SOA Exam P is a 3-hour, 30-question computer-based probability exam, and 44.7% of candidates passed it at the May 2026 sitting. The January 2026 rate was 49.2%, and both sittings used a pass mark of 71% average correct (SOA). So more than half of candidates fail on a given attempt. For most people it's the first actuarial exam.

It was my first one too. I took it the summer after my junior year, put in about 200 hours over 10 weeks and passed on the first try. The people who pass tend to do the same handful of things.

What's on the exam and how long to study

The SOA syllabus has three sections. General probability (23 to 30% of the exam) covers set functions, independence, combinatorics, conditional probability and Bayes' theorem. Univariate random variables (44 to 50%) covers PDFs, CDFs, expected value, variance, MGFs, the common distributions (binomial, Poisson, normal, exponential, gamma, beta, uniform) and insurance applications like deductibles, policy limits and expected claim costs. Multivariate random variables (23 to 30%) covers joint, marginal and conditional distributions, covariance, correlation and transformations.

Nobody asks you to recite a formula. Every question asks you to set up a problem and combine several ideas under time pressure, so practice setting up integrals and identifying distributions until you do it without thinking.

I'd budget 250 to 300 hours. The SOA doesn't publish a recommended number, and your math background moves it more than anything else:

Scroll horizontally to view all columns.

Background Hours Timeline
Math, stats or physics major 150 to 200 8 to 10 weeks at 2 to 3 hours a day
Engineering, econ or CS 250 to 300 12 to 14 weeks at 2.5 to 3.5 hours a day
Business, liberal arts or little probability 300 to 400 16 to 20 weeks at 2.5 to 3 hours a day

If you can solve a double integral over a joint density without hesitating, you're in the first group, and your hours go to the insurance topics and to speed. A strong undergrad math background carried a lot of my 200 hours, and plenty of people need 350 or more and pass just fine.

I did 2 to 3 hours every morning before class. Studying daily for 12 weeks works better than cramming 8-hour days in the last 3 weeks, because the concepts build on each other. Reading passively for 300 hours also gets worse results than working problems for 200. If you can't solve problems without your notes open, you haven't learned the material yet.

How I'd spend 12 weeks

This assumes 20 to 25 hours a week, about 3.5 to 4 hours a day, 6 days a week.

Scroll horizontally to view all columns.

Weeks Hours Topics
1 to 2 40 to 50 Set theory, counting, axioms of probability, conditional probability, Bayes' theorem, independence
3 to 5 60 to 75 Discrete distributions (binomial, Poisson, geometric, negative binomial), continuous distributions (uniform, exponential, normal, gamma, beta), MGFs, transformations
6 to 8 60 to 75 Joint, marginal and conditional distributions, covariance, correlation, order statistics, convolutions
9 to 10 40 to 50 Deductibles, policy limits, loss models, expected claim costs, mixtures
11 to 12 40 to 50 A full 30-question practice exam every 2 to 3 days, at about 6 minutes per question

In the first two weeks, a good day is 45 minutes reading, 90 minutes working 15 to 20 problems and 45 minutes on the ones you missed, with no pressure on speed yet. Spend real time on conditional probability and Bayes' theorem. During the distribution weeks, write your own reference sheet for each distribution (PDF, CDF, mean, variance, MGF). Multivariate problems are where a lot of candidates plateau, so practice double integrals for joint distributions until they feel routine. The insurance topics are probability problems in an insurance setting, and after 15 to 20 examples the patterns click.

Don't save every practice exam for the end, since a bad score 4 weeks out is the most useful data you can get. Full exams also build the stamina that topic practice doesn't.

Throughout, attempt every problem before you read the solution, and write down your approach even when you're stuck. Review wrong answers right away and sort each one: setup error, distribution mixup or calculation mistake. There's no set number of problems. I'd plan on about 1,000 as a rough target, and keep going until every syllabus topic feels routine and your timed practice exam scores have stopped swinging much from one sitting to the next.

You're close when you're clearing 71% on full timed practice exams, finishing most problems within about 6 minutes and spotting the distribution and technique quickly. If the exam is close and you're not there, pick the 2 or 3 highest-weight topics where you're weakest, switch to pure practice and stop reading new material.

FreeFellow, which I run, has 1,188 free Exam P practice questions by topic and difficulty, each with a full solution, plus a free readiness score. That score is your accuracy on an exam-shaped mix of FreeFellow questions, on a 0 to 10 scale. Each topic counts at its SOA syllabus weight, and each difficulty level at my estimate of the exam's mix, since the SOA doesn't publish one. Exam P's Ready line is 7.5. Clearing it with enough recent practice across the topics means you've met FreeFellow's practice criteria. It doesn't mean you'll pass, since it only sees the questions you've answered here, so read it alongside your timed practice exam scores. Fellow adds analytics by topic and difficulty, focus practice on your weakest learning objectives and timed 30-question, 3-hour mocks.

Formulas, exam day and free resources

You still need these memorized:

  • Bayes' theorem and the law of total probability
  • Var(X)=E[X2]−(E[X])2\mathrm{Var}(X) = E[X^2] - (E[X])^2 and the conditional variance formula
  • Mean, variance and MGF for the normal, exponential, Poisson, binomial, gamma and uniform
  • Cov(X,Y)=E[XY]−E[X]E[Y]\mathrm{Cov}(X,Y) = E[XY] - E[X]E[Y]
  • Transformations by the CDF method and the Jacobian method
  • Expected claim cost with a deductible dd and limit uu, and the loss elimination ratio

On exam day, if a problem passes 8 minutes, mark it and move on. Spending 12 minutes on a hard one early leaves you rushing at the end. If you have to guess, cutting even one wrong choice helps your odds. Watch for "at least one" versus "exactly one," and conditional versus marginal distributions. Draw diagrams for conditional probability and write out the support of every distribution.

The SOA publishes 718 sample questions with solutions. Paul's Online Math Notes is a good calculus and integration refresher, and MIT OpenCourseWare 18.05 has free probability course materials. For practice, start with FreeFellow's free Exam P questions.