Conditional Expected Value Calculator
Knowing a deal closed, a claim was filed, or a market regime held changes what you should expect on average, and regular expected value cannot use that information. This calculator finds E[X|C], the expected value of an outcome given that a condition occurred, starting from the numbers people actually have: each outcome’s joint probability with the condition.
Enter up to six outcomes with P(outcome AND condition) for each, plus the probability of the condition itself. The tool divides through by P(condition), shows every outcome’s conditional probability, and warns you if your joint probabilities add up to more than the condition that contains them.
Each outcome’s probability is P(outcome AND condition), a number between 0 and 1 that cannot exceed P(condition). If your probabilities already sum to 1, they are conditional probabilities: set P(condition) to 1 and the tool returns their plain weighted average.
Quick Answer
Key takeaways
- The per-outcome inputs are joint probabilities, P(outcome AND condition), not conditional ones. Dividing each by P(condition) is what turns them into the conditional distribution.
- The numerator, the weighted sum before dividing, is not the answer. It is the condition’s contribution to the overall expected value, and confusing the two is the most common error in this corner of probability.
- Joint probabilities must sum to at most P(condition). The tool checks this and warns instead of printing a confident number built on impossible inputs.
- If your probabilities already sum to 1, they are conditional probabilities. Set P(condition) to 1 and the same tool returns the plain weighted average.
- Quick Answer
- How to use the conditional expected value calculator
- The formula: a weighted average, rescaled
- Worked example: expected payoff given the deal closes
- The consistency check, and the P(C) = 1 shortcut
- Three conditional tools, three different questions
- What competing calculators miss
- Frequently Asked Questions
- Add This Calculator to Your Website
- References
How to use the conditional expected value calculator
The one decision that matters is which kind of probability you have. This tool expects joint probabilities: how likely each outcome is together with the condition, counted against all cases, not just the conditioned ones.
Add one row per outcome. Enter the outcome value and P(value AND condition). Two rows are ready; add up to six, or remove rows you do not need.
Enter P(condition). This is how often the condition occurs overall. In the worked example the condition holds in half of all cases, so P(C) = 0.5.
Calculate and read all three outputs. E[X|C] is the headline. The per-outcome conditional probabilities show the renormalized distribution, and the unconditional contribution shows the pre-division sum for comparison.
A quick self-test before you trust any inputs: add the outcome probabilities. The total cannot exceed P(condition), because the outcomes are pieces of the condition. If your total comes to exactly 1, you almost certainly hold conditional probabilities already, and the fix is to set P(condition) to 1 rather than to rework every row. The standard textbook treatment of conditional expectation, such as the LibreTexts chapter on conditional expectation and regression, builds every example exactly this way: multiply each value by its joint probability, sum, then divide by the probability of the condition.
The formula: a weighted average, rescaled
E[X|C] is the ordinary expected value computed on the conditional distribution. Building that distribution takes one division per outcome, and the calculator shows each one.
Why divide at all? Because the joint probabilities only account for the fraction of cases where C happened. In the worked example they sum to 0.5, not 1, so as weights they describe half a distribution. Dividing by P(C) = 0.5 stretches them back to a full distribution (0.6 and 0.4) that sums to 1, and the weighted average over that distribution is the honest answer to “given C, what do we expect?” Skip the division and you have answered a different question: how much this condition contributes to the overall expectation, which is the law of total expectation bookkeeping, E[X] = E[X|C] × P(C) + E[X|not C] × P(not C). Both numbers are useful. Only one of them is the conditional expected value.
Worked example: expected payoff given the deal closes
An investment pays 100 if a deal closes on good terms and 50 if it closes on weak terms. The joint probabilities of each payoff together with closing are 0.3 and 0.2, and deals close half the time. The calculator is pre-filled with this case.
| Outcome x | P(x AND C) | P(x | C) = p / 0.5 | x × P(x | C) |
|---|---|---|---|
| 100 | 0.3 | 0.6 | 60 |
| 50 | 0.2 | 0.4 | 20 |
| Total | 0.5 = P(C) | 1.0 | E[X|C] = 80 |
Read the answer the way a decision-maker would: across all deals, this opportunity contributes 40 to expected payoff. Across the deals that close, the expected payoff is 80. Both statements are true, and quoting the 40 where the 80 belongs understates the opportunity by half. That gap is exactly the base-rate correction that conditioning performs, the same correction the conditional probability calculator applies to single events with Bayes’ theorem.
The consistency check, and the P(C) = 1 shortcut
Joint probabilities obey a hard ceiling: the outcomes inside a condition cannot be more probable, in total, than the condition itself. Enter outcome probabilities summing to 0.7 against P(C) = 0.5 and something in your inputs is wrong, no matter what any calculator prints. This tool still shows the arithmetic, because you may be testing the boundary on purpose, but it attaches a plain warning naming the two totals so you can find the bad input.
The shortcut matters just as much in practice. Many data sources hand you the conditional distribution directly: claim sizes given that a claim occurred, returns given a bull market, test scores given a pass. Those probabilities sum to 1 by construction. Entering P(C) = 1 makes the division a no-op and the tool returns the weighted average of the distribution you supplied, with the per-outcome lines confirming each weight unchanged. One tool, both input conventions, and the sum of your rows tells you which convention you are holding.
