What Is Probability? A Beginner's Guide (With Examples)
Probability is a number between 0 and 1 that measures how likely an event is — 0 means impossible, 1 means certain, and 0.5 means a coin flip. This guide explains it in plain English with worked examples: two heads in a row is 1/4, at least one heads in three flips is 7/8, and a 99% accurate medical test with a 1% base rate gives a positive result that is only 16.7% likely to be real.
Free calculator
Try the free Probability Calculator →
Work out probabilities for coins, dice, cards and repeated events — with every step of the working shown.
Calculate a probability freeWhat Is Probability, in Plain English?
Probability is a number between 0 and 1 that measures how likely an event is. A probability of 0 means the event cannot happen; a probability of 1 means it is certain; a probability of 0.5 means it is a coin flip — equally likely to happen or not. The same number can be written as a fraction (1/2), a decimal (0.5) or a percentage (50%): they all say the same thing, and this guide switches between them as convenient.
The basic rule is simple: probability = favourable outcomes ÷ total possible outcomes, when every outcome is equally likely. Flip a fair coin and the probability of heads is 1 favourable outcome out of 2 possible, so 1/2 = 0.5. Roll a fair six-sided die and the probability of a 4 is 1 out of 6, or about 0.1667. Draw a card from a shuffled deck and the probability of an ace is 4 out of 52, or about 0.0769. The Probability Calculator runs these calculations for any numbers you enter, and the Dice Roller lets you test them with real rolls.
Probability is not a prediction about one event. A probability of 0.5 for heads does not mean the next flip will be heads — it means that over many flips, about half will be heads. The more trials you run, the closer the observed fraction gets to the true probability, a result called the law of large numbers. Ten flips might give 7 heads; ten thousand flips will give very close to 5,000. Probability describes the pattern in the long run, not the outcome of the next try.
The Basic Rules: Complement, Addition, Multiplication
Three rules handle most probability problems. The complement rule says that the probability of something not happening is 1 minus the probability of it happening: P(not A) = 1 − P(A). If the chance of rain is 0.3, the chance of no rain is 0.7. This rule is the key to the "at least one" problems in a later section, where the direct calculation is messy but the complement is easy.
The addition rule says that for events that cannot happen at the same time (mutually exclusive events), the probability of either one happening is the sum: P(A or B) = P(A) + P(B). Roll a die: the probability of a 1 or a 6 is 1/6 + 1/6 = 2/6 = 1/3. If the events can overlap, you subtract the overlap so it is not counted twice: P(A or B) = P(A) + P(B) − P(A and B). Draw a card: the probability of a heart or a king is 13/52 + 4/52 − 1/52 = 16/52 = 4/13, because the king of hearts belongs to both groups.
The multiplication rule says that for independent events, the probability of both happening is the product: P(A and B) = P(A) × P(B). Two coin flips both landing heads: ½ × ½ = 1/4 = 0.25. The word "independent" is doing the work here — the rule only applies when the first event does not change the odds of the second, which is the subject of the next section.
One more thing the rules make clear: a probability can never exceed 1. The total probability of all possible outcomes is exactly 1 — something has to happen — so no single event can claim more than the whole. If your calculation produces a probability above 1, it is wrong: you have double-counted, added events that overlap without subtracting, or multiplied where you should have added. A probability of 1.2 is not a strong likelihood; it is an error.
Independent vs Dependent Events
Two events are independent when the outcome of the first does not change the probability of the second. Coin flips are the textbook example: a coin has no memory, so after five heads in a row the next flip is still 50/50. Believing otherwise — that tails is "due" — is the gambler's fallacy, and casinos profit from it. Each flip starts fresh.
Events are dependent when the first outcome changes the odds of the second. Drawing cards without replacing them is the classic case. Draw one card from a full deck: the probability it is an ace is 4/52. If it is an ace and you keep it, the deck now has 3 aces in 51 cards, so the probability the second card is also an ace is 3/51. The probability of two aces in a row is therefore (4/52) × (3/51) = 12/2652 = 1/221 ≈ 0.00452, or about 0.45%.
Compare that with drawing with replacement — putting the first card back and reshuffling before the second draw. Now the events are independent again: the probability is (4/52) × (4/52) = 16/2704 = 1/169 ≈ 0.00592, or about 0.59%. The difference is small but real, and it comes entirely from whether the first draw changes the deck. In real life, most repeated events are somewhere between the two: drawing names from a hat is dependent, rolling dice is independent, and weather from one day to the next is dependent in complicated ways that simple models cannot capture.
| Situation | Independent or dependent? | Probability |
|---|---|---|
| Two heads in a row (coin) | Independent | ½ × ½ = 1/4 = 0.25 |
| Two aces, card replaced | Independent | (4/52)² = 1/169 ≈ 0.00592 |
| Two aces, card kept | Dependent | (4/52) × (3/51) = 1/221 ≈ 0.00452 |
| Sum of 7 with two dice | Independent (6 of 36 outcomes) | 6/36 = 1/6 ≈ 0.1667 |
| Doubles with two dice | Independent (6 of 36 outcomes) | 6/36 = 1/6 ≈ 0.1667 |
“At Least One”: The Complement Trick
Some of the most useful probability questions are "at least one" questions: at least one heads in three flips, at least one sale from ten calls, at least one six in a few rolls. The direct approach is painful, because "at least one" means one or two or three or more, and each case has to be calculated separately. The complement trick turns it into one easy calculation: P(at least one) = 1 − P(none).
