About This Methodology
USMLE Step 3 is a two-day examination. Day 1, Foundations of Independent Practice (FIP), has about 232 multiple-choice questions in twelve 30-minute blocks. Day 2, Advanced Clinical Medicine (ACM), has about 180 multiple-choice questions in nine 30-minute blocks plus 13 to 14 Computer-based Case Simulations. The passing score is 200 (raised from 198 effective January 1, 2024), and USMLE has scheduled a review of that standard for November 2026, with any change applying from January 1, 2027. On March 10, 2026, USMLE moved to shorter multiple-choice blocks; the number of CCS case simulations was not reduced.
No single self-assessment captures the full mix of multiple-choice knowledge, case management and CCS timing, so the tool combines several inputs: your Step 2 CK score, UWorld percent correct, UWSA 1 and 2, NBME CCMSA Forms 6, 7 and 8, the Free 137, and an optional CCS self-rating.
Why Step 2 CK Is the Strongest Anchor
The study “Investigating the Relationship Between a Clinical Science Composite Score and USMLE Step 2 Clinical Knowledge and Step 3 Performance” (Morrison, Barone, Baker, Ross and Pak; Medical Science Educator, 2020; PMC8368809) followed students from US and Canadian medical schools who took all five NBME clinical science subject exams for the first time between 2013 and 2017. In the group of 27,118 who also had Step 3 scores, Step 2 CK correlated with Step 3 at r = 0.70 — the highest of any measure in the study (Step 1: r = 0.61; a five-subject-exam composite: r = 0.69).
- Same content domain: both exams emphasize clinical decision-making, diagnostic reasoning and patient management.
- Universal data: every Step 3 examinee has taken Step 2 CK.
- Not the same scale: USMLE says Step 2 CK and Step 3 scores are not directly comparable. In the study, the same examinees averaged 243 on Step 2 CK and 225 on Step 3.
- A correlation, not a validation of this tool: the study measured how the two exams relate. It did not test our predictor.
The Data Behind the Tool
The tool compares your scores with a reference set of about 4,500 records, including 1,216 student-reported results. The student-reported records average a few points above USMLE's reported mean, which is why the formula includes a fixed correction and why we state every adjustment as an assumption.
Calibrated against real score reports. We calibrated the Step 3 estimate on 2,015 Step 3 score reports that students shared publicly (part of 4,700+ reports across all three Steps), and checked it on reports from July to October 2026. NBME CCMSA percent-correct scores now move the estimate. The tested results are on the methodology page.
How the Estimate Is Built, Step by Step
- 1. Formula part: each input you enter is turned into a Step 3 projection (see the cards below), and the projections are averaged using fixed weights that rescale to whichever inputs you gave. A fixed +5-point correction is then added, because the records the weights were tuned on run above USMLE’s reported mean.
- 2. Similar-record part: the tool finds up to 50 records with the closest Step 2 CK, UWorld, Free 137, UWSA and NBME scores and takes a closeness-weighted average of their Step 3 results.
- 3. Blend: about 55% similar-record part, about 33–40% formula part, and 5–12% pulled toward USMLE's mean of 227 — more when few student-reported records are close to your scores. With no similar records, it uses 88% formula and 12% mean.
- 4. Spread correction: blending pulls estimates toward the middle, so the result is stretched slightly (by 5%) away from 231, the middle of the reference records.
- 5. CCS self-rating: an optional fixed shift (details below).
- 6. Estimate range: based on how spread out the similar records' scores are, with a minimum of about ±11 points; it widens if you enter only one score, skip Step 2 CK, or few similar records exist. It is a heuristic range, not a formal 95% confidence interval.
- 7. Pass probability and verdict: the chance of scoring 200 or more (the current passing score, under scheduled review), worked out from the same estimate range you see — a normal curve centred on the predicted score, with a spread equal to the range's half-width divided by 1.96. A wider range therefore gives a lower probability for the same predicted score. It is capped at 99% (and floored at 1%). The verdict comes from that probability: Safe at 95% or more, Borderline from 50% to 94%, Risky below 50%. Candidate group does not change it.
- 8. Percentile: a normal-curve approximation using USMLE's mean of 227 and standard deviation of 15. USMLE's own norm table (below) is the official reference.
How Each Input Is Used
Weights are for the formula part of the model and rescale to whichever inputs you enter. Apart from the published Step 2 CK correlation, every conversion below is our own assumption — there are no official practice-score-to-Step 3 conversion tables.
Step 2 CK Score
45% weightNotes: Each point above or below 245 moves this projection by about 0.7 points around a Step 3 mean of 227 — so a 255 projects to about 234, not 255.
