If you want a final measurement, UWorld Self-Assessment 2 gives you more to measure: 160 questions, a wrong-answer input, and a full score range. If you want a rehearsal of the real interface and pacing, the Free 120 is the closer match, but the models that convert it are deliberately blunt: the Step 1 version cannot output above 220 even at 100% correct. Most students who have time for both should rehearse with the Free 120 and measure with UWSA 2.
The two instruments are not measuring the same thing
The Free 120 is a 120-question official practice set. Its value is that the questions and the delivery resemble the exam you are about to sit. Its weakness as a measurement is sample size: fewer items, and no published conversion from NBME.
UWSA 2 is a 160-question UWorld self-assessment sold as a predictive instrument. It reports on a three-digit scale of its own, which is what makes a conversion possible in the first place, and also what makes a correction necessary.
Neither is an equated exam. Both estimates on this site come from conversion lines built on community-reported data, and the difference in their design shows up directly in the numbers below.
What each point is worth
Score resolution is the clearest practical difference. On this site's models:
| Instrument | Questions | Score movement per question |
|---|---|---|
| Free 120 (Step 1) | 120 | ~0.63 points |
| Free 120 (Step 2 CK) | 120 | ~1.42 points |
| UWSA 2 (Step 1) | 160 | ~1.10 points |
| UWSA 2 (Step 2 CK) | 160 | 0.85 points |
The Step 2 CK Free 120 model is the twitchiest of the four: three extra misses move the estimate more than four points. The Step 2 CK UWSA 2 model is the steadiest, at 0.85 points per question, noticeably flatter than the CCSSA forms, which sit near 1.1.
Flatter is not more accurate. It means the instrument's own scale is stretched relative to the exam scale, which is part of why its estimate needs adjusting.
The Free 120 ceiling, in numbers
The Step 1 Free 120 model here is percent-in, score-out, and it compresses hard at the top:
- 60% correct → 190
- 68% correct → 196 (the passing standard)
- 80% correct → 205
- 90% correct → 213
- 100% correct → 220
A perfect run returns 220. That is not a bug in the calculator; it is a refusal to pretend that 120 questions can distinguish a 230 from a 260. If you want to know whether you are safely clear of the line, this is a reasonable instrument. If you want a target-score reading, it cannot give you one. You can see the whole curve on the Step 1 percent-correct calculator.
The Step 2 CK version behaves differently, because Step 2 CK is scored rather than pass/fail: 67% correct converts to roughly the 218 passing standard, 70% to 223, 80% to 240 and 90% to 257. It has range, but its pass-probability curve is deliberately flat, topping out at 97% and bottoming at 3%. Even an estimate of 240 reads only 85%. The Step 2 CK percent-correct calculator shows that reluctance directly.
The UWSA 2 correction, in numbers
The Step 2 CK UWSA 2 model on this site corrects downward. A converted estimate carries a realistic range that sits at or below the number:
| Wrong of 160 | Converted estimate | Realistic range |
|---|---|---|
| 30 | 255 | 247–255 |
| 40 | 246 | 238–246 |
| 50 | 238 | 230–238 |
| 65 | 225 | 217–225 |
Its pass curve is also shifted: the 50% point sits at 223 rather than 218, so an estimated 218 returns 32%. Read that as the model refusing to call a borderline UWSA 2 result safe. The full explanation lives on the UWSA 2 Step 2 CK page.
For Step 1, UWSA 2 converts on a straight line (55 wrong of 160 lands at 237, 65 wrong at 226), and the model carries no published downward correction, which is itself a limitation rather than a reassurance.
Which to take last
Take the Free 120 last if you are anxious about pacing, block structure, or the interface; you have already collected two or three full-length assessments; and you understand that a good percentage here is weaker evidence than the same percentage on a full form. Sitting it two or three days out gives you a rehearsal without leaving time to spiral about the number.
Take UWSA 2 last if you need a decision, whether that is sit or postpone, target reached or not, and you can afford a result that might unsettle you. Give yourself at least a week afterwards so a bad reading is still actionable.
If you can do both, the usual ordering is UWSA 2 about a week out and the Free 120 in the final few days. That sequence puts the measurement where you can still act on it and the rehearsal where it does the most good.
Reading a Free 120 percentage without over-reading it
With 120 items, every question is worth 0.83 percentage points. On the Step 2 CK model that translates to a coarse scale: 72% correct converts to 226, 75% to 232. Four questions, six points.
That granularity is the whole caution. A student who reports "I got 75% on the Free 120" is reporting a number that would have been 226 instead of 232 if four items had gone the other way, and four items is well within the range of a bad block, a misread stem, or two lucky guesses.
The same arithmetic applies on the Step 1 side, where the compression works in your favour by refusing to over-claim: 72% is 199 and 75% is 201. Two points apart. The model is telling you it cannot resolve the difference, which is more honest than a tidy number would be.
A sequencing plan for the final ten days
One workable shape, assuming you have both instruments available and roughly ten days left. Adjust it to your own calendar rather than following it literally.
- Ten days out: sit UWSA 2 (or your last full CCSSA form) under exam conditions: full length, correct block timing, no reference material.
- Nine to seven days out: review every miss, sorted by cause rather than by subject. This is where the assessment earns its cost.
- Six to four days out: targeted work on the two weakest areas the review exposed, plus daily timed blocks to hold pacing.
- Three days out: sit the Free 120 as a rehearsal. Note the pacing and the interface; convert the percentage, then put the number down.
- The final two days: light review of your own notes, no new material, and no further assessments.
The reason UWSA 2 goes first is actionability. A result you cannot respond to is entertainment; a result with a week behind it can still change what you study. The reason the Free 120 goes late is that its value is rehearsal, and rehearsal is worth most when it is fresh on exam day.
What neither can tell you
Nobody has published a verifiable comparison of how well these two instruments predict real outcomes. Claims that one is accurate "within a few points" are not independently checkable, including when a vendor makes them about its own product.
What can be said is structural. The Free 120 gives a smaller sample and a compressed scale. UWSA 2 gives a larger sample on a stretched scale that needs correcting. Both are converted here by lines built on community-reported data, not by NBME's equating, and no official wrong-answers-to-score formula exists for these self-assessments.
Quick answers
Is the Free 120 easier than the NBME forms? It is widely regarded as running easier, which is why this site treats a strong Free 120 percentage as a floor rather than a ceiling. That is a conservative qualitative position, not a measured offset.
Should I take the Free 120 the day before the exam? Taking it very late leaves no room to respond to the result. Two to three days out is a common compromise; the rehearsal value survives, and a surprise still leaves a day to steady yourself.
My UWSA 2 and my last NBME form disagree. Which counts? Weight the more recent sitting, then check whether the gap exceeds the bands both models carry. If it does, putting both results on one Step 2 CK scale is more useful than picking a winner.
Does a 90% on the Free 120 mean a 250? On the Step 1 model, no: the output stops at 220. On the Step 2 CK model, 90% converts to about 257, but from a 120-question sample that number deserves much less confidence than the same estimate from a 200-question form.
Can I convert the Free 120 from a wrong-answer count? The models here take a percentage. Convert your wrong-answer count to a percentage of 120 first; the arithmetic is exact, so nothing is lost.