Score conversion

NBME 16 to Step 2 CK

The newest form, and one of the strongest single signals you can get.

NBME 16 carries no correction. NBME 16 prints a few points higher than the older forms for the same student, which makes people suspicious of it. The data does not support discounting it: students who print high on 16 really do score high on the exam. It is counted at face value, with no correction.

NBME 16 conversion table

Printed score to projected Step 2 CK range, for a form taken in the last month. The range is about a 68% band: roughly two in three students land inside it, which is why it is a range and not a single number. Take the exam further out and the band widens.

NBME 16 printedProjected Step 2 CKMidpointPercentile
200209 to 2312204th
205212 to 2342235th
210216 to 2382278th
215220 to 24223111th
220224 to 24623516th
225228 to 25023922nd
230232 to 25424330th
235236 to 25824739th
240240 to 26225150th
245244 to 26625560th
250248 to 27025971st
255252 to 27426381st
260256 to 27826789th
265260 to 28227195th
270264 to 28627598th

Rows shaded amber have a projected floor below the 218 pass line. Percentile is against LCME first-takers. Your printed score is never restated as a lower number: where a correction applies it lives inside the math, not on your report.

Blend several forms and add your exam date

How the projection works

Three steps, and the constants are published rather than described. Then it is shrunk toward the population mean of 250 with a slope of 0.79, which is what stops a single high form from projecting a fantasy. Then 9 points are added for the measured gain from practice form to real exam.

Fit on 258 real posted score reports and tested blind on 32 it had never seen: average miss 4.1 points, bias statistically zero. Reddit posters skew toward good news, so treat the range as honest rather than generous. Full methodology, including where the model stops being trustworthy.

Questions

Is NBME 16 inflated?
It prints higher than older forms, but that is not the same as inflated. High printers on 16 go on to score high, so counting it at face value predicts better than discounting it would. No correction is applied.
Is NBME 16 accurate?
It is vendor-equated onto the same scale as the other modern forms, so it needs no correction here. Accuracy in the sense that matters, how close the projection lands to a real score, was tested blind on 32 score reports the model had never seen: average miss 4.1 points.
How much does the date I took it matter?
A lot, and more than most converters admit. Forms are weighted by recency with a roughly 21 day half-life, and the width of the projected range depends on how far out your exam is. A form from three months ago tells you much less than one from last week.

Other forms

Get Step Gunner