Three conditional tools, three different questions
Conditioning answers three questions that are easy to blur. Each has its own calculator, and using the wrong one returns a correct answer to a question you did not ask.
How likely is the event now?
The conditional probability calculator (linked from the worked example above) computes P(A|B) and its Bayes flip. Use it when the answer you need is a single probability: the chance of disease given a positive test, of rain given a forecast.
What is the average outcome now?
This page. E[X|C] averages the values of a random variable over the conditioned cases. Use it when outcomes carry amounts: payoffs, claim sizes, returns, scores.
How spread out is it now?
The conditional variance calculator computes Var(X|C), the spread around this page’s mean. Use it when the risk matters as much as the average, because two distributions can share an E[X|C] of 80 and behave nothing alike.
The three connect in sequence. Conditional probability builds the renormalized weights, this page averages with them, and conditional variance measures what is left around that average. For events that chain together rather than condition, the compound probability calculator covers the and/or combinations of independent events.
What competing calculators miss
The tools and lessons ranking for conditional expected value were checked before this rewrite, including the textbook treatments.
Miss 1: conditional inputs only. The CalculatorGek conditional expected value tool takes values with their conditional probabilities P(Value | Y) and multiplies straight through. That works if you already hold the conditional distribution, but most real data starts as joint frequencies or joint probabilities, and there is no P(condition) input and no division step. Here the joint path is the main path, and the P(C) = 1 shortcut covers the conditional case in the same tool.
Miss 2: the mislabeled product. A rival formula documented in earlier research on this page multiplies P(C) by E[X|C] and presents the product as the conditional expected value. That product is the unconditional contribution, 40 in the worked example, not 80. This calculator prints both numbers with their correct names side by side, so the distinction is visible instead of buried.
Miss 3: no consistency check. Nothing stops you entering joint probabilities that sum past the condition in most tools, and the impossible inputs produce a plausible-looking answer. The warning here fires whenever the outcome total exceeds P(condition), which catches the single most common data-entry error in this calculation: pasting conditional probabilities into joint slots without setting P(C) to 1.
Frequently Asked Questions
It is the average outcome of a random variable given that a specific condition has occurred: E[X|C] = sum(x × p) / P(C), where each p is the joint probability P(X = x AND C). It answers “now that we know C happened, what should we expect?” rather than averaging over all cases.
P(outcome AND condition): the joint probability that the outcome occurs together with the condition, between 0 and 1 and no larger than P(condition). If the probabilities you have already sum to 1, they are conditional probabilities P(x|C); enter them as they are and set P(condition) to 1.
Because the condition does not always happen. The joint probabilities sum to P(condition), not to 1, so they underweight every outcome. Dividing rescales them into a proper distribution over just the cases where the condition held, and the weighted average over that distribution is the conditional expected value.
Your outcome probabilities sum to more than P(condition), which cannot happen for joint probabilities: the parts of a condition cannot total more than the condition itself. Recheck the inputs, or, if your probabilities sum to 1, set P(condition) to 1. The result shown used your entries exactly as given.
No. Conditioning on an event that never occurs is undefined, because the formula divides by P(condition). The tool blocks zero and asks for a positive value. If your condition is extremely rare but real, enter its small positive probability and expect the conditional probabilities to be sensitive to rounding.
Regular expected value averages over every case. Conditional expected value averages only over the cases where the condition held, after renormalizing their probabilities to sum to 1. The two can differ sharply: a payoff that averages 40 across all deals can average 80 across the deals that close.
It is sum(x × p) before the division: this condition’s contribution to the overall expected value, since E[X] = E[X|C] × P(C) + E[X|not C] × P(not C). Some calculators mislabel this number as the conditional expected value. It is a useful figure, but it is not E[X|C].
The tool holds six outcome rows. For more, group small outcomes into a combined row with their summed probability and a representative value, or extend the same sum by hand. For continuous variables the concept is identical with integrals replacing sums, but this calculator is discrete-only.
Add This Calculator to Your Website
References
- LibreTexts, “Applied Probability: Conditional Expectation and Regression” – the standard construction used here: multiply each value by its joint probability, sum, and divide by the probability of the conditioning event. LibreTexts: Conditional Expectation, Regression. Accessed October 2026.
- University of Puerto Rico, “Conditional Expectation” course notes – worked joint-distribution examples showing conditional distributions built by renormalizing joint probabilities. UPRM: Conditional Expectation Notes. Accessed October 2026.
- Technion, “Probability and Statistics: A Survival Guide” – expectation, variance and the law of total expectation stated formally. Technion: Probability and Statistics Survival Guide. Accessed October 2026.
- LibreTexts, “Statistical Analysis of Data for Engineers, Chapter 2” – conditional probability definitions that the renormalization step depends on. LibreTexts: Histograms, Statistical Measures and Probability. Accessed October 2026.
- StatTrek, “Bayes Theorem” – the single-event version of the same conditioning correction, for readers moving between P(A|B) and E[X|C]. StatTrek: Bayes Theorem. Accessed October 2026.