UWorld Step 3
13% weightNotes: Cumulative percent correct on the UWorld Step 3 Qbank. Values below 50% are treated as 50%, because early-prep percentages say little about the final result. Most useful once you have worked through a good share of the Qbank.
UWSA 1 & 2
10% / 22%Notes: Student-reported comparisons suggest UWSA scores tend to run below real Step 3 scores, and UWSA does not include CCS. The +11 offset was tuned on our reference records; it is not an official conversion. UWSA 2, usually taken late in preparation, counts for more than UWSA 1.
NBME CCMSA 6, 7 & 8
5% weightNotes: NBME's Comprehensive Clinical Medicine Self-Assessment reflects the multiple-choice part of Step 3. NBME's sample performance profile describes a 10–800 score scale and says the CCMSA is not intended to predict Step 3 performance, so our mapping is an assumption and carries little weight in the blend. Enter the result as % correct (forms 6, 7 and 8 are accepted); the estimate is then moved by about 0.85 points per percentage point your CCMSA average sits away from 70%, a term fitted to real score reports. It does not capture CCS readiness.
Free 137
5% weightNotes: USMLE's free Step 3 practice test. Best used as a final-week stamina and question-style check rather than a primary score predictor; with relatively few items, a single result is noisy.
What USMLE Says About Step 3 Score Precision
USMLE's Score Interpretation Guidelines (updated 17 August 2026) give the numbers that matter for reading any Step 3 estimate:
- Mean and spread: first-takers from LCME-accredited schools averaged 227 with a standard deviation of 15 in each of 2023, 2024 and 2025.
- Standard error of measurement (SEM) — about 5 points: plus or minus one SEM covers about two-thirds of the scores an examinee would get across repeated testing at the same true level.
- Standard error of difference (SED) — about 7 points: two examinees' scores need to differ by two or more SEDs before the difference is likely to be real.
- Standard error of the estimate (SEE) — about 7 points: your official score report shows your score plus or minus one SEE — the range your score would fall in two-thirds of the time if you tested again without learning or forgetting.
- Not comparable across Steps: a 220 on Step 2 CK is not equivalent to a 220 on Step 3.
In other words, even the real exam can't pin your proficiency down to a single point. A prediction built from practice scores carries far more uncertainty than that, which is why the tool shows a range.
Step 3 Norm Table (Selected Scores)
| Step 3 score | Percentile rank (share scoring lower) |
|---|---|
| 200 | 4 |
| 210 | 12 |
| 220 | 31 |
| 225 | 43 |
| 230 | 56 |
| 235 | 69 |
| 240 | 80 |
| 245 | 88 |
| 250 | 93 |
| 260 | 99 |
Source: USMLE Score Interpretation Guidelines, Table 2 — first-takers from LCME-accredited medical schools testing between January 1, 2023 and December 31, 2025 (N = 64,576). From 2025, Canadian schools are no longer in this group. USMLE updates the norm table every year; use the latest version.
March 2026 Step 3 Format Update
For Step 3 exams from March 10, 2026, USMLE's new test-delivery software brought these changes:
- MCQ blocks restructured — Day 1: 6 blocks of 38–39 items → 12 blocks of 18–20 items. Day 2: 6 blocks of 30 items → 9 blocks of 20 items.
- Same total — the total number of items (about 232 on Day 1 and 180 on Day 2) and the overall length of each test day did not change.
- CCS case count unchanged — USMLE still specifies 13 to 14 case simulations (10 or 20 minutes each). The popular “13 → 9 cases” claim mistakes the new Day 2 MCQ block count for the case count.
- Updated interface — USMLE describes an updated design, improved keyboard navigation and adjustable image contrast across the Step exams.
Almost all student-reported records in our reference set were taken before this change, so the tool does not adjust for exam format and cannot tell whether the new format moves scores. When we tested filtering to new-format records, it made estimates less accurate against real outcomes, so we dropped it.
How Our Step 3 Predictor Compares
| Approach | What It Uses | Main Limitation |
|---|---|---|
| USMLEPredictor (this site) | Step 2 CK plus practice scores, a stated formula and similar records | No published accuracy figure; every estimate comes with a range |
| UWSA score alone | A single Qbank vendor's self-assessment | No CCS; tends to run below real scores in student reports |
| Copying your Step 2 CK score | Step 2 CK only | USMLE says the two scales are not directly comparable |
| NBME CCMSA | NBME's official Step 3 self-assessment | NBME says it is not intended to predict Step 3 performance; no CCS |
We are not affiliated with the NBME, USMLE, FSMB, UWorld, or any Q-bank vendor.
