BREAK·THE·TEST
Methodology

How a question gets here, and what has to pass before you see it.

A big practice library is only useful if the answer keys are right and the questions behave like the real exam. Both are easy to claim and hard to prove, so this page describes the actual process, including the parts that are limits rather than strengths.

1. Written to a specific target

Nothing gets written as a generic practice question. Each one starts from a named skill on the current exam blueprint (for example, digital SAT "words in context" or ACT Science research summaries), a target difficulty, and one specific way the real test makes a wrong answer look right. That last part is what the question is built to teach.

Because the target comes first, the mix of skills in a full-length practice test tracks the published proportions on the real exam rather than whatever happened to be easiest to write.

2. The answer is checked, not asserted

Every math answer runs through a symbolic solver that evaluates all four options independently, so a keyed answer that does not actually satisfy the problem cannot ship. Questions with a figure, table, or data set get the same treatment against the underlying numbers, which catches charts that disagree with their own question.

3. Solved blind by independent solvers

Machine checks confirm arithmetic. They do not catch the harder failure: a question where the keyed answer is correct arithmetic for the wrong question, or where two options are equally defensible. So before a batch ships, the answer keys are stripped and the questions are solved from scratch, independently, more than once. Disagreement with the key blocks the batch until a human resolves it. Questions that turn out to be genuinely ambiguous are cut, not patched.

We also ran this over the entire existing library in one pass rather than only on new questions, and fixed or retired everything it caught.

4. Reading questions must be answerable from the passage shown

A reading question is rejected if the evidence for its answer is not present in the passage printed on your screen. This sounds obvious. It is the single most common way a plausible-looking reading question is actually broken, and it is checked mechanically rather than trusted.

5. Duplicate and near-duplicate screening

Two checks run on every batch: one for questions a student would recognize as identical, and one for questions that are the same skeleton with the nouns swapped. Volume that comes from re-skinning the same question is not volume, so those are rejected before they reach the library.

6. Reviewed for the current exam, twice, by different reviewers

Exams change. The digital SAT is a different test from the paper SAT, and the ACT changed again in 2026. Questions are reviewed against the format that is actually being administered now, and every question is reviewed by more than one independent reviewer, arranged so that no question is graded by whoever wrote it.

7. Reported problems get fixed and stay fixed

You can flag any question in one tap while you practice. A person reads it. If the question is wrong, we correct or retire it, and the change is versioned, so a fixed question does not quietly come back. If you are right and we were wrong, that is a useful day for us.

Where the scores come from

Full-length digital SAT practice tests are scored using the College Board's published raw-to-scaled conversion, 200 to 800 per section. The tests route the same way the real one does: your first module decides whether the second is the harder or the easier one, and taking the easier one caps how high that section can score. Full-length ACT tests are scored 1 to 36 per section with a composite, on the published scale for the current format.

One number on that page is an estimate and you should know which. The College Board confirms that missing questions in the first module puts the top of the scale out of reach, but it has never published the routing threshold or where the easier path actually caps out. Every figure in circulation, including ours, comes from people observing practice tests, and those figures disagree with each other by about a hundred points. We cap the easier path at 600 per section and route on 19 of 27 in Reading and Writing, 15 of 22 in Math. Those are our best reading of the public evidence, not facts we were given, and we moved them in July 2026 because our earlier numbers were more generous than anyone else's and were flattering the lower path.

The conversion is official. The questions are ours. So treat the number as a well-grounded estimate, not an official score. Real scores move between test forms and test days, and any practice score should be read with roughly a 60-point band on the SAT.

What we do not claim

  • These are not real exam questions. Nothing here is leaked, licensed, or reproduced from a real test, and none of it is meant to be memorized.
  • We are not affiliated with the College Board, ACT, Inc., or the Common Application, and none of them endorse us.
  • We do not publish score-improvement claims. We have no study that would support one, and we would rather say nothing than invent a number.
  • Difficulty labels are our judgment against the published blueprint, not a measured statistic from real test-takers.
  • Coverage is not even. The SAT Reading, Math, and Grammar libraries are the deepest; AP essay and application essay practice are smaller, and we would rather tell you that than pad them.

How the blog is written and checked

The same standard applies to what we publish. Posts are written by the people who build the question bank, not commissioned to a content agency, which is why they can cite our own measurements.

Before a post goes up, every claim about the current exam is checked against a primary source, usually the College Board or ACT directly, because formats change and a confident sentence written last year can be wrong this year. Numbers carry their denominators so you can judge the sample yourself. Research we cite from outside is read rather than summarised second-hand, and we name it so you can go and disagree.

Every post is then reviewed by more than one independent reviewer, arranged so that nothing is checked only by whoever wrote it, and the brief is to hunt for overclaiming rather than to polish. That process has repeatedly caught real errors in our own drafts, including advice that would have been impossible to follow on the real test. When a review is right, the post changes or does not run.

When something we published turns out to be wrong, we correct the post rather than quietly deleting it, and time-sensitive pieces about admissions policy carry the date they were last checked. If you find an error, the support address at the bottom of the page reaches a person.

Current size

19,600+ questions across the digital SAT, the ACT, AP, and college application essays, plus 50+ full-length practice tests. The count on the home page is generated from the library itself at build time, so it cannot drift from what actually ships.

Try a questionCommon questions