Probability and Combinatorics Toolkit: A Practical Guide

Probability questions sneak into everyday decisions more often than people notice. Should we worry that three servers failed in the same week? What are the odds a given lottery ticket wins anything? How many users should we sample before we can trust the result? These are probability questions dressed up as business questions, and getting them wrong leads to either wasted caution or expensive surprise. The math behind them is not hard — the hard part is recognizing which tool fits the situation.

This guide covers the small set of tools that handle most real probability problems: counting with permutations and combinations, basic probability laws, conditional probability, Bayes' theorem, and a few standard distributions. Each section has a worked example and a browser-based calculator so you can run the numbers without leaving the page.

Counting: permutations and combinations

Combinatorics is about counting arrangements. Two questions cover most of what you need. First, how many ways can you arrange n things in order? That is n factorial, written n!, and it grows terrifyingly fast — ten items have 3.6 million orderings. Second, how many ways can you pick k things out of n where the order does not matter? That is "n choose k," written C(n,k), and it is the most useful formula in applied probability.

Permutations

Permutations are arrangements where order matters. The number of ways to pick k items from n with order is n! divided by (n-k)!. If you are seating 5 people in 3 chairs, the answer is 5! / 2! = 60. Permutation Calculator handles the factorial arithmetic, which stops being hand-computable somewhere around n=13.

Combinations

Combinations are groupings where order does not matter. Picking a committee of 3 from 10 candidates is a combination — you get the same committee regardless of the order you pick. The formula is n! / (k! × (n-k)!). For the 3-from-10 case, that is 120. Combination Calculator runs the arithmetic and handles the overflow that catches hand-computation on large values.

Wolfram MathWorld has rigorous definitions and extensive cross-references for the combinatorial identities, and Khan Academy's probability track walks through worked examples if you need a refresher from first principles.

Probability laws worth memorizing

Four rules cover almost every basic probability question:

• Sum rule: P(A or B) = P(A) + P(B) − P(A and B). The subtraction handles the overlap you would otherwise double-count.
• Product rule: P(A and B) = P(A) × P(B | A). If A and B are independent, this simplifies to P(A) × P(B).
• Complement rule: P(not A) = 1 − P(A). Often the easier way to compute "at least one" events.
• Total probability: P(A) = sum of P(A | B_i) × P(B_i) across disjoint cases B_i that cover all possibilities.

The complement rule is the one people forget, and it is the one that makes "at least one" problems tractable. Instead of computing P(at least one success in 10 trials), compute 1 − P(zero successes), which is almost always easier. Compute either directly with Probability Calculator, which handles the common distributions and independent-event combinations without requiring you to memorize the formulas.

Conditional probability

Conditional probability is the probability of A given that B has happened, written P(A | B). It is where most probability confusion lives. The classic trap: P(A | B) is not the same as P(B | A), and treating them as interchangeable is the source of the "base rate fallacy," which is responsible for an absurd number of bad conclusions in medicine, law, and business.

The canonical example: a disease affects 1 in 1,000 people. A test is 99% accurate (correctly positive for 99% of sick people and correctly negative for 99% of healthy people). You test positive. What is the probability you have the disease? Most people guess 99%. The correct answer is about 9%. The math is straightforward once you write it out, but the intuition is the opposite of what people expect, because the base rate — how rare the disease is — dominates.

This brings us naturally to Bayes' theorem.

Bayes' theorem in plain language

Bayes' theorem is:

P(A | B) = P(B | A) × P(A) / P(B)

In plain English: the probability of a hypothesis given the evidence equals the probability of the evidence given the hypothesis, times the prior probability of the hypothesis, divided by the overall probability of the evidence. It is a way of updating beliefs in light of new data, and once you see it applied a few times, it becomes hard to unsee.

Back to the disease example. Let A = "has disease," B = "tests positive." P(B | A) = 0.99 (the test is accurate for sick people). P(A) = 0.001 (the prior — disease is rare). P(B) = P(B | A)×P(A) + P(B | not A)×P(not A) = 0.99×0.001 + 0.01×0.999 ≈ 0.011. Plug in: P(A | B) = 0.99 × 0.001 / 0.011 ≈ 0.09. About 9%. The rare disease is still rare even after a positive test, because most positive tests are false positives from the enormous healthy population.

