About This Methodology
USMLE Step 1 outcomes have been reported as Pass/Fail only for exams taken on or after January 26, 2022 (USMLE policy). Since then, predicting Step 1 readiness has meant estimating a probability of passing rather than a 3-digit numeric score. Any tool that still reports a Step 1 3-digit score is guessing at a number USMLE no longer reports.
Our Step 1 tool maps an NBME Comprehensive Basic Science Self-Assessment (CBSSA) score — the equated percent correct, or EPC — to an estimated probability of passing, using a smooth curve that we set by hand. It is our own approximation of the relationship between CBSSA scores and passing. It is not an NBME table. NBME already gives every CBSSA taker its own estimated probability of passing; that number is the one to trust.
What NBME Actually Publishes
NBME's guidance on gauging readiness for Step 1 (current version dated September 2026) explains how to read the CBSSA and CBSE score report. It describes four things — and no score-to-probability table:
- Your equated percent correct (EPC): an estimate of the share of the content you have mastered, on a 0–100 scale. Scores are statistically equated for small differences in form difficulty, so you can track progress across forms.
- Your likely score range: if you tested again without learning or forgetting anything, your score would fall within about 4 points of the reported EPC two-thirds of the time.
- The Step 1 low-pass range: performance that passes, but only by a small margin. NBME advises extra caution whenever your likely score range overlaps it.
- Your estimated probability of passing (1%–99%): calculated with a statistical model built on a large group of examinees who tested within one week before taking Step 1 for the first time. The CBSE report adds a 4-week probability. NBME updates the model every year, and the date of the latest update is on your report.
NBME also notes that a 90% estimated probability still means about 1 in 10 examinees with comparable performance would not be expected to pass, that the probability is not the same as the percentage of content mastered, and that results are less reliable if you test more than a week before Step 1, use self-paced timing, consult resources, or repeat the same form.
How Our Tool Works, Step by Step
- 1. Most recent form: the tool takes the highest-numbered CBSSA form you enter (Forms 29–34) as your most recent result. NBME 34, the newest form, is counted as up to 5 EPC points lower than entered (never below 30%), because it reads about that much higher than the other forms for the same person.
- 2. Curve value: it reads that EPC off our approximation curve — for example, about 87% at an EPC of 60, about 92% at 64, about 97% at 68 and about 98–99% at 72 or above. The headline number the tool shows is a little lower at the top end (see step 5).
- 3. Readiness tier: it assigns a tier from the same EPC (thresholds in the Readiness Tiers section below). The thresholds are deliberately conservative near the passing line.
- 4. Cross-checks: a low Free 120 or UWorld score can move the tier down; neither can move it up.
- 5. Headline number: the percentage shown is the average of the curve value and a target set by your tier (the middle of the tier, shifting smoothly near the tier edges), so the number and the verdict never contradict each other. The raw curve value for your score is shown in the EPC curve card. In practice that gives about 89% at an EPC of 64, 94% at 68, 95% at 72, 97% at 76 and 98–99% at 78 or above.
- 6. Estimate band: a hand-set band of about ±3 to ±4 percentage points around the headline number. It narrows by one point when you add a second NBME form that is in line with the first; Free 120 and UWorld never narrow it. It is a rough guide, not a statistical confidence interval.
- 7. Trend: with two or more forms, a change of 2 or more EPC points per form is reported as improving or declining; a declining trend adds a warning.
- 8. Your group (optional): choosing US-MD, US-DO or IMG changes only the advice text, never the probability.
Limitations You Should Know About
- Hand-set, not fitted: NBME's model is built on outcome data we do not have. Our curve is an approximation, and its numbers can differ from the probability on your report.
- No test-conditions check: the tool cannot tell whether you took the form timed, in one sitting, without resources, or how far you are from test day — all things NBME says affect how reliable a CBSSA result is.
- One form is noisy: with a likely score range of about ±4 EPC points, a single result can swing your tier. Two consecutive recent forms tell you much more.
- Group pass rates don't describe individuals: cohort figures are context only.
In short: an estimate, not a guarantee. Use it to understand your margin and trend, and treat the probability on your NBME report as the reference.
How Each Practice Score Is Used
The tool is built around NBME CBSSA scores. Free 120 and UWorld act only as cross-checks. The curve is a hand-set approximation rather than a fitted model. NBME publishes no conversion table, so this curve is our own approximation, and the probability printed on your NBME report is the reference. Our calibration uses score reports that students shared publicly (1,742 for Step 1, 294 of them fails, part of 4,700+ reports across all three Steps); they are self-reported, not official outcome data. See how we calibrate.
NBME CBSSA (Forms 29–34)
Primary inputNotes: Your most recent form drives the estimate. NBME's own probability is built for a CBSSA taken within about a week of Step 1, timed and in one sitting — the closest match to how you will test.
