How Accurate Are NBME Score Predictors?

Students call NBME practice exams "predictors", but NBME is careful about what it claims. This guide sets out what each NBME product reports, using NBME's own pages and guides, and what that means for deciding whether you are ready.

By Dr. Sahith Valluru. For The Boards Professor, founded by Dr. Sean Moazezi, DO. Published October 21, 2026. Updated October 28, 2026.

What each NBME report converts to

Checked October 28, 2026. Product details change; the linked NBME page is the authority.

Product	What the report shows	Conversion NBME provides	Source
CBSSA (Step 1)	Estimated probability of passing Step 1; equated percent correct (EPC) overall and by content area	NBME’s common questions mention a lookup table for USMLE scores for the CBSSA and CCSSA only	NBME: CBSSA
CCSSA (Step 2 CK)	Estimated probability of passing Step 2 CK; EPC overall and by content area	As above; no estimated three-digit score is stated on the product page	NBME: CCSSA
Clinical Science Mastery Series (shelf)	Score on a 50-question form, with answer explanations	A conversion table to an approximate total on the Subject Exam score scale	NBME: Clinical Science Mastery Series

1. Know what each NBME product reports

Sources: NBME's CBSSA and CCSSA pages and its CCSE score interpretation guide. On the self-assessment pages NBME states that the estimated probability is not a promise of how you will do on the real exam.

2. Read the CCSE estimate the way NBME intends

NBME describes the total CCSE score as estimated performance on Step 2 CK if the examinee had taken both exams under the same conditions and with the same knowledge. It also says many factors, such as changing levels of knowledge, may make the real Step 2 CK score higher or lower. The CCSE score carries a standard error of the estimate of 8 points: re-tested without learning or forgetting, you would land within 8 points about two times out of three.

3. Add the real exam's own error

The real exam has error too. USMLE gives Step 2 CK a standard error of the estimate of about 8 points and a standard error of measurement of about 6 (USMLE Score Interpretation Guidelines). Put the two together and a single practice score is best read as a band of plausible results, not a point.

That is why two students with the same practice score can land 10 or 15 points apart on test day without either predictor being "wrong".

4. Use several results, not one

This part is our advice. A single form tells you little; a trend across three or four tells you a lot. Look at whether scores are rising, flat or falling, and whether the same content areas stay weak. A steady trend above the level you need is more reliable than one high score.

5. Treat third-party predictors with care

Question banks also offer self-assessments. UWorld, for example, says its Step 2 practice exams give a three-digit score estimate (UWorld). Those estimates use each company's own methods; they are not NBME or USMLE figures. Unofficial "conversion" charts for the Free 120 have no published method at all.

6. Decide on the date

In our experience, the question to ask is not "what will I score?" but "is my recent trend comfortably above where I need to be, under real timing?" If yes, keep the date. If the trend is close to the line, or depends on one good form, moving the date is usually cheaper than an extra attempt, since attempts are limited and stay on your record.

See which NBME forms to take and when, Step 2 CK percentiles and our NBME practice exams page.

NBME score conversion: what is official and what is not

The only official conversion is the one printed in your NBME score report. For the Step 1 and Step 2 CK self-assessments that means an estimated probability of passing plus equated percent correct scores (NBME CCSSA); for the shelf-style Mastery Series it means a table that maps your form score to an approximate Subject Exam score (NBME Mastery Series).

Everything else is somebody’s estimate. Online "NBME to Step 2 CK" converters and spreadsheets are usually built from scores students report about themselves, with no published method, no error band and no check against real results. Searches for a conversion for one numbered form, such as an old NBME 25, run into the same problem: a chart made for one form and one year of students does not carry over to another.

