You can't predict one flip — but a thousand flips hide a number that never moves.
Flip a coin and, for one heartbeat before it lands, the universe genuinely does not know whether it will read heads or tails — and neither do you. This is a random phenomenon: a thing you can repeat under the same conditions, whose single outcome you cannot call in advance. And yet chance is not chaos. Repeat that same unpredictable flip a hundred times, a thousand times, and a startling order rises out of the noise: the share of heads stops wandering and parks itself near one fixed number. That number is the probability — the event's built-in likelihood — and the whole of this stage is the story of how to find it, name it, and reason with it. We begin where every probability begins: by emptying the bag of all possible outcomes onto the table.
Some things are certain: drop a stone, it falls. Some are impossible: the stone does not float away. But most interesting things sit between — a coin toss, a roll of a die, tomorrow's weather, the draw of a card. We call each a random phenomenon when it has two features: you can repeat it under the same conditions, and you cannot predict the single result. Repeatable, but unpredictable — that is the home turf of probability.
The first move is always the same. List every outcome that could happen, exactly once, and gather them into one collection. That collection is the sample space, written Ω (the Greek capital "omega"), and each outcome inside it is a sample point. For a coin, Ω = {H, T} — two points. For one die, Ω = {1, 2, 3, 4, 5, 6} — six. The outcomes need not be equally likely (a bent coin is still random), but they must be complete and mutually exclusive: nothing left out, no two happening at once.
A random phenomenon is repeatable but unpredictable. Its sample space Ω is the bag of all possible outcomes, each listed once. Every probability question starts by writing down Ω.
Toss two coins. The outcomes are HH, HT, TH, TT, so Ω = {HH, HT, TH, TT} — four points. (Note HT and TH are different outcomes; the order of the two coins matters when we list Ω.) Roll two dice and there are 6 × 6 = 36 ordered pairs in Ω. Counting Ω is the doorway to counting probabilities — exactly the permutations-and-combinations toolkit from Stage 32, waiting to walk in.
Conclusion. Random means repeatable-yet-unpredictable; the sample space Ω is the complete bag of outcomes, and you write it first.
"An even roll." "At least one head." "A sum of 7." Each of these phrases doesn't pick out a single outcome — it describes a chunk of the bag. "An even roll" of a die means the outcome is 2, or 4, or 6; that is the subset A = {2, 4, 6} sitting inside Ω = {1,…,6}. This is the definition that makes everything tick: a random event is a subset of the sample space. We say the event A occurs exactly when the actual outcome is one of the points inside A.
Two events sit at the extremes. The whole bag Ω is the certain event — whatever happens, the outcome is somewhere in Ω, so Ω always occurs. The empty set ∅ is the impossible event — it contains no outcomes, so it can never occur. Rolling "a number between 1 and 6" is Ω (certain); rolling "a 7" is ∅ (impossible). Every other event lives in between, and "how big a slice of Ω is it?" is the question probability answers.
One die, Ω = {1,2,3,4,5,6}. The event A = "greater than 4" is the subset {5, 6} — two outcomes. The event B = "prime" is {2, 3, 5} — three outcomes. The event "a 7" is ∅: no face shows 7, so it is the impossible event and can never occur. The event "1 through 6" is the whole bag Ω: certain.
Conclusion. An event is a subset of Ω; it occurs when the outcome lands inside it. Ω is certain, ∅ is impossible.
Knowing which outcomes make up an event still doesn't tell us how likely it is. For that we have to actually do the experiment. Flip a fair coin n times and count the heads, m. The ratio
f = mn = number of times A happenednumber of trials
is the frequency (or relative frequency) of the event. It's the share of trials on which A actually occurred — a number you can measure, always between 0 and 1. Flip 10 times and get 4 heads: f = 4/10 = 0.4. Flip another 10 and get 7 heads: now f = 7/10 = 0.7. The frequency wobbles. For small n it lurches around, because a single lucky streak can swing a short run a long way.
After five heads in a row, is the coin "due for a tail"? No. A fair coin has no memory — each flip is still 0.5, streak or no streak. The frequency only steadies because later trials dilute an early imbalance, never because the coin "corrects" it. Believing chance owes you a make-up is the gambler's fallacy — the most expensive mistake in this whole stage.
Conclusion. Frequency f = m/n is the measured share of trials on which A occurred; for small n it wobbles.
Here is the quiet miracle. Keep flipping. As n grows — 100, 1000, 10000 — the wobble shrinks, and the frequency m/n stops wandering. It settles down and hovers ever more tightly around one fixed number. That bullseye, the value the frequency homes in on as the trials pile up, is the probability of the event, written P(A). It is not something we measure on any single run; it is the event's built-in likelihood, the target the frequencies are aiming at all along.
This long-run-steadiness is so dependable that we take it as the very definition: the probability P(A) is the number the frequency m/n approaches as n → ∞. For a fair coin the heads-frequency settles on P(heads) = 0.5; for a fair die the frequency of "even" settles on P(even) = 0.5 and of any one face on 1/6 ≈ 0.167.
Because every frequency m/n is a share of trials, it can never dip below 0 or climb above 1 — and neither can the number it approaches. So three facts come free:
(1) For any event A, 0 ≤ P(A) ≤ 1. (2) The certain event has P(Ω) = 1. (3) The impossible event has P(∅) = 0. The closer P(A) sits to 1, the more often A occurs in the long run; the closer to 0, the rarer it is.