PGY-1 vs Pre-Residency & IMG Considerations
Step 3 examinees include US residents in PGY-1 or PGY-2, international graduates applying for residency, and physicians at other stages. USMLE's 2025 performance data report these first-time Step 3 pass rates: graduates of US MD-granting schools 96% (21,589 tested), US DO-granting schools 97% (only 129 tested, so treat that figure with care), and non-US schools 88% (12,480 tested). From 2025, Canadian schools are grouped with non-US schools.
PGY-1 / PGY-2 Residents
- USMLE recommends — but does not require — completing, or nearly completing, at least one year of accredited US postgraduate training before Step 3.
- Many residents test during PGY-1 or early PGY-2, while Step 2 CK knowledge is still fresh.
- The tool does not adjust for training year; your practice scores carry that information.
IMG / Pre-Residency Examinees
- Graduates of schools outside the US need a valid ECFMG Certificate, plus passing Step 1 and Step 2 CK scores, to be eligible for Step 3.
- Candidate group (IMG, US-DO, US-MD, Canada) is context only in the tool: it does not change the score, the estimate range or the pass probability, because the reference data gave no reliable basis for an adjustment.
- Group pass rates are averages; they don't describe an individual with strong practice scores.
Score Verdict Tiers
Beyond the point estimate, our USMLE Step 3 score predictor classifies each prediction into a verdict tier from its pass probability, the chance of scoring 200 or more (the current passing score, under scheduled review) worked out from the estimate and its range (capped between 1% and 99%):
Safe
Pass probability of 95% or more. The estimate sits well above 200, and at worst the low end of its range is only just under 200 on this model. It is still an estimate, not a guarantee.
Borderline
Pass probability of 50% to 94%. The estimate is more likely than not to be a pass, but the range reaches below the passing score. Reasonable to test if scheduling pressure exists; more preparation builds a safer margin.
Risky
Pass probability below 50%. The estimate is below the 200 passing threshold. Strongly consider deferring the exam and addressing weakest content areas — particularly CCS practice if not yet attempted.
How We Handle Computer-based Case Simulations (CCS)
USMLE does not publish a percentage for CCS. The Bulletin of Information says that your performance on the case simulations affects your Step 3 score and could affect whether you pass, and that the proportional contribution of the CCS score is no greater than the proportional contribution of the time allotted to CCS. Any specific percentage you see quoted is a community estimate, not an official figure. CCS is part of your single Step 3 score — it is not reported as a separate score — and your score report shows whether your CCS performance was lower, the same as, or higher than your overall performance. No widely available self-assessment measures CCS readiness on a comparable scale, which makes it one of the biggest sources of uncertainty in any Step 3 prediction.
Our predictor offers an optional CCS self-rating that shifts the estimate by a small fixed amount: Great +3, Good +2, OK −2 and Struggled −3 points. These are our own assumptions, not measured effects, and are kept small because student-reported CCS showed only a weak link to final scores. We don't use CCS practice percentages to find similar records, because student-reported CCS percentages proved too noisy to help.
If you leave CCS unrated, the estimate is not adjusted, and a caveat reminds you that CCS still counts. The estimate range does not widen for missing CCS, so remember that the range already leaves out anything the tool can't see.
Sources and Further Reading
Primary sources for the facts on this page (checked 25 October 2026):
- Morrison C, Barone M, Baker G, Ross L, Pak S. Investigating the Relationship Between a Clinical Science Composite Score and USMLE Step 2 CK and Step 3 Performance. Med Sci Educ. 2020;30(1):263–269 (PMC8368809) — Step 2 CK–Step 3 correlation r = 0.70 (n = 27,118).
- USMLE Score Interpretation Guidelines (PDF, updated 17 August 2026) — means, standard deviations, the norm table, SEM, SED and SEE.
- USMLE — Performance Data — first-time Step 3 pass rates by group (2025 figures used here).
- USMLE Bulletin of Information — exam structure, Step 3 eligibility and how CCS contributes to the score.
- USMLE — Examination Results and Scoring — the Step 3 minimum passing score of 200 (from January 1, 2024; currently under scheduled review).
- USMLE — Test Delivery Software Updates Coming in 2026 — the March 10, 2026 Step 3 block changes, with case simulations listed as “no change.”
- USMLE — Scheduled Review of the Step 3 Passing Standard — review at the November 2026 meeting; any change effective January 1, 2027.
- NBME — Comprehensive Clinical Medicine Self-Assessment (CCMSA) — NBME's Step 3 self-assessment.
Frequently Asked Questions
How accurate is the Step 3 score predictor?
We calibrated the Step 3 estimate against 2,015 real Step 3 score reports shared by students and checked it on newer reports; our methodology page shows how it performed. Beyond that: the published correlation between Step 2 CK and Step 3 is r = 0.70, which leaves plenty of room for individual differences, and USMLE itself puts the standard error of measurement for Step 3 at about 5 points. Treat the estimate and its range as a rough guide, not a guarantee.
