Unlike P, FM, FAM and SRM, ASTAM is mostly writing. You don't pick a letter from A through E. You derive the answer, justify your method, show every intermediate calculation and write up the result, and it's graded against a published rubric. The math has to be right, but presentation drives the score.
The only public training set for that format is the 20 constructed-response sample items the SOA published in its November 2022 Sample Questions and Solutions PDF. FreeFellow has all 20 free in a simulator. You read the stem and type your answer to each sub-part, and when you submit, the SOA reference solution appears next to it with the rubric point weights showing.
Start practicing the 20 ASTAM sample items
The 20 items by topic
| Topic area | Approx. items featuring | What you write |
|---|---|---|
| Frequency and severity modeling | ~6 items | Method-of-moments and MLE fits, goodness-of-fit tests (chi-squared, Kolmogorov-Smirnov) |
| Aggregate loss models | ~5 items | Compound Poisson manipulation, Panjer recursion, stop-loss premium calculations |
| Credibility | ~4 items | Bühlmann and Bühlmann-Straub setup, credibility-weighted predictions |
| Ruin theory | ~3 items | Adjustment coefficient computation, ruin probability bounds, surplus process analysis |
| Loss reserving | ~4 items | Chain ladder, Bornhuetter-Ferguson, Bühlmann-Straub triangle reserving |
| Reinsurance | ~3 items | Quota share, excess-of-loss, stop-loss premium and expected recovery |
| Risk measures | ~3 items | VaR, TVaR, comparing risk measure properties (coherence, subadditivity) |
Items often combine two or three areas. One might open with a severity fit, move to an aggregate loss calculation, and finish with a reinsurance premium that uses the aggregate distribution from the previous sub-part.
The math in ASTAM is no harder than the FAM-S half of FAM, and most candidates who fail lose their points on presentation. Someone who can solve every problem but doesn't write the answer the way the rubric rewards routinely scores 50 to 60 percent, well below the 70 percent threshold most expect for a passing score.
Writing to the rubric
Practice the writing as deliberately as the math, which means two passes on each item. Solve the math first. Then rewrite your solution to follow the structure of the SOA reference solution, since that's the structure rubrics reward. State your assumption, justify your model choice in one sentence, derive the formula or apply the theorem with named steps, show the intermediate values and interpret the result.
The most useful self-grading you can do is to map your answer to each rubric point after you submit. Each point in the published breakdown is worth 1, 2 or 3 points, and each one should match a specific sentence in your response. If a rubric point has no matching sentence, you didn't earn it, however right your final number was.
Spend your time in proportion to the points. Sub-parts carry explicit point weights, so a 5-point sub-part gets roughly 5x the time of a 1-point one. Practicing that on the samples builds the time management the 3-hour exam needs. Most candidates spend too long on the early sub-parts and run out of time on the high-value late ones.
The simulator shows the weights so you can grade yourself. If you want AI grading against the same rubric, Fellow includes five AI-graded attempts a day, and anyone can copy a grading prompt into their own ChatGPT or Claude.
You won't get much from the set on newer Lee-Carter mortality model extensions (a 2024+ syllabus addition that's sparse in the 2022 set), catastrophe-style aggregate loss problems with heavy-tail severity and explicit catastrophe layers, or reserving under explicit inflation indexes. For those, use the FreeFellow original ASTAM bank, which has 971 multiple-choice questions across the syllabus for drilling the underlying computations.
Other versions of the set
| Source | ASTAM Sample Items | Format | Cost |
|---|---|---|---|
| FreeFellow | All 20 in interactive simulator | Type-submit-compare with rubric weights visible | Free |
| Coaching Actuaries ASTAM Manual | All 20 plus commentary plus original CR items | PDF in manual | Check current package price |
| ACTEX ASTAM | Selected items | PDF in study guide | Check current package price |
| SOA only | All 20 | Static PDF on soa.org | Free |
FreeFellow is the only free interactive product for ASTAM constructed-response practice. No commercial provider has released a competing set of 20-plus written-answer items with point-weighted rubrics. The closest paid product is the Coaching Actuaries ASTAM Manual, which has the same 20 SOA-published items plus original written-answer items their team writes. That's useful, but the SOA's 20 are the only items whose rubrics reflect exactly how the exam is graded.
If you haven't sat FAM yet, it builds the math ASTAM tests, and Free Exam FAM Sample Questions covers its set. Coaching Actuaries pricing across the SOA exams is in FreeFellow vs Coaching Actuaries (2026).