Joint Probability Calculator

Two events, one deceptively simple question: how likely are they to happen together, on the same trial, in the same world, at the same time? The answer — the joint probability P(A and B) — depends entirely on how the events are related, and this calculator honours all three relationships: independent events multiply, dependent events multiply through a conditional, and mutually exclusive events refuse to co-occur at all. Enter the probabilities, choose the relationship honestly, and get the joint probability with the union P(A or B) computed and printed beside it.

Three Relationships, Three Multiplications

The joint probability is not one formula but a small decision tree, and the calculator’s relationship selector is the tree’s first branch. Independent events (two coins, a die and a coin, today’s rain and a roulette spin) satisfy P(A and B) = P(A) × P(B) — their probabilities simply compound. Dependent events (drawing two cards without replacement, rain tomorrow given rain today) need the conditional: P(A and B) = P(A|B) × P(B) — the probability of the second world, priced inside the world where the first already happened. Mutually exclusive events (a single die showing 3, or showing 5) have no shared outcomes, and their joint probability is structurally zero. Most errors in applied probability are relationship errors — multiplying as if independent what is merely both-unlikely, or adding what should multiply — which is why this calculator makes the relationship an explicit, chosen input rather than a hidden assumption buried silently inside a formula the user never sees.

Worked Examples: The Three Branches Priced

The defaults run the independent branch: P(A) = 0.5, P(B) = 0.4 gives P(A and B) = 0.20 and P(A or B) = 0.5 + 0.4 − 0.20 = 0.70 — the union subtracting the joint so the overlap is not counted twice. Switch to dependent and enter P(B) = 0.4 in the first box with P(A|B) = 0.7 in the second: the joint is 0.7 × 0.4 = 0.28 — dependence raised the conjunction, as positive correlation does; had A|B been 0.3, the joint would fall to 0.12, dependence as discouragement. The exclusive branch with the same marginals returns joint 0 and union 0.90: the events partition their world almost completely while never sharing a crumb of it. Three relationships, three utterly different conjunctions from identical marginal probabilities — the page’s thesis, computed.

The Union’s Discount and the Venn Picture

Every union formula in probability is the addition rule wearing the relationship’s discount: P(A or B) = P(A) + P(B) − P(A and B). The subtraction is the Venn diagram’s overlapping lens, counted twice by the naive sum and restored once by the joint — which is why this calculator never prints a union without the joint that priced it. The picture repays memorising because it scales: three events add three singles, subtract three pairwise joints, and add back the triple (the inclusion–exclusion the sibling 3-events calculator runs), and the pattern continues as far as events do. Hold the picture and the formulas become bookkeeping; lose it and each is a separate spell to misremember under examination conditions.

Conditional Probability: The Dependent Branch’s Engine

The dependent case deserves its own lens because conditionals are where intuition goes to be mugged. P(A|B) — A priced inside the world where B is known — can differ wildly from P(A): a positive medical test given disease (sensitivity) versus disease given a positive test (the posterior everyone actually wants) are different conditionals pointing opposite directions, related only through Bayes’ theorem and the base rate. This calculator asks you to supply the conditional directly (the second box, in dependent mode), because no tool can invent it from the marginals: P(A) and P(B) alone do not determine the joint for dependent events — the conditional is genuinely extra information about how the events lean on each other, and any source that prices a dependent joint without one has smuggled in independence under a false moustache.

The Chain Rule: Joints of Many Events

The dependent branch generalises into the chain rule, probability’s master key: P(A∩B∩C) = P(A) × P(B|A) × P(C|A∩B) — each factor priced in the world the previous factors built. Weather supplies the friendly example: P(rain Saturday) = 0.3; given Saturday rain, P(rain Sunday) = 0.6; given both, P(rain Monday) = 0.5. The three-day joint is 0.3 × 0.6 × 0.5 = 0.09 — against the 0.027 that blind independence (0.3³, had each day stood at 0.3) would quote, a threefold understatement of the wet weekend, delivered by weather’s famous memory. Every Bayesian network and Markov model is this rule with bookkeeping; the two-event dependent mode on this page is its first link, and the discipline — never multiply marginals where conditionals live — is the whole of the law.

Reading a Joint Table

Applied work usually meets joint probabilities as a 2×2 table, and this page’s quantities are its margins and cells. Take the dependent example as a population: of all days, A∩B = 0.28 (both events), A∩¬B = 0.22, ¬A∩B = 0.12, ¬A∩¬B = 0.38. The cells sum to 1; the row and column sums are the marginals (A totals 0.50, B totals 0.40); and the conditional is a cell divided by its margin — P(A|B) = 0.28/0.40 = 0.70, recovering the input. Dependence is visible at a glance: were the events independent, the both-cell would hold 0.50 × 0.40 = 0.20, and its actual 0.28 is the signature of their leaning. Any joint distribution you are handed — survey cross-tabs, diagnostic tables, risk matrices — submits to this same reading, and the calculator’s three modes are the three stories such a table can tell about its cells.

A Worked Sanity Ledger

Before trusting any hand-computed joint, run the three ledger checks this page’s structure enforces: the joint cannot exceed the smaller marginal (0.28 ≤ 0.40 ✓ in the dependent example); the union must lie between the larger marginal and the sum of the marginals (0.70 sits between 0.50 and 0.90 ✓); and all four Venn regions must be non-negative and sum to one. Thirty seconds of ledger work catches nearly every slip the topic offers, and it is the same conscience the calculator applies internally before printing.

Related Tools

More from the Statistics Calculators hub: the Probability of 3 Events Calculator — this page’s logic, one event further — and the AND Probability Calculator for chains of independent events.

Frequently Asked Questions

What is the joint probability of two independent events?

The product: P(A and B) = P(A) × P(B) — 0.5 and 0.4 give 0.20 in the worked example. Independence is a strong claim (one event’s occurrence carries no information about the other), and the product is licensed only by it; the calculator makes you declare independence explicitly, which is the habit that separates using the formula from being used by it.

Can a joint probability exceed either event’s own probability?

Never — the conjunction is a subset of each event, so P(A and B) ≤ min(P(A), P(B)) always, with equality only when one event implies the other entirely. It is the cheapest sanity check in this corner of statistics: any computed joint larger than its smallest parent exposes an arithmetic or relationship error instantly, and the calculator’s outputs obey it by construction on every run it performs.

How do I find P(A and B) if I know P(A), P(B) and P(A or B)?

Rearrange the addition rule: P(A and B) = P(A) + P(B) − P(A or B). The union and the joint are two views of the same overlap, so any three of the four quantities determine the fourth. (0.5 + 0.4 − 0.70 = 0.20 recovers the worked example’s joint from its union.) Examiners love this rearrangement because it tests the Venn picture rather than the multiplication table.

Are mutually exclusive events independent?

The classic trap: no — quite the opposite. Exclusive events are maximally informative about each other (if A happened, B certainly didn’t), which is the essence of dependence. The only way an event pair can be both exclusive and independent is for one of them to have probability zero. If your model calls two events “exclusive and independent” with positive probabilities, one of the words is wrong, and the joint probability it computes will be wrong in both dialects (0 by exclusivity, the product by independence).

What changes with more than two events?

The pattern generalises at a price: joints multiply through chains of conditionals (P(A∩B∩C) = P(A) × P(B|A) × P(C|A∩B)), and unions blossom into full inclusion–exclusion (the sibling 3-events calculator runs the independent case completely). Independence, when it genuinely holds, remains the great simplifier — every joint a plain product — which is why experimental design works so hard to manufacture it (randomisation, replacement, separation) rather than merely assume it.

Joint probability versus conditional probability: which do I need?

Ask what is being counted. The joint counts worlds where both happen, as a share of all worlds (both rains: 0.28 of all days). The conditional counts A-worlds as a share of B-worlds only (rain tomorrow among days following rain: 0.7). Reports, risk registers and medical leaflets confuse them at scale; the calculator’s dependent mode keeps them in separate boxes precisely so the confusion has to be chosen, not inherited.