Earlier CBSSA Forms
TrendNotes: Earlier forms don't change the probability; they show your trajectory. A rising trend is reassuring; a falling one triggers a warning to look at what changed.
Free 120
Downgrade onlyNotes: USMLE's own free practice test. Best as a final-week check of stamina, interface and question style. Below 60% it caps a Safe or Very Safe verdict at Likely Pass; below 50% it caps a Likely Pass or better verdict at Borderline.
UWorld
Downgrade onlyNotes: UWorld is a learning tool, often used untimed or in tutor mode, so we use it only to flag a sharp gap with your NBME results. Below 55% it caps a Safe or Very Safe verdict at Likely Pass; below 45% it caps a Likely Pass or better verdict at Borderline.
NBME Form-Specific Notes
Forms the Tool Accepts (29–34)
- No form ranking: NBME equates CBSSA scores to account for small differences in form difficulty, so the tool reads an EPC the same way on every accepted form.
- Most recent = highest number: the tool treats the highest-numbered form you enter as your latest result.
- Timing: for the closest match to NBME's own 1-week probability, take your final CBSSA about a week before your test date, timed, in one sitting, without resources.
- Don't repeat a form: NBME says results from a form you have already taken are less reliable for judging readiness.
Not Accepted by the Tool
- NBME 25–28: left out as a scope choice to keep the tool's inputs consistent — not because the forms are miscalibrated. Read the estimated probability on that form's own score report.
- CBSE / COMSAE: the CBSE is NBME's school-administered exam, and its report already includes 1-week and 4-week pass probabilities. COMSAE is a different exam on a different scale. Our curve is set for CBSSA scores, so we don't accept either as an input.
How Our Step 1 Predictor Compares
| Source | How the Probability Is Produced | Data Behind It | Updated |
|---|---|---|---|
| Your NBME CBSSA score report | NBME's own statistical model | A large group of examinees who took Step 1 within a week of testing (NBME) | Every year (date on your report) |
| USMLEPredictor (this site) | Our hand-set approximation curve plus readiness tiers | Our own approximation; NBME publishes no conversion table | October 2026 |
| AMBOSS | Proprietary Qbank model | Not published | Not published |
| Forum spreadsheets | User-submitted scores | Varies | Inconsistent |
| Q-bank built-in predictors | Internal Q-bank performance | Single-vendor | Varies |
We are not affiliated with the NBME, USMLE, FSMB, or UWorld. Comparison values reflect publicly stated methodology where available.
Step 1 for International Medical Graduates (IMGs)
USMLE's 2025 performance data report these first-time Step 1 pass rates: US MD 93% (23,028 tested), US DO 89% (4,623 tested) and non-US schools 75% (22,066 tested). From 2025, USMLE groups students from Canadian schools with non-US schools, so these figures are not directly comparable with earlier years. They are group averages and say nothing about an individual with a given NBME score.
How We Handle IMG Status
- Individual probability is unchanged: the tool gives the same estimate to anyone with the same scores. An IMG examinee and a US-MD examinee with an EPC of 70 on Form 31 see the same number.
- Advice text shifts: for Likely Pass and lower tiers, IMG examinees see a note that 75% of first-time takers from non-US schools passed in 2025, and that an extra preparation cycle is often a defensible choice when the NBME score is borderline.
- Selecting IMG is optional: the group field affects only advice wording. Leaving it blank produces neutral guidance.
Readiness Tier System
A raw probability is easy to misread — a "90% pass" estimate sounds safe even though it implies a real chance of failing. So the tool also places each result in one of five readiness tiers, set from the EPC on your most recent form. The bracketed figures give the raw value of our approximation curve and the headline number the tool shows (the average of the curve value and a conservative target for the tier).
Very Safe
EPC 78 or above (curve ~99%; the tool shows about 98–99%). Focus the final week on retention, sleep and exam-day timing.
Safe
EPC 72–77 (curve ~98–99%; the tool shows about 95–98%). Likely to pass; a fresh CBSSA within a week of the exam gives a sharper read.
Likely Pass
EPC 64–71 (curve ~92–98%; the tool shows about 89–95%). A good signal with real downside risk. One more NBME close to exam day builds margin.
Borderline
EPC 56–63 (curve ~78–91%; the tool shows about 73–89%). Consider extending the dedicated study window, then re-test with a CBSSA in about 2 weeks to confirm the direction of progress.
High Risk
EPC below 56 (curve below ~78%; the tool shows below about 73%). Our advice: consider rescheduling 4–6 weeks out, focus on your weakest content areas, and take a fresh CBSSA in 2 weeks to track progress.
A low Free 120 or UWorld score can move you down one or more tiers, and the headline percentage averages the curve value with a target set by your tier. These tiers are our own guide — NBME's guidance instead compares your likely score range with the Step 1 low-pass range on your report.
