P(A given B) usually differs from P(B given A) because each probability uses a different comparison group. Both count the outcomes where A and B happen together. But P(A given B) divides by the probability of B, while P(B given A) divides by the probability of A.
Reversing the letters changes the question, even when the same two events appear. A large share of VIP tickets might include backstage access. That does not mean a large share of backstage tickets must be VIP tickets. The size of each group matters. This lesson uses an invented ticket model to show exactly why.
You can explore related ideas in the statistics and probability hub. Here, the main task is learning which group belongs in the denominator. Once that choice is clear, the formulas become easier to read and check.
What Does “Given” Mean In A Probability?
“Given B” means we restrict attention to outcomes where B happens. We then ask what share of that group also meets A. The event after the vertical bar supplies the comparison group. Read P(A | B) as “the probability of A, given B.”
For example, “backstage access given VIP” starts with VIP tickets only. “VIP given backstage access” starts with backstage tickets only. The questions look similar, but they begin with different collections. You should name that collection before doing any arithmetic.
The word “given” describes available information. It does not say that B causes A. It also does not say B happens first. You can learn a ticket’s access level before learning its type, or learn its type first. Neither order changes the meaning of the two mathematical questions.
A practical reading trick is to say “among” instead of “given.” P(VIP | backstage) asks, “Among backstage tickets, what fraction are VIP?” That wording makes the denominator easier to spot. It also keeps you from silently answering the reverse question.
How Can The Same Tickets Give Two Different Answers?
Imagine 100 tickets: 10 VIP and 90 standard. Eight VIP tickets and nine standard tickets include backstage access. These are invented counts, not survey results. Assume each individual ticket has the same chance of being selected from the full collection.
| Ticket type | Backstage | No backstage | Total |
|---|---|---|---|
| VIP | 8 | 2 | 10 |
| Standard | 9 | 81 | 90 |
| Total | 17 | 83 | 100 |
Among the 10 VIP tickets, eight have backstage access. Therefore, P(backstage | VIP) = 8/10 = 0.8, or 80%. The denominator is 10 because knowing “VIP” leaves only those 10 tickets in the comparison group.
Among the 17 backstage tickets, only eight are VIP. Therefore, P(VIP | backstage) = 8/17, which is about 0.4706, or 47.06%. The same eight tickets appear in the numerator. The denominator changes because knowing “backstage” leaves 17 tickets.
Why isn’t the answer 80% both ways? There are more standard tickets to begin with. Even though backstage access is less common within that group, its nine backstage tickets outnumber the eight VIP backstage tickets. Group size can outweigh a higher within-group percentage.
The equal-selection assumption matters. If VIP tickets were more likely to be selected, simple count fractions could give the wrong answer. In a model with unequal chances, add the probability weights of the relevant tickets instead. The conditional formula still works, but raw counts alone may not.
Why Do The Formulas Have Different Denominators?
The formulas divide the shared probability by the probability of the group being considered. Write A ∩ B for the intersection, meaning both A and B. The intersection does not change when you reverse A and B. The condition after the bar does change.
P(A | B) = P(A ∩ B) / P(B), provided P(B) > 0.
P(B | A) = P(A ∩ B) / P(A), provided P(A) > 0.
Stanford’s conditional probability lecture presents the intersection-over-condition formula and illustrates why reversing the condition can change the answer. The key distinction is the group used for comparison.
In the ticket model, let A mean VIP and B mean backstage. Then P(A) = 10/100 = 0.1, P(B) = 17/100 = 0.17, and P(A ∩ B) = 8/100 = 0.08. These are probabilities across the full collection, before restricting the group.
Dividing 0.08 by 0.1 gives 0.8 for backstage given VIP. Dividing 0.08 by 0.17 gives about 0.4706 for VIP given backstage. Both calculations agree with the table. The formula simply turns the full-collection probabilities back into within-group fractions.
If the condition has probability zero, this ordinary definition is undefined. You cannot divide by zero or treat an empty comparison group as having a zero-percent answer. A probability of zero for the condition and a probability of zero for the requested outcome are different issues.
How Does Bayes’ Rule Reverse A Conditional Probability?