Why is Step 2 CK the strongest Step 3 predictor?
In a study of 27,118 examinees from US and Canadian medical schools (Morrison et al., Medical Science Educator, 2020; PMC8368809), Step 2 CK correlated with Step 3 at r = 0.70, the highest of any measure studied (Step 1: r = 0.61). Both exams test clinical knowledge. That does not mean you will score the same on both: USMLE says Step 2 CK and Step 3 scores are not directly comparable, and in that study the same examinees averaged 243 on Step 2 CK and 225 on Step 3.
Does UWSA underpredict Step 3?
Student-reported comparisons suggest UWSA scores tend to run below real Step 3 scores, and UWSA does not include Computer-based Case Simulations, which count toward the real Step 3 score. There is no official conversion. Our tool adds a fixed +11 points to UWSA inputs — a model assumption tuned on our own reference records, not a measured or official correction.
What changed with the March 2026 Step 3 format update?
From March 10, 2026, the USMLE restructured the multiple-choice blocks: Day 1 went from 6 blocks of 38–39 items to 12 blocks of 18–20 items, and Day 2 from 6 blocks of 30 items to 9 blocks of 20 items. The total number of items and the overall length of the test did not change. The CCS case count did not change either — USMLE still specifies 13 to 14 case simulations, each allotted a maximum of 10 or 20 minutes. A widely repeated claim that cases dropped to 9 confuses the new Day 2 MCQ block count (6 → 9) with the case count; USMLE’s own change table lists case simulations as “no change.” Almost all student-reported records in our reference set were taken before that change, so the tool does not adjust for exam format.
What is the Step 3 passing score?
The Step 3 passing score is 200, raised from 198 effective January 1, 2024. USMLE has scheduled a review of the Step 3 passing standard for its November 2026 meeting; any change would apply to exams taken on or after January 1, 2027. Our predictor reports pass probability relative to the current threshold, with a verdict tier read off that probability: “Safe” at 95% or more, “Borderline” from 50% to 94%, and “Risky” below 50%.
Can I take Step 3 before residency as an IMG?
Yes, if you meet USMLE’s eligibility rules: passing Step 1 and Step 2 CK, an MD or DO degree (for graduates of schools outside the US, the MD equivalent plus a valid ECFMG Certificate), and the other criteria in the USMLE Bulletin of Information. USMLE recommends — but does not require — having completed, or nearly completed, at least one year of accredited US postgraduate training. The tool works the same way whether or not you are in residency; nothing in it adjusts for training year.
How does the predictor handle CCS readiness?
Through an optional self-rating. Great adds 3 points to the estimate, Good adds 2, OK subtracts 2 and Struggled subtracts 3. These small amounts are our own assumptions, not measured effects. If you leave CCS unrated, no adjustment is made and a caveat reminds you that CCS still counts toward your Step 3 score.
Why are estimate ranges wider at score extremes?
USMLE reports a Step 3 mean of 227 with a standard deviation of 15, so very low and very high scores are uncommon: about 4% of first-takers from LCME-accredited schools score below 200, and about 1% score 260 or higher. There are fewer comparable records at the extremes, and blending pulls estimates toward the middle. The tool applies a small spread correction and widens the range when it finds fewer than 10 similar records, but estimates at the extremes remain less reliable.
What if my real Step 3 score differed from the prediction?
An estimate can miss by more than its range. Test-day performance, CCS and changes in your knowledge all matter, and USMLE notes that scores vary even on repeat testing: the standard error of the estimate for Step 3 is about 7 points. Your official score report shows your relative strengths and weaknesses, including how your CCS performance compared with your overall Step 3 performance.
Related reading
- USMLE Step 3 passing score (2026) — the 200 standard in context, pass rates by cohort, and the four-attempt limit.
- CCS cases on Step 3: how many you actually get — the 13–14 case count and the official scoring rules.
- NBME CCMSA to Step 3 conversion — how practice percentages relate to Step 3, and why the conversion is an estimate.
Accuracy for other exams
- Step 2 CK prediction accuracy — how the Step 2 CK estimate is built and how to read it.
- Step 1 pass/fail prediction accuracy — how the pass-probability estimate works next to your NBME report.
Disclaimer
USMLEPredictor.com is an independent educational tool. We are not affiliated with the National Board of Medical Examiners (NBME), the Federation of State Medical Boards (FSMB), the United States Medical Licensing Examination (USMLE), UWorld, or any Q-bank vendor. Score predictions are estimates built from published correlation research, stated assumptions and our reference dataset; individual outcomes vary based on test-day performance, CCS readiness, clinical experience, and other factors. Use this tool as one data point in your preparation strategy alongside official NBME self-assessments and clinical experience tracking.