Sources and Further Reading
Primary sources for the facts on this page (checked 25 October 2026):
- NBME — CBSSA and CBSE: Gauging Readiness for USMLE Step 1 (PDF, September 2026) — how to read EPC, the likely score range, the low-pass range and NBME's estimated probability. It does not contain a score-to-probability table.
- USMLE — Performance Data — first-time Step 1 pass rates by group (2025 figures used here).
- USMLE — Examination Results and Scoring — Step 1 outcomes reported as Pass/Fail for exams on or after January 26, 2022.
- USMLE Bulletin of Information — official policies on scoring and score reporting.
- IAMSE 2024 conference session — Predictive Value of the NBME CBSE for Step 1 Pass/Fail Outcomes — a conference abstract, not a peer-reviewed paper: a study from Ponce Health Sciences University of 172 students who took Step 1 between April 2022 and October 2023, reporting a 96.4% pass probability at a CBSE score of 53 or above. It concerns the CBSE, not the CBSSA, and was not used to build our curve.
- PubMed — Step 1 prediction research — a search for published research on NBME self-assessments and Step 1.
Frequently Asked Questions
How accurate is the Step 1 pass-probability predictor?
It is an estimate, not a guaranteed forecast. The tool maps your NBME CBSSA score to a pass probability using our own hand-set approximation curve. NBME does not publish a score-to-probability table, so this curve is our own approximation. Our calibration uses score reports that students shared publicly (1,742 for Step 1, 294 of them fails), which are self-reported and not official outcome data. The estimated probability printed on your NBME score report comes from NBME's own model, which NBME updates every year, and it is the authoritative number. Use this tool to read your margin and trend, not as a guarantee.
Why does the predictor only accept NBME forms 29–34?
It is a scope choice that keeps the tool's inputs consistent. NBME equates CBSSA scores to account for small differences in form difficulty, so older forms are not miscalibrated or harsher by design. If you have taken Forms 25–28, read the estimated probability printed on that form's own score report.
Will you provide a 3-digit score for Step 1?
No. USMLE has reported Step 1 outcomes as Pass/Fail only for exams taken on or after January 26, 2022. There is no current 3-digit Step 1 score to predict, so the tool reports an estimated probability of passing instead.
What is considered a "safe" NBME score for Step 1?
There is no official cut-off. On our approximation curve, an equated percent correct (EPC) of about 68% corresponds to roughly a 97% pass probability and 64% to roughly 92%; the tool's headline number is a little lower (about 94% at 68% and 89% at 64%) because it is averaged with a conservative tier target. Below 56% the tool flags high risk. NBME's own guidance is to compare the likely score range on your report (your EPC plus or minus about 4 points) with the Step 1 low-pass range — any overlap warrants extra caution.
How are IMG candidates handled?
The tool gives the same probability to anyone with the same scores; your group never changes the number. USMLE's 2025 performance data show first-time Step 1 pass rates of 93% for US MD students, 89% for US DO students and 75% for students from non-US schools. When you select IMG, the advice text mentions that base rate, but the probability itself is unchanged.
Does UWorld first-pass percentage matter?
It is a useful cross-check, but UWorld is a learning tool rather than an assessment. In this tool a UWorld average can only lower your readiness tier: below 55% it caps a Safe or Very Safe verdict at Likely Pass, and below 45% it caps a Likely Pass or better verdict at Borderline. It can never raise the tier above what your NBME score indicates.
How does the predictor handle multiple NBME scores?
When two or more NBME forms are entered, we use the most recent form (highest form number) as the primary anchor and compute a trend across the others. An improving trend (≥2 EPC points per cycle) is treated as a positive indicator; a declining trend triggers a recommendation to investigate what changed before test day.
What if my real Step 1 outcome differed from the prediction?
An estimated probability is not a guarantee. NBME itself notes that a 90% estimated probability still means about 1 in 10 examinees with comparable performance would not be expected to pass, and many factors — changes in knowledge, or practice conditions that differ from the real exam — affect the outcome. If your result differed from the estimate, the probability and content-area feedback on your NBME report are the best places to start your review.
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), or any Q-bank vendor. Pass probability is an estimate from our own approximation curve, not an NBME figure; individual outcomes vary based on test-day performance, content preparation depth, and other factors. The estimated probability on your NBME score report is the authoritative number. Use this tool as one data point in your preparation strategy alongside official NBME self-assessments.
Accuracy for other exams
- Step 2 CK prediction accuracy — how the Step 2 CK estimate is built and how to read it.
- Step 3 prediction accuracy — how the Step 3 estimate is built, and USMLE's own measurement-error figures.
Related reading
- How to calculate your Step 1 pass probability — a form-by-form walkthrough of NBME 25–33 and the failed-NBME recovery protocol.
- Which NBME form is most accurate? — how each practice assessment relates to the real exam.