changes.
Let me tell you about a medical test.
There's a disease that affects 1% of the population. There's a test for it, and the test is quite good: if you have the disease, it comes back positive 90% of the time. If you don't have the disease, it correctly comes back negative 90% of the time (meaning it falsely says "positive" 10% of the time).
You take the test. It's positive. Quick — how likely is it that you actually have the disease?
You take the test. It's positive. How likely is it that you actually have the disease?
Most people's gut reaction is somewhere north of 80%. The test catches 90% of sick people, after all. But the actual answer is about 8%. You're roughly eleven times more likely to be healthy than sick, despite the positive test.
If that feels wrong, good. That means we're about to learn something. And the tool that makes it clear is — you guessed it — counting.
Let's make it concrete. Imagine 1,000 people walk into a clinic.
Of those 1,000, exactly 10 have the disease (1%). Of those 10, the test correctly identifies 9 (90% sensitivity). Of the 990 healthy people, 10% get false positives: that's 99 people.
Now line everyone up who got a positive test result. You've got 9 truly sick people and 99 falsely flagged healthy people. That's 108 positive tests total. Of those, only 9 actually have the disease.
$P(\text{disease} \mid \text{positive test}) = 9 / 108 \approx 8.3\%$.
No formulas needed. No rounding. Just counting people in groups.
This is Bayes' theorem, and it's simply the arithmetic of counting an overlap from two different directions. Here's the formal version:
$P(A \mid B) = \frac{P(B \mid A) \cdot P(A)}{P(B)}$
But you don't need to memorize it. It falls directly out of the definition of conditional probability. We know $P(A \mid B) = P(A \cap B) / P(B)$, and we also know $P(B \mid A) = P(A \cap B) / P(A)$. Both formulas describe the same overlap — $|A \cap B|$ — just viewed from different sides. Solve the second for $P(A \cap B)$ and plug into the first, and Bayes drops out.
The reason the medical test fools people is the base rate fallacy. We fixate on the test's sensitivity (90%!) and forget to count how many people are sick in the first place. When the disease is rare, even a good test produces more false positives than true positives, simply because the healthy group is so much larger. The prior — $P(A)$ — matters enormously, and our intuition tends to ignore it.
Here's the beautiful thing: you don't need to trust this argument on faith. You can just count. Fill in the numbers. See who ends up in which group. Bayes' theorem is an accounting identity — count the overlap, divide by the margin, get the answer.
So far, every experiment has been a one-shot deal — one roll, one draw, one test. But most of the questions that actually matter involve repetition. A shooter at the free-throw line. A drug tested on a thousand patients. What happens to our counting when the same experiment runs over and over?
1,000 people: color encodes disease status, rings mark positive tests. Adjust the sliders to see how prevalence and sensitivity shape the answer.
A basketball player makes 70% of her free throws. In a crucial 10-shot sequence, what's the probability she makes exactly 7?
We'll get to that answer in a moment. But first, let's build the idea with the cleanest possible example.
Let's flip a coin. Not once — four times. What's the probability of getting exactly 2 heads?
First, the sample space. Each flip has 2 outcomes, and you flip 4 times: $2 \times 2 \times 2 \times 2 = 16$ total sequences. (We can list them: HHHH, HHHT, HHTH, ... all the way to TTTT.)
Now, the favorable outcomes — sequences with exactly 2 heads. This is a counting question: in how many ways can you choose which 2 of the 4 flips come up heads? That's $C(4, 2) = 6$. So:
$P(\text{exactly 2 heads in 4 flips}) = \frac{C(4, 2)}{2^4} = \frac{6}{16} = \frac{3}{8}$
Let me list all six: HHTT, HTHT, HTTH, THHT, THTH, TTHH. Each of these 6 sequences is equally likely, and together they account for 6 of the 16 equally likely outcomes.
The combination $C(4, 2)$ is doing the heavy lifting here — it's counting the number of arrangements of 2 heads among 4 positions. The denominator, $2^4$, counts the total sequences. Favorable over total. Counting again.
For a fair coin, this is pure combinatorics. But what about an unfair coin — say, one that lands heads 70% of the time?
The logic is the same, but now each sequence has a different weight. A sequence with 2 heads and 2 tails has probability $(0.7)^2 \times (0.3)^2 = 0.0441$. Since there are $C(4,2) = 6$ such sequences, and each has the same weight:
$P(\text{exactly 2 heads}) = C(4, 2) \times (0.7)^2 \times (0.3)^2 = 6 \times 0.0441 = 0.2646$
This is the binomial distribution. In general:
$P(X = k) = C(n, k) \cdot p^k \cdot (1-p)^{n-k}$
It looks like a lot of notation. But read it as a sentence: "the number of ways to arrange k successes in n trials, times the probability of any one such arrangement." The combination counts the arrangements. The powers of p and (1-p) weight each one. Still counting — with weights.
