Normal Distribution

The normal distribution is the classic bell-shaped model. On Exam P, it usually matters less as a standalone object and more as the distribution you standardize, approximate with, or use to model sums and averages.

The SOA syllabus also notes that a normal table is provided, which is a signal that standardization and table lookups are part of the expected workflow rather than optional extras.

Suppose claim severity is modeled as normal with mean 100 and standard deviation 15. The probability a claim is less than 130 is the probability that Z is less than (130 - 100) / 15 = 2.

That means the problem becomes a table lookup for P(Z < 2), which is about 0.977. The key move was not the lookup itself. It was turning the raw value into a z-score first.

Normal questions often appear in one of three ways: direct probability with a given mean and variance, approximation to a discrete model, or linear combination and aggregation logic.

That makes the normal distribution one of the highest-impact Exam P pages because it connects univariate formulas to multivariate and approximation questions.

A common mistake is forgetting that standardization needs the standard deviation, not the variance. Another is using a normal approximation without checking whether the setup is actually reasonable for the underlying count or sum.

Candidates also lose easy points by treating the table as mysterious. The table is just the last step. The real skill is setting up the standardized value correctly.

In statistics and ML, the normal distribution shows up in error models, Gaussian assumptions, approximation arguments, and uncertainty summaries. On Exam P, the same intuition appears in a more elementary form through z-scores and normal approximation.