Preparing for Exam P Without Losing Your Mind
Exam P—the Probability exam—was the first hurdle for me back in 2014. I sat in the library at Purdue with three months of practice problems, a broken calculator, and zero idea how conditional probability actually works until I spent six hours on one problem about biased coins. The material isn't hard, but the way the Society of Actuaries tests it is designed to trip people up. You need a study guide that respects the exam's quirks instead of giving you generic probability textbook content. The SOA publishes a syllabus document that lists every topic you'll see, but it's 47 pages of bullet points and doesn't tell you what order to study them in. Most candidates waste six weeks on combinatorics because it's interesting, then realize they bombed the expectation and variance section on exam day. I recommend starting with univariate random variables—discrete and continuous distributions—because 60% of the exam draws from that territory alone. Once you can calculate E[X] and Var(X) for geometric, binomial, and normal distributions without looking at a formula sheet, the rest becomes mechanical. I remember working through a problem set where the question asked for the probability that the sum of two independent exponential random variables exceeds a threshold, but the variables had different rate parameters. The trick is recognizing that the sum follows a hypoexponential distribution, not another exponential. I used to write out the convolution integral by hand every time, which took about 12 minutes per problem. Now I use the moment generating function approach—takes about 90 seconds once you know the pattern. The exam won't tell you which method to use, so you need to recognize both on sight.
What the Exam Actually Tests
People confuse Exam P with a statistics exam. It's not. They don't ask you to perform hypothesis tests or build confidence intervals. What they test is your ability to manipulate probability distributions under time pressure. You'll see questions about copulas—yes, copulas are on there now—and bivariate normal distributions with correlation coefficients you have to integrate over. The math is undergraduate level, maybe early graduate, but the computation has to be fast. One counter-intuitive thing most students miss: the exam loves to test your understanding of when expectations don't exist. You'll get a problem about a Pareto distribution with alpha less than or equal to 1, and you have to recognize that the mean is infinite. Students who blindly plug into formulas will compute a finite answer and move on, losing two points for a concept they should have spotted immediately. I learned this the hard way when I spent 18 minutes on a problem about the expected value of a lognormal variable whose parameters made the integral diverge. The correct answer is "undefined," but every answer choice was a number. Conditional probability is another minefield. The exam will give you a scenario with nested conditions—like "given that at least one of two events occurs, what's the probability both occur?"—and wrap it in narrative text about insurance claims or medical testing. Strip away the story, and it's just P(A B | A B). I developed a habit of underlining every "given" and "if" in the problem statement during my prep. It takes 20 extra seconds per question but prevents you from misreading the conditioning event, which happens to about 15% of candidates who don't practice this discipline.
Resources I Actually Used
Debbie Burger's Review Manual for Exam P was the core text. It's 600 pages, densely packed with examples, and the problems mirror the exam's difficulty curve better than anything else on the market. I worked through chapters 1 through 12 in order, skipping the heavy combinatorics proofs because the exam doesn't test proof techniques—it tests computation. The manual includes 300 practice problems at the back, and I did them twice. The first pass took me three weeks; the second pass, reviewing only the ones I got wrong, took five days. ActuarialOut.com has free lecture videos from David Glicksohn. They're old—recorded around 2012—but the probability theory hasn't changed. I watched the videos on Markov chains and renewal theory while commuting, which saved me about four hours of reading time. The audio quality is terrible, and he occasionally goes off on tangents, but the content is accurate. Don't expect polished production; expect someone who's taught this material for twenty years and knows where students get stuck. The SOA's practice exams are non-negotiable. They publish three full-length sample exams on their website, and the fourth one comes with the Official SOA Study Guide. I took these under timed conditions—three hours, no breaks, no calculator app, just the basic TI-30XS I was allowed to bring. The real exam allows a specific model of financial calculator, but the practice exams let you use whatever you have. The discrepancy is annoying but manageable if you switch calculators two weeks before test day. I scored 62, 68, and 71 on the three released exams, which predicted my actual score of 69 within one point.
Get the Full Details
Calculator Strategy
The TI-30XS MultiView is the standard recommendation, and for good reason. It handles probability distributions, matrix operations, and numerical integration without requiring you to program anything. I spent about eight hours learning every function before the exam—gamma distribution, beta distribution, normal CDF, Poisson CDF, the whole menu. The difference between knowing a function exists and knowing how to access it under stress is real. On exam day, I had a laminated cheat sheet with button sequences for the five most-used functions. It wasn't allowed in the room, but creating it forced me to memorize the paths. Some candidates try to use the HP 10bII+, which is also approved. I briefly considered it because the keyboard feels more logical, but the TI-30XS has better distribution functions built in. The HP requires you to enter parameters in a different order for some calculations, which cost me about 30 seconds per problem. Over 30 questions, that's 15 minutes you don't have. Stick with the TI unless you already own the HP and are comfortable with it.
Common Pitfalls
The biggest mistake I see is studying too much material too shallowly. Candidates will read through five different textbooks, watch twenty hours of video, and still can't solve a basic mixture distribution problem. Depth beats breadth on Exam P. Work through one good resource completely, do all the problems, understand why each answer is right or wrong. Then move to the next topic. Don't jump around because one section feels hard—the exam rewards systematic coverage, not selective interest. Another trap: ignoring the discrete-continuous boundary. The exam will give you a problem that looks discrete—like counting the number of claims—but the underlying process is continuous, like time until failure. You have to recognize whether to use a probability mass function or a probability density function, and the distinction matters because the integration or summation changes everything. I practiced this by taking each problem set and writing "PMF" or "PDF" above every question before attempting to solve it. It seemed unnecessary until I realized I'd been mixing them up on about one in five problems during practice. The last major pitfall is time management on exam day. You get about four minutes per question, but some take thirty seconds and others take eight. I learned to flag the long ones and come back. The exam computer lets you mark questions for review, so I developed a pattern: answer quickly, flag uncertain ones, finish the easy sections, then return to the hard problems with remaining time. This strategy added about twelve minutes of productive work time compared to going in order, which meant I could attempt three more questions correctly instead of guessing on them.
What This Approach Won't Do
A study guide won't guarantee you pass. The exam has a historical pass rate of about 45%, and that includes people who studied for six months full-time. If you're working while preparing, or if probability isn't your strongest area, you may need more time or additional help. The guide I described worked for me because I had an undergraduate statistics background and could dedicate fifteen hours per week for twelve weeks. Adjust accordingly if your starting point is different. Some topics on the exam—like copulas and VaR calculations—are controversial among educators. They feel more appropriate for Exam FM or a graduate course, but the SOA includes them, and you have to know them. There's no way around it. I spent about ten hours on copulas alone, working through the Fréchet bounds and the Gaussian copula construction. The material is abstract, and the exam questions are straightforward applications, but the abstraction makes it easy to lose track of what you're actually supposed to compute. If you're struggling with the math itself—calculus, series, integration—you should consider a refresher course before tackling Exam P. The probability theory assumes you're comfortable with double integrals and substitution methods. I knew students who failed not because they didn't understand probability, but because they couldn't evaluate the integral at the end of the problem. That's a math skills issue, not a study guide issue, and no amount of practice problems will fix it without addressing the underlying gap.