Now back to our basketball player: $P(\text{exactly 7 makes in 10 shots at 70\%}) = C(10, 7) \times (0.7)^7 \times (0.3)^3 = 120 \times 0.0824 \times 0.027 \approx 0.267$. About a 27% chance. The combination $C(10, 7) = 120$ counts the arrangements of 7 makes among 10 shots. The probability weights each arrangement. Same formula, real stakes.
Remember Pascal's triangle? Row n of the triangle gives you $C(n, 0), C(n, 1), \ldots, C(n, n)$ — exactly the counts that drive the binomial distribution. When $p = 1/2$, the binomial probabilities are Pascal's triangle, normalized by $2^n$. The triangle isn't an abstract curiosity. It's a map of how likely each outcome is.
One last thing. As n gets large, the shape of the binomial distribution approaches a smooth bell curve — the normal distribution. But that's a story for another page.
How many people do you need in a room before two of them probably share a birthday? The answer has been surprising people for decades. And it falls right out of counting.
The binomial distribution: C(n,k) counts the arrangements, p weights each one. Adjust n and p to see the shape change.
We've spent this whole page building a single idea: probability is counting. Let's take that idea for a victory lap and see how it dissolves some of the most famous puzzles in probability.
Here's a question that has shocked people at cocktail parties for decades: how many people do you need in a room before there's a better-than-even chance that two of them share a birthday?
Before I tell you — take a guess. How many people do you think it takes? Hold that number in your mind.
The answer is 23. Just 23 people in a room of 365 possible birthdays, and the probability of a shared birthday exceeds 50%. That feels absurdly low. But counting makes it clear.
The trick is to count the complement — the probability that nobody shares a birthday. The first person can have any birthday: 365/365 options. The second person needs a different birthday: 364/365. The third: 363/365. And so on. For 23 people:
$P(\text{all different}) = \frac{365}{365} \times \frac{364}{365} \times \frac{363}{365} \times \cdots \times \frac{343}{365} \approx 0.493$
So $P(\text{at least one shared birthday}) = 1 - 0.493 = 0.507$.
Why does this happen? Because we're not asking "does someone share your birthday" — that would require about 253 people for a 50% chance (a different calculation entirely). We're asking "does any pair share a birthday." With 23 people, there are $C(23, 2) = 253$ possible pairs. Each pair is a separate chance for a match. The counting of pairs is what makes the number so surprisingly small.
How many people do you need in a room for a better-than-even chance two share a birthday?
The birthday calculator: watch the probability rocket upward as pairs multiply.
You're on a game show. Three doors: behind one is a car, behind the other two are goats. You pick a door. The host, who knows what's behind each door, opens a different door to reveal a goat. Should you switch to the remaining door?
Think about it for a moment. Does it matter whether you switch?
Yes. Switching wins 2/3 of the time. Here's the counting argument.
When you first picked, you had a 1/3 chance of picking the car and a 2/3 chance of picking a goat. The host's action (opening a goat door) doesn't change those initial odds — it just eliminates one of the losing options. If you picked a goat (2/3 of the time), switching gets you the car. If you picked the car (1/3 of the time), switching gets you a goat. So switching wins in 2 out of 3 scenarios.
That's it. Three equally likely initial configurations, and in two of them, switching wins. Count the scenarios, take the ratio. The answer falls out.
Even poker hand rankings are a counting statement. A flush (5 cards of one suit) beats a straight (5 cards in sequence) because flushes are rarer: there are 5,108 possible flushes versus 10,200 possible straights.2 Fewer flushes exist, so they're harder to get, so they rank higher. Hand rankings are literally a sorted count.
Every one of these "surprising" results — and dozens more you'll encounter in your mathematical life — yields to the same approach. Define the sample space. Count the favorable outcomes. Divide.
Pick a door.
Play the Monty Hall game. Track your results. Switching wins 2/3 of the time.
Remember the wardrobe?
Three shirts, two pairs of pants, one question: what are the chances you grab the blue shirt? You counted — two outfits out of six — and you had the answer in seconds. Everything since then — sample spaces, combinations, Bayes' theorem, the binomial distribution — was the same act, performed on bigger stages. The notation scaled up. The counting never stopped.
But here's what I really want to leave you with.
We counted discrete things on this page. Dice faces. Card hands. People in rooms. Finite sets, every one of them. What happens when the set isn't finite? When you need the probability that a dart lands in a particular region of a target, or that a measurement falls within some range?
Mathematicians didn't abandon counting. They refined it. Integration — the centerpiece of calculus — is counting infinitesimal slices and summing them up. Instead of counting cells in a grid, you measure area under a curve — but the logic is identical. Favorable region divided by total region. The grid became a landscape, the counting became measuring, and the fraction stayed a fraction. The leap from discrete to continuous probability isn't a departure from what we did here. It's the same idea, taken to its limit. Literally.
The concept on this page — favorable over total, ratio of part to whole — isn't just where probability begins. It's the seed of the entire subject.
You already knew how to count before you opened this page. Now you know where that leads.