Free · No sign-up · Shows its work

Step 2 score predictor that shows its work

Any NBME 9 to 16 or UWSA score becomes a projected Step 2 CK range, fit on 258 real posted score reports and tested blind on 32 more: average miss 4.1 points. The range appears as you type.

dated forms weigh more
Your conversion appears here

Pick a form and enter the printed score. No sign-up, it answers as you type.

FORMS
this one
each dated form tightens this range
The part you control

Your form is the floor. The next three weeks decide the beat.

+9median beat
+10 to 11with 3+ weeks of grind

The gap between a practice form and the real thing is not luck. It is the reps you put in after. Step Gunner runs those weeks: 7,000+ Step 2 CK questions, a projection that moves with your accuracy, and a planner that says exactly what your next form needs.

Get Step Gunner free
Free to start. This predictor stays free either way.
146 tracked reports. The biggest beats belong to students who kept working.
02

The receipts

258reports, in the open
The formula is printed on this page.Fit on 258 real posted score reports, methodology public, nothing to take on faith.
4.1avg miss, blind
Frozen, then tested on 32 unseen reports.Bias +0.97, statistically zero. Nine in ten actual scores landed within 8 points.
3forms decoded
It knows which forms lie.NBME 9 prints 5.5 low, UWSA 2 prints 4.8 hot, UWSA 3 prints 6.3 low. A flat lookup chart hands you the wrong number.
+11at 3+ weeks out
It knows your runway.A form taken three weeks out means a bigger beat than one taken the night before, so the date changes the math.
100%of misses public
It grades itself in public.Every new real score gets checked against our projection, and the misses stay on the page.
Read the error bars honestly.

Score-release threads skew toward good news, so our published misses are a floor, not a ceiling. No practice form can promise a score. That is why everything on this page is a range.

How the conversion works

Most NBME conversion charts are a single lookup table with no dates and no source. This one runs the same calibrated engine as the Step Gunner app, in three steps:

1. Scale translation. Three forms print off the common scale and get translated before anything else: old NBME 9 prints about 5.5 low (credited back), UWSA 2 prints about 4.8 hot, UWSA 3 about 6.3 low. Forms 10 through 16 are equated by NBME itself and enter at face value. Your printed number is never shown reduced.

2. Recency weighting. With multiple forms, each one is weighted by exp(minus age over 30 days). A form from this week dominates; a form from two months ago barely whispers. This is also the honest answer to which form predicts best: the one you took last.

3. The calibrated map. The blended level is mapped through 250 + 0.79 x (level minus 250) + 9, fit on real reported outcomes. The +9 is the measured practice-to-real gain, which is why your projection usually sits above your latest form. The result is shown as a range, because a point estimate would be pretending.

The blind test

In August 2026 we froze the model and scored 32 new score reports it had never seen, from the newest score release thread. Result: average miss 4.1 points, bias +0.97 (statistically zero), 88 percent of actual scores inside plus or minus 7 and 94 percent inside plus or minus 8. Also: 26 of the 32 scored at or above their freshest NBME, which is the pattern the +9 gain encodes. Full write-up and reproduction scripts on the methodology page.

Questions students actually ask

How accurate is this Step 2 score predictor?

Tested blind on 32 score reports it had never seen (the August 2026 release thread): average miss 4.1 points, bias +0.97 (statistically zero), and roughly 9 out of 10 actual scores landed within 8 points of the projection. The fit comes from 258 real posted score reports. Reddit posters skew toward good news, so treat the range as honest but not generous.

Is NBME 16 accurate, and does it convert high?

NBME 16 prints a few points higher than the older forms for the same student. But people who print high on 16 really do score high on the exam, so this predictor counts your 16 at face value. It is one of the strongest single forms you can take.

Do UWSA scores inflate?

UWSA 2 prints about 5 points hot relative to the NBME scale, so this converter translates it down before blending. UWSA 1 sits roughly on scale. UWSA 3 runs a few points low on thinner data. Your printed number is never changed on screen, only inside the math.

Is old NBME 9 still worth anything?

Form 9 genuinely prints about 5 to 6 points low, so it gets credited back in the conversion. It is old and it fades fast under the recency weighting, but it is not garbage.

Which NBME form predicts Step 2 best?

Mostly whichever one you took last. Once you control for when forms are taken, forms 10 through 16 are statistically indistinguishable as predictors. Recency is what matters: a form from this week says far more about you than a form from six weeks ago.

What score do I need on my next NBME?

That depends on your goal, your other forms, and your timing. The Step Gunner app includes an NBME Planner that answers it exactly: pick the form you have not taken, and it shows the required score for each goal on your timeline.

One form at a time

Every form, what it prints, and a full conversion table.

The predictor is a snapshot. The app is the film.Get the app