Basic rules
A probability is a number from 0 (impossible) to 1 (certain). The probabilities of all mutually exclusive and exhaustive outcomes add to 1.
| Rule | Formula |
|---|---|
| Multiplication (joint probability) | P(AB) = P(A | B) × P(B) |
| Addition | P(A or B) = P(A) + P(B) − P(AB) |
| Independent events | P(A | B) = P(A), so P(AB) = P(A) × P(B) |
| Mutually exclusive events | P(AB) = 0, so P(A or B) = P(A) + P(B) |
Odds express the same idea differently: odds for an event = P(E) / [1 − P(E)]. A probability of 0.25 is odds of 1 to 3.
Total probability rule
When the world can be in one of several scenarios, an unconditional probability is the weighted sum of the conditional ones:
P(A) = P(A | S1) × P(S1) + P(A | S2) × P(S2) + …
Example: a 60% chance of growth and 40% chance of recession. A company beats earnings forecasts with probability 70% in growth and 30% in recession. P(beat) = 0.7 × 0.6 + 0.3 × 0.4 = 0.54.
Expected value and variance
The expected value is the probability-weighted average of the outcomes:
E(X) = Σ P(xi) × xi
Var(X) = Σ P(xi) × [xi − E(X)]²
Example: growth (60%) gives a 12% return, no growth (40%) gives −4%.
- E(R) = 0.6 × 12 + 0.4 × (−4) = 5.6%.
- Var = 0.6 × (12 − 5.6)² + 0.4 × (−4 − 5.6)² = 0.6 × 40.96 + 0.4 × 92.16 = 61.44, so σ = 7.84%.
Conditional expected values are the same idea within one scenario. The total expected value equals the probability-weighted sum of the conditional expected values.
Probability trees
A tree lays out scenarios step by step. Each branch carries a probability; multiply along a path for the joint probability and add the paths that lead to the outcome you want. Trees make total probability and expected value problems easy to set up and check.
Bayes’ formula
Bayes’ formula updates a prior belief with new evidence:
P(A | B) = P(B | A) × P(A) / P(B)
Example. 10% of firms in a sector cut their dividend. A model flags 90% of the firms that will cut and, wrongly, 20% of the firms that will not.
- P(flag) = 0.9 × 0.1 + 0.2 × 0.9 = 0.09 + 0.18 = 0.27.
- P(cut | flag) = 0.09 / 0.27 = 33%.
Even with a model that catches 90% of cuts, a flagged firm has only a one-in-three chance of cutting, because cuts are rare and false alarms are common. This “base rate” effect is what the exam likes to test.
Counting rules
Sometimes you need the number of possible outcomes:
- Factorial: n! ways to arrange n items. 4! = 24.
- Combinations (order does not matter): nCr = n! / [(n − r)! r!]. Choosing 3 stocks from 10: 120 ways.
- Permutations (order matters): nPr = n! / (n − r)!. Ranking 3 stocks from 10: 720 ways.
- Multinomial: n! / (n1! × n2! × …) to split n items into labelled groups.
Exam focus
Expect an expected value and standard deviation from a small scenario table, a total-probability calculation, and a Bayes question with a test or a model that flags something. Draw a quick tree: it prevents most mistakes.