Conversion	Made by	Method published?	How we use it
NBME score report (CBSSA, CCSSA)	NBME	Described in NBME’s score interpretation materials	Main input for a test-date decision
Mastery Series conversion table	NBME	Included with the score report	Shelf readiness during the clerkship
QBank self-assessment estimate	The QBank company	Company’s own method	One extra data point
Online converters and spreadsheets	Websites and students	Usually not	Not for decisions

Our rule in sessions: if a number did not come from NBME or USMLE, it can raise a question but it cannot answer one. A converter that says 255 when your last three NBME forms show a flat trend near 240 is telling you about the converter, not about you.

Converting a shelf practice score

Shelf exams are scored differently from the Steps. NBME reports Clinical Science subject exams as equated percent correct on a 0 to 100 scale, the share of the content you have mastered, and sets no passing score of its own: schools use NBME’s grading guidelines to decide pass and honors (NBME Subject Exam Program Guide).

The practice tool for that scale is the Clinical Science Mastery Series: 50 questions per form, 1 hour 15 minutes at standard pace, $21, with a conversion table to the approximate Subject Exam score (NBME). Take it timed, then compare the converted score with your school’s grade cut-offs, which your clerkship coordinator can give you. For percentile ranks on the shelf scale, see NBME shelf exam percentiles.

Worked examples: three students, one exam

These examples are illustrations, not real students. The error figures come from NBME and USMLE (NBME CCSE guide, USMLE guidelines).

Turning an estimate into a percentile band

An estimate of 250 with an 8-point error gives a band of 242 to 258. On the current USMLE norm table, 240 is the 23rd percentile and 260 the 71st, so the honest reading of that single estimate is "somewhere between the mid-20s and about the 70th percentile". Several results narrow that range; one does not.

Why practice scores and real scores differ

• Measurement error on both tests, as above.
• Learning between the practice test and the exam, which usually raises scores.
• Test-day conditions: sleep, nerves, an unfamiliar centre.
• Form effects: some practice forms feel harder or easier than others.

A readiness checklist

• At least three recent NBME results, spaced out.
• A steady or rising trend.
• The lowest recent result still above your target band.
• Completed under real timing.
• Weak areas narrowing, not widening.

Step 1: reading a probability of passing

Because Step 1 is pass or fail, the CBSSA reports an estimated probability of passing rather than a score. The same logic applies: one high probability on one form is encouraging; a high probability that holds across several forms, taken under standard timing, is what to rely on. NBME says the estimate is not a promise of your result (NBME).

What the content-area report adds

Each self-assessment also reports performance by content area against a comparison group. NBME notes that content-area scores are less precise than the total (NBME). Use them to choose what to study, not to predict the total. If the same area is low on three forms, it is real; if it is low once, it may be noise.

When to move the date

These are our rules of thumb, not NBME or USMLE rules. The deadline that matters most is usually when programs need to see your score.

Mistakes that make predictors look wrong

• Taking the form in pieces or with extra breaks.
• Using self-paced mode for a date decision.
• Retaking a form you have already seen.
• Taking two forms on consecutive days and averaging them.
• Comparing a QBank estimate directly with an NBME estimate.

Why NBME uses a probability rather than a score

The CBSSA and CCSSA report an estimated probability of passing because a probability says directly what most students need to know: how likely a pass is if they tested with the same knowledge. The CCSE, which schools administer, adds a total score that estimates Step 2 CK performance (NBME CCSE guide). Both are estimates with error, and NBME says neither is a promise of your result (NBME).

A worked trend, form by form

This is an illustrative student, not a real record. Each step is within or close to the 8-point error band, but the direction over five forms is clear, and that direction is what the decision rests on.

Combining NBME, QBank and Free 120 results

These weights are our judgment, not an official ranking.

Questions to ask yourself after each form

• Did I take it under standard timing, in one sitting?
• Is the result inside the 8-point error band of my last one?
• Which content areas moved, and which did not?
• Did misses cluster at the end of sections, pointing to timing?
• What will I change before the next form?

If your NBME scores are falling

A downward run across two or three forms usually means fatigue, too many new questions without review, or a schedule that has squeezed out sleep. In our experience the fix is fewer questions with fuller review for a week, then another form. If the trend has not turned, talk to your advisor about the date.