A 13-question set of hard, original math questions at the top difficulty band, timed at 21 minutes — the same per-question pace as the real thing. It is a diagnostic, not a full-length practice test: 13 questions can show you which traps catch you and where your pacing breaks, but they cannot measure how you hold up across a whole module.

The 13 questions are a focused diagnostic rather than a complete picture of a student's mathematics or test performance. The result is a set of observed answers and trap patterns, not a standardized score.

Work begins with a concept specification: the skill, the constraint that matters, and the kind of reasoning the question should require. The drafter solves the intended problem, then designs the wrong choices around plausible operations that could fail.

The question is not ready because its wrong choices look tempting. Each wrong choice must be reachable by a described mistake, and the derivation must be checked from that mistake to that exact value. A question that cannot support that chain is revised or discarded.

Trapline names the earliest operation that went wrong, rather than classifying a distractor by how close its number looks. The taxonomy covers setting up the model incorrectly, choosing a relation that does not apply, leaving out a required step, slipping while executing a correct method, failing to check a candidate, and reporting the wrong quantity.

That distinction matters in the results: a named trap is meant to describe a route to a wrong choice, not to label a student's ability or character.

For every wrong choice, the review record keeps a derivation of the associated mistake. Reviewers ask whether the stated operation really produces that value, whether the correct choice remains uniquely defensible, and whether the question's conditions rule out the tempting path.

This page does not publish answer keys, per-question trap mappings, or solver comments. Those are operational content used for grading and review, not public claims needed to explain the method.

Trapline can generate a question from a concept specification, then sends the student-visible question and answer contract through a validation graph. Independent blind-solver and diagnostic-auditor model families answer their assigned projections without inheriting the generator's private context.

A deterministic skeptic node checks the passage, answer uniqueness, distractor patterns, explanation grounding, and agreement with the independent evidence. A merge node records the graph status and content hashes; a blocked or inconclusive node prevents acceptance.

These checks reduce the chance of a wrong key, ambiguity, or misleading diagnosis. They are review evidence, not a promise that a question can never contain an error.

Each shipped question has a provenance record containing the generator identity, concept specification, graph node families, graph hash, and date of final validation. The build gate rejects missing graph roles, correlated answering families, stale evidence, or a generator family reviewing its own output.

The public promise is about recorded verification, not human origination. The page intentionally does not turn internal provenance records into public per-question receipts.

Every question here is checked by an independent validation graph before it ships, and logged with its provenance. No College Board material is copied, paraphrased, or used anywhere in the process — including as input to any AI tool.

Trapline uses PSAT/NMSQT® descriptively to explain the preparation context. It does not reproduce, paraphrase, or present official test material.

A 13-question set cannot measure full-module endurance, establish population norms, forecast a student's score, or establish official difficulty. Trap labels are explanations for designed wrong choices, not a diagnosis of a person and not a substitute for instruction.

Review is evidence that the stated checks were performed. It does not guarantee that results contain no errors, certify an expert, or present this material as official.

No account, no name, no login. A random ID is generated in your browser and stored with your answers, your timings, your device type, and the link you arrived from. It is not linked to your identity — but it is pseudonymous rather than anonymous: if you pay the $10 reservation, Stripe collects your email for delivery and that payment carries the same random ID, which connects the two.

No advertising trackers, no third-party analytics, no data sold or shared. The five events we record (arrived, started, finished, viewed results, clicked reserve) go to our own database and nowhere else.

The result link is a bearer-like link: anyone who has it may be able to view the result. Email the correction channel below if you need a result or stored data removed.

PSAT/NMSQT® and National Merit® Scholarship Program are registered trademarks of the College Board and the National Merit Scholarship Corporation, neither of which sponsors, endorses, or is affiliated with Trapline. Those names are used here only to describe what this practice material prepares you for.

If a question, trap explanation, privacy statement, or other methodology claim looks wrong, email prep.app.admin@gmail.com with the page or question identifier and what you found. That is the correction channel for this project; it is not a promise that every report will result in a change.

Last updated: 2026-08-21.