A weather service records that, of the last 1000 days that looked like today, it rained on 340. The frequency is f = 340/1000 = 0.34, steady enough over so many days that we report P(rain) ≈ 0.34. We did not deduce this from a formula — we read it off the stabilized frequency. That's the empirical face of probability, and it's exactly what a "34% chance of rain" means.
Conclusion. As n grows the frequency steadies near one number, the probability P(A), with 0 ≤ P(A) ≤ 1, P(Ω) = 1, P(∅) = 0.
We now have a complete language. A random phenomenon gives a sample space Ω; an event A is a subset of it; the frequency m/n measures how often A happens; and the number it steadies on is the probability P(A). This empirical definition is honest but slow — you'd have to flip forever to nail P exactly.
The next lesson finds a shortcut. When the outcomes of Ω are equally likely — a fair coin, a fair die, a well-shuffled deck — you don't need to flip at all: P(A) is simply the fraction of outcomes that are favorable, a ratio of counts you can read straight off Ω. The flipping has told us what probability is; Stage 33.2 will let us compute it. And the whole counting toolkit from Stage 32 — permutations and combinations — walks straight in as the numerator and denominator.
Conclusion. Frequency defines probability; equally-likely outcomes will let us compute it by counting, next lesson.
| Idea | What it is | Symbol / fact |
|---|---|---|
| Random phenomenon | repeatable, but its single outcome is unpredictable | — |
| Sample space | the bag of all possible outcomes, each listed once | Ω |
| Event | a subset of Ω; occurs when the outcome lands inside it | A ⊆ Ω |
| certain / impossible | all of Ω / the empty set | Ω / ∅ |
| Frequency | measured share of trials on which A happened | f = m/n |
| Probability | the number the frequency steadies near as n grows | P(A) |
| the bounds | never below 0, never above 1 | 0 ≤ P(A) ≤ 1 |
| the anchors | certain is 1, impossible is 0 | P(Ω)=1, P(∅)=0 |
Write the sample space Ω for tossing two coins, then list the event A = "exactly one head" as a subset. How many outcomes are in Ω, and how many in A?
Ω = {HH, HT, TH, TT}, so |Ω| = 4. "Exactly one head" happens for HT and TH, so A = {HT, TH} with |A| = 2. (HH has two heads, TT has none — neither belongs to A.)
For one roll of a die, classify each event as certain, impossible, or neither: (a) "rolling a number ≤ 6"; (b) "rolling a 0"; (c) "rolling an odd number."
(a) Certain — every face is ≤ 6, so this event is the whole bag Ω and P = 1. (b) Impossible — no face shows 0, so it is ∅ and P = 0. (c) Neither — it is the subset {1, 3, 5}, a genuine in-between event.
A student flips a coin 8 times and gets H H H H H T H T (6 heads). What is the frequency of heads? Does this prove the coin is biased? Explain in one sentence.
Frequency = 6/8 = 0.75. No, it does not prove bias: with only n = 8 trials the frequency wobbles a lot around the true 0.5, and 6 heads in 8 is well within ordinary luck. You'd need many more flips before a steady drift away from 0.5 became convincing evidence.
A spinner is divided so the blue region covers 3/8 of the disc. After a huge number of spins, roughly what frequency of "lands in blue" do you expect, and what is P(blue)?
In the long run the frequency settles on the wedge's area share, so it approaches 3/8 = 0.375. That long-run value is the probability: P(blue) = 3/8 = 0.375. (Any one short run will differ — only the long-run frequency pins down P.)
A roulette wheel has just landed on red five times in a row. A gambler reasons, "black is overdue, so I'll bet black." Name the error and state the truth.
This is the gambler's fallacy. The wheel has no memory: each spin is an independent random phenomenon, so the probability of black is exactly the same after five reds as before — the past streak does not make black "due." The frequency steadies in the long run by diluting imbalances, not by reversing them.
Over 2000 attempts a basketball player makes 1700 free throws. (a) What is her free-throw frequency? (b) Why is it reasonable to call this her probability P(make)? (c) Could a single game show 0 makes in 4 attempts without contradicting it?
(a) Frequency = 1700/2000 = 0.85. (b) Across 2000 trials the frequency has steadied, so we report P(make) ≈ 0.85 — that's the long-run definition of probability. (c) Yes: a 4-shot game is a tiny sample, and a short cold streak (0/4) is perfectly possible; it wobbles around 0.85 without contradicting the long-run value.
Six questions to lock it in. Tap the answer you think is right.
This opening lesson builds the vocabulary of probability from the ground up: a random phenomenon, the sample space Ω, an event as a subset of Ω (with the certain event Ω and the impossible event ∅), the relative frequency m/n, and its long-run stabilization toward the probability P(A) with 0 ≤ P(A) ≤ 1. It directly serves CCSS HSS-CP.A.1 (describe events as subsets of a sample space) and HSS-IC.A.2 (understand long-run frequency / relative-frequency interpretation of probability). The recurring red trap — the gambler's fallacy — is worth dwelling on: it is the misconception that most often survives into adulthood, and naming it explicitly here pays off in every later lesson. Encourage learners to actually run the flip and spinner widgets several times, since seeing the early wobble give way to a steady value is what makes the definition of probability feel earned rather than asserted.