Flip a coin three times. What is the probability of at least one heads? The complement — no heads at all — means three tails in a row: ½ × ½ × ½ = 1/8. So P(at least one heads) = 1 − 1/8 = 7/8 = 0.875, or 87.5%. There are 8 possible sequences of three flips, and 7 of them contain at least one heads — only TTT does not. The complement calculation gives that answer in one line.
Here is a sports version. A basketball player makes 70% of her free throws. What is the probability she makes at least one of three? The chance of missing all three is 0.3 × 0.3 × 0.3 = 0.027, so the chance of making at least one is 1 − 0.027 = 0.973, or 97.3%. Even a mediocre shooter is almost certain to make at least one of three — repetition is powerful, and the complement trick shows exactly how powerful.
One more: what is the probability of rolling at least one six in four rolls of a die? The chance of no six in one roll is 5/6, so the chance of no six in four rolls is (5/6)⁴ ≈ 0.4823, and P(at least one six) = 1 − 0.4823 = 0.5177, or about 51.8% — just better than even. Four rolls is the tipping point: with three rolls the answer is 1 − (5/6)³ ≈ 0.4213, below half. These small calculations explain why "at least one" events feel more common than people expect — and why the complement is the first tool to reach for whenever you see the words "at least".
Conditional Probability and Bayes' Theorem
Conditional probability is the probability of an event given that another event has happened, written P(A | B) — the probability of A given B. Drawing a second ace after the first was an ace is a conditional probability: P(second ace | first ace) = 3/51. The condition changes the sample space, and the probability changes with it.
Bayes' theorem reverses a conditional probability: it tells you how to update a belief when new evidence arrives. The formula is P(A | B) = P(B | A) × P(A) / P(B). The most famous illustration is a medical test, and it is famous because the answer surprises almost everyone — including doctors.
Here is the classic example, with the numbers stated clearly. A disease affects 1% of the population (the base rate). A test for it is 99% sensitive — it correctly identifies 99% of people who have the disease — and it has a 5% false-positive rate — it wrongly says 5% of healthy people have the disease. You test positive. What is the probability you actually have the disease? Most people guess around 99%, or at least 90%. The real answer is 16.7%.
The reason is the base rate.
Worked Example: 10,000 People Take the Test
Imagine 10,000 people taking the test. Of these, 100 have the disease (1%) and 9,900 do not. The test catches 99 of the 100 sick people — 99 true positives. But it also flags 5% of the 9,900 healthy people — 495 false positives. So there are 594 positive results in total, and only 99 of them are real: 99 / 594 = 0.1667. The false positives swamp the true positives because healthy people vastly outnumber sick ones. Bayes' theorem gives the same answer in one line: (0.99 × 0.01) / (0.99 × 0.01 + 0.05 × 0.99) = 0.0099 / 0.0594 = 0.1667.
| Group (of 10,000 people) | People | Test positive | Test negative |
|---|---|---|---|
| Have the disease (1%) | 100 | 99 (true positives) | 1 |
| Do not have the disease (99%) | 9,900 | 495 (false positives) | 9,405 |
| Total | 10,000 | 594 | 9,406 |
The same table shows the good news: a negative result is very reassuring. Of the 9,406 negative results, 9,405 are healthy people, so P(no disease | negative) = 9,405 / 9,406 ≈ 0.9999, or 99.99%. The test is far better at ruling the disease out than at ruling it in — a common pattern with screening tests for rare conditions. The lesson is general: when the base rate is low, even a good test produces mostly false positives. That is why a positive screening result is always followed by a second, more specific test. Try your own numbers in the Bayes' Theorem Calculator — change the base rate to 10% and watch the answer jump, which shows how much the base rate matters.
The Binomial Distribution
The binomial distribution describes the number of successes in a fixed number of independent trials, where each trial has the same probability of success. Flip a coin 10 times and count the heads; take 20 shots and count the baskets; survey 100 customers and count the satisfied ones — all binomial, as long as the trials are independent and the success probability stays the same.
The probability of exactly k successes in n trials is P(X = k) = C(n,k) × p^k × (1 − p)^(n−k), where C(n,k) counts the ways to choose which k trials succeed. Take the coin example: n = 10 flips, p = 0.5, and ask for exactly 5 heads. C(10,5) = 252, so P(X = 5) = 252 × 0.5¹⁰ = 252 / 1024 ≈ 0.2461, or 24.6%. Five heads is the single most likely outcome — yet it still happens less than a quarter of the time, because the probability is spread across all eleven possible outcomes from 0 to 10 heads.