Bayes is not just a trick for exam questions. Every time you update a belief in response to evidence — in medicine, law, machine learning, A/B testing, or hiring — you are doing Bayesian reasoning whether you know it or not. The discipline is making the priors and likelihoods explicit.

Three distributions to know

Binomial

The binomial distribution describes the number of successes in n independent trials, each with probability p of success. "How many of 100 emails will get opened if the open rate is 25%?" is a binomial question. The expected value is np, the variance is np(1-p). Probability Calculator handles binomial probabilities directly.

Normal

The normal (Gaussian) distribution shows up everywhere because of the central limit theorem: sums and averages of many independent random variables converge to a normal shape regardless of what the original variables looked like. Most statistical tests you will ever run assume, somewhere, that some quantity is approximately normal. Z-Score Calculator handles the normal-distribution lookups that used to require a table in the back of a textbook.

Poisson

The Poisson distribution describes the count of events happening at a constant average rate — arrivals per hour, defects per thousand units, typos per page. If the average is λ, the probability of exactly k events is (λ^k × e^(-λ)) / k!. Useful whenever you have "events that just happen" with a roughly stable rate.

Sample size, confidence, and planning

The most actionable application of probability for day-to-day analytics is sample size planning. Before you run a survey, a test, or an experiment, you should know roughly how many observations you will need to detect the effect you care about. Too small and your test will fail to detect a real effect; too large and you are wasting data collection budget.

The key inputs are the effect size you want to be able to detect, the statistical power you want (typically 0.8, meaning 80% chance of catching a real effect), and your alpha level (typically 0.05). Sample Size Calculator takes these inputs and tells you how many observations you need. For a typical "detect a 5 percentage point change in conversion rate" question with 80% power and 5% alpha, you will usually land somewhere between 1,000 and 1,600 observations per group.

For percentage and proportion arithmetic that supports these calculations — converting "15% of 2,000" or "what percentage is 30 out of 200" — Percentage Calculator handles the common conversions, and Statistics Calculator computes the descriptive statistics that go into variance-based sample-size formulas.

Adjacent tools worth bookmarking

Related calculators in the same track: Standard Deviation Calculator for measuring dispersion in sampled data, Confidence Interval Calculator for reporting uncertainty around estimates, and Linear Regression Calculator when a probability question turns into a modeling question.

Related pillar guide

This cluster post is part of the comprehensive tools track. For the broader foundation on choosing and using free online tools, see The Complete Guide to Free Online Tools.

FAQ

When do I use permutations vs combinations?

Use permutations when order matters (arranging people in a line, finishing positions in a race). Use combinations when order does not matter (picking a committee, choosing which 3 of 10 products to review). The same set of people gives different counts depending on whether arrangements are distinguishable.

What is the law of large numbers?

As the number of trials grows, the observed average approaches the true underlying mean. It is why casinos always win in the long run and why A/B tests stabilize as you collect more data. It does not mean a streak is "due to reverse" — that is the gambler's fallacy, and it is wrong.

Are Bayesian methods better than frequentist?

Better for what? Bayesian methods give you a probability of a hypothesis given data, which is what people actually want. Frequentist methods give you a probability of data given a hypothesis, which is more conservative and doesn't require specifying a prior. Use whichever produces a clearer answer for your decision.

How do I handle dependent events?

Use the product rule with conditional probability: P(A and B) = P(A) × P(B | A). If A affects the probability of B, you cannot treat them as independent. Drawing cards without replacement is the classic example — the second draw depends on what you pulled first.

What is the birthday problem and why does it matter?

The probability that at least two people in a room of 23 share a birthday is about 50%. It feels wrong because most people compare themselves to others rather than counting all the pair comparisons (253 of them in 23 people). It is a reminder that intuitions about probability are often wrong and the math is worth doing.

Closing thought

Probability is the mathematics of honest uncertainty. The tools in this post are the ones that pay for themselves the first time a decision actually hinges on whether the data is surprising or expected. Start with the counting rules and Bayes, apply them to problems you already care about, and the rest of the field becomes a natural extension of habits you already have.