Bayes’ rule reverses a conditional probability by including the sizes of the two groups. Knowing P(B | A) alone is generally not enough. You also need P(A) and P(B), or enough information to calculate them. For positive P(A) and P(B), the rule is:
P(A | B) = P(B | A) × P(A) / P(B).
The reason is short: P(B | A) × P(A) gives P(A ∩ B). Dividing that shared probability by P(B) produces P(A | B). The rule reconnects the reverse question with the correct denominator. It does not let you ignore the original group sizes.
In the example, P(VIP) = 0.1 is the prior: the VIP chance before learning the access level. P(backstage | VIP) = 0.8 is the likelihood: the backstage chance within the VIP group. These names describe parts of the calculation, not three separate kinds of evidence.
P(VIP | backstage) is the posterior, the updated VIP chance after learning that the ticket includes backstage access. The calculation is 0.8 × 0.1 / 0.17 = 8/17 ≈ 0.4706. The update raises the VIP chance from 10% to about 47.06%.
If you do not know the overall backstage chance, calculate it from both ticket types. Standard tickets have a backstage chance of 9/90 = 0.1. Since every ticket is either VIP or standard, P(backstage) = 0.8 × 0.1 + 0.1 × 0.9 = 0.17.
This is the law of total probability in this two-group model. Each term supplies one group’s contribution to backstage tickets. Include both terms: using only the VIP contribution would leave out the nine standard backstage tickets and produce the wrong denominator.
What Happens If VIP Tickets Become More Common?
The reversed probability changes when the original group sizes change, even if both within-group backstage rates stay fixed. Suppose a second invented collection has 50 VIP tickets and 50 standard tickets. Keep backstage access at 80% within VIP and 10% within standard.
Now 40 VIP tickets and five standard tickets have backstage access. There are 45 backstage tickets in total. Therefore, P(VIP | backstage) = 40/45 = 8/9, or about 88.89%. Yet P(backstage | VIP) remains 40/50 = 80%.
The prior, or base rate, is the starting share of VIP tickets. Increasing it supplies more VIP tickets that can enter the backstage group. That changes the composition of the group you use for the reverse question. A conditional rate does not describe that starting mix by itself.
This comparison also explains why transferring a reversed percentage between collections can fail. The access rates may match while the ticket types occur in different proportions. Before reusing an answer, check whether both the conditional rates and the base rates fit the new model.
When Are The Two Conditional Probabilities Equal?
When both conditions have positive probability, the reversed probabilities are equal if the marginal probabilities P(A) and P(B) are equal. They are also equal when the intersection has probability zero. Here, “marginal” simply means each event’s probability before conditioning on the other.
If P(A ∩ B) is positive, dividing that same number by two different positive denominators gives different results. Thus equal reversed answers require equal denominators in that case. Equality is possible, but it follows from the model, not from the similar-looking notation.
For a small example, imagine 20 equally likely tickets. Six are VIP, six have backstage access, and three have both features. Then P(backstage | VIP) = 3/6, and P(VIP | backstage) = 3/6. The answers match because the two comparison groups each contain six tickets.
Independence does not automatically make the reverse answers equal. For independent events, P(A | B) = P(A), while P(B | A) = P(B), assuming the conditions have positive probability. Those answers still differ whenever the original event probabilities differ.
If the events cannot happen together and both have positive probability, both conditional answers are zero. For example, VIP and standard ticket types are mutually exclusive in our model. The lesson on mutually exclusive events explains that separate relationship.
How Can You Check A Reversed Probability?
Start by rewriting the question as an “among” sentence. Circle the event after the bar, then make its count or probability your denominator. Next, find the overlap. That sequence checks the meaning before you enter numbers, which helps catch a correctly calculated answer to the wrong question.
Check that the overlap is no larger than either group. In our first collection, eight shared tickets fit within both 10 VIP tickets and 17 backstage tickets. A numerator larger than the denominator signals a counting or labeling error, since a conditional probability must lie between zero and one.
Keep exact fractions until the final display. The value 8/17 is exact; 0.4706 and 47.06% are rounded. Also use decimals consistently in Bayes’ rule. Mixing 80 with 0.1 treats a percent as a whole number and can distort the result by a factor of 100.
Use the probability calculator for relevant numerical checks after identifying your events and comparison group. The key question remains: “Out of which outcomes?” Reversing a conditional probability keeps the overlap but changes that group, so it usually changes the answer.