The distribution's centre and spread have simple formulas. The mean is np = 10 × 0.5 = 5, and the standard deviation is √(np(1 − p)) = √2.5 ≈ 1.58. Those are the same ideas of average and spread from our statistics guide, applied to a probability distribution instead of a dataset. And the tail is informative too: the probability of 8 or more heads in 10 flips is P(8) + P(9) + P(10) = 0.0547, about 5.5% — unusual, but not extraordinary. That kind of tail probability is the basis of significance testing: an outcome in the far tail is evidence that something other than chance is at work. The Binomial Distribution Calculator computes any of these probabilities for your own n and p.
Counting: Permutations and Combinations
Probability problems often come down to counting: how many ways can this happen, out of how many total ways? Two counting tools cover most cases. A permutation counts arrangements where order matters; a combination counts selections where it does not.
Five runners finish a race, and you want the number of possible 1st- and 2nd-place finishes. Order matters — Alice first and Bob second is different from Bob first and Alice second. That is a permutation: 5P2 = 5 × 4 = 20. But if you are choosing a 2-person committee from 5 people, order does not matter — Alice and Bob are the same committee as Bob and Alice. That is a combination: 5C2 = 10, exactly half the permutations, because every pair can be ordered two ways.
Combinations explain why lotteries are so hard to win. A typical lottery asks you to choose 6 numbers from 49, and the number of possible combinations is C(49,6) = 13,983,816 — nearly 14 million. A single ticket has a 1 in 13,983,816 chance of matching all six. Buying 100 tickets improves that to 100 in 13,983,816, still about 1 in 140,000. The Permutation and Combination Calculator handles the counting for any n and r, and the same counting is what produces the C(10,5) = 252 in the binomial example above — counting and probability are the same subject, seen from two sides.
Probability in Everyday Life
Probability shows up wherever uncertainty does, and a few habits make it easier to reason about. Translate percentages into counts. "A 1% risk" is abstract; "1 in 100 people" is concrete, and "10 in 1,000" is easier still. The medical-test table above works because it turns percentages into people.
Ask for the base rate. Any claim about a test, a prediction or a detection system is incomplete without the base rate. A 99% accurate test for a 1-in-10,000 condition produces almost entirely false positives; the same test for a common condition is highly informative. The base rate is not a technicality — it is often the whole story.
Distrust streaks as evidence. Random sequences contain streaks. In 100 coin flips you should expect a run of 6 or 7 heads somewhere; it is not evidence that the coin is biased or that tails is due. Human brains are pattern-finders, and randomness is the one pattern they systematically misread.
Remember that rare events happen to someone. A 1-in-a-million event, applied to 300 million people, happens to about 300 of them. The lottery winner is not lucky in any meaningful sense — someone had to win. The probability that a specific person wins is tiny; the probability that someone wins is high. Keeping those two questions separate is the difference between a fair assessment of risk and a misleading one.
Sources: all figures in this guide were recomputed in Python from the examples stated in the text (coin and dice outcomes, the complement and binomial calculations, and the Bayes medical-test example with its stated base rate of 1%, sensitivity 99% and false-positive rate 5%). Last reviewed: 15 November 2026.
Frequently Asked Questions
What is probability in plain English?
Probability is a number between 0 and 1 that measures how likely an event is: 0 means impossible, 1 means certain, and 0.5 means a coin flip. For equally likely outcomes it is favourable outcomes divided by total outcomes — the chance of heads on a fair coin is 1 out of 2, or 0.5.
What is the difference between independent and dependent events?
Independent events do not affect each other — coin flips are independent, so two heads in a row is ½ × ½ = 1/4. Dependent events do affect each other: drawing two aces without replacing the first card is (4/52) × (3/51) = 1/221, because the first draw changes the deck.
How do I calculate "at least one" over several tries?
Use the complement: P(at least one) = 1 − P(none). For at least one heads in three coin flips, P(no heads) = 1/8, so the answer is 7/8 = 0.875. For a 70% free-throw shooter taking three shots, it is 1 − 0.3³ = 0.973.
What is Bayes' theorem?
Bayes' theorem updates a probability when new evidence arrives: P(A | B) = P(B | A) × P(A) / P(B). In the classic medical-test example — a 1% base rate, a 99% sensitive test and a 5% false-positive rate — a positive result means only a 16.7% chance of actually having the disease, because false positives outnumber true ones.
What is a binomial distribution?
It describes the number of successes in a fixed number of independent trials with the same success probability. For 10 coin flips, the chance of exactly 5 heads is C(10,5) × 0.5¹⁰ = 252/1024 ≈ 0.2461. The mean is np = 5 and the standard deviation is 1.58.
Why can't a probability be more than 1?
Because the probabilities of all possible outcomes add up to exactly 1 — something has to happen. No single event can claim more than the whole. If a calculation gives a probability above 1, it is wrong: you have double-counted, or added overlapping events without subtracting the overlap.
Last updated: