Eighteen reports analyze privacy-safe aggregate practice behavior across every exam part, all 13 domains, and 330 subjects. Every result shows its sample, method, and limits.
Community accuracy weights every eligible learner equally. The top-performer line shows the highest-accuracy quartile within each part, not the learners with the most attempts.
The top-quartile cutoff is the minimum learner accuracy within that part's reproducible top cohort.
| Part | Eligible learners | Attempts | Community accuracy | Top learners | Top accuracy | Top cutoff | Active questions |
|---|---|---|---|---|---|---|---|
| Part 1 | 501 | 195,735 | 63.4% | 126 | 80.1% | 71.8% | 2,534 |
| Part 2 | 192 | 87,837 | 63.7% | 48 | 81.9% | 74.8% | 2,147 |
| Part 3 | 145 | 60,303 | 70.1% | 37 | 86.9% | 78.7% | 1,486 |
Data through August 13, 2026. The snapshot contains aggregate results only and excludes names, email addresses, notes, question text, session identifiers, and learner-level rows.
Original evidence about difficult subjects, exam pressure, review flags, and common misconceptions.
The lowest learner-weighted accuracy among strongly sampled subjects is Non-Business Bad Debts in Part 1 at 47%, Distributable Net Income vs. Accounting Income in Part 2 at 52%, and Internal Revenue Code and Regulations in Part 3 at 55.6%. Each result uses current questions and learners with at least 25 attempts in that part.
Read the reportBusiness tax preparation in Part 2 carries the highest weighted difficulty in the current data: 37 of 85 scored questions multiplied by a 35.2% learner-weighted miss rate. The index ranks study priority; it does not predict an official score.
Read the reportAmong matched learners, answered practice-exam accuracy trails test-bank accuracy by 5 points in Part 1, 6.2 in Part 2, and 5.2 in Part 3. Unanswered slots lower the recorded exam score further.
Read the reportFlagged answered items are less accurate in every part: 51.6% versus 64% in Part 1, 41.6% versus 62.2% in Part 2, and 58.3% versus 65.4% in Part 3. A flag identifies uncertainty; it does not prove that changing an answer helps.
Read the reportUnderstanding the Unified Credit has the highest qualifying wrong-answer concentration: one incorrect choice captures 69.1% of 110 wrong responses across 3 well-used questions. The report identifies where errors converge without publishing question or answer wording.
Read the reportObserved practice behavior, repeat work, simulations, pacing, study time, and notes.
Across repeated current-question pairs, latest recorded accuracy rises from 34.7% to 88.3% in Part 1, 37.4% to 89.5% in Part 2, and 44.2% to 91.5% in Part 3. The adaptive system deliberately resurfaces misses, so this is recovery, not causal proof.
Read the reportOn the second recorded attempt, accuracy is 63.7% for same-day repeats, 75.7% after 1-6 days, 72.2% after 7-29 days, and 71.2% after 30+ days. Starting accuracy differs sharply across groups, so the chart is not an experiment about the best interval.
Read the reportPart 1 half-exam learners are the only progression cohort above 50 learners: 56 candidates average 46.1% on their first comparable exam and 54.8% on their latest, an 8.7-point gain. Smaller cohorts are shown with their sample sizes.
Read the reportFull practice exams have higher blank rates than half exams in all three parts. Part 1 rises from 7.3% blank on half exams to 13.9% on full exams; Part 2 rises from 8.5% to 20.6%; Part 3 rises from 8% to 12%.
Read the reportBlank answers rise near the end of full exams. In questions 91-100, the blank rate is 17.8% for Part 1, 23.8% for Part 2, and 12.1% for Part 3. The data shows position, not the reason for each blank.
Read the reportAfter excluding zero, negative, and over-four-hour events, the recorded duration totals 5,591.5 hours in the test bank, 698.6 hours in practice exams, and 215.9 hours in Power Notes. These are valid browser-event durations, not independently verified active study time.
Read the reportLearners with recorded note or Power Notes activity complete far more practice: 772.5 versus 89.3 average Part 1 attempts, 912.6 versus 139.3 in Part 2, and 704.5 versus 206.3 in Part 3. Their raw accuracy is slightly lower, consistent with heavier work on weak areas.
Read the reportCoverage, item analysis, editorial review, and transparent performance benchmarks.
The top quartile exceeds the full eligible cohort in all 13 domains. The largest domain separation is Specialized Returns and Taxpayers in Part 2, where 62.7% community accuracy compares with 83% for the top cohort.
Read the report638 current questions have at least 30 active learners and 100 attempts. Of those, 36 fall below 40% accuracy, 222 fall between 40% and 54.9%, and 15 reach 85% or higher. Difficulty is a review signal, not a quality verdict.
Read the reportThe active bank contains 6,167 question IDs: 2,534 in Part 1, 2,147 in Part 2, and 1,486 in Part 3. It covers all 13 current domains and 330 subjects.
Read the reportEditorial confidence is not calibrated to observed question difficulty in the current bank. Well-sampled questions marked 100 confidence average 57.6% observed accuracy, compared with 58.2% below 50 confidence and 61% in the 50-74.9 band. Most active questions are marked 100.
Read the reportAnnual candidate trends and careful validation of Passboard readiness signals.
The stable historical pattern is stronger than a single headline trend. In 2023-2025, learner-weighted Part 1 accuracy remains near 63%, while Part 2 rises to 67.4% and Part 3 to 71.5% in smaller 2025 cohorts. The 2026 rows are year-to-date and contain only 8-15 eligible learners per part.
Read the reportCurrent Passboard bands align with higher current practice accuracy and higher historical practice-exam scores, but the data is not time-ordered. Learners in the 70-79.9 band average 73.6% current practice accuracy and 57.3% across historical completed exams; the database does not contain verified real SEE outcomes.
Read the report