Nobody can make this call for you, but you can make it far less vague. A score gap converts to a specific number of questions: closing 15 points takes about 13 extra correct answers on one Step 1 form and about 18 on the Step 2 CK UWSA 2 model. Once the gap is expressed in questions, the real question becomes whether the weeks you would buy are likely to produce that many, and what those weeks cost you elsewhere.
Step one: convert the gap into questions
Every conversion model here has a cost per wrong answer. Divide your gap by that cost and you have the size of the task.
| Assessment | Points per question | Questions needed for +10 | For +15 |
|---|---|---|---|
| NBME 28 (Step 1) | 1.05 | ~10 | ~14 |
| NBME 30 (Step 1) | 1.15 | ~9 | ~13 |
| NBME 14 (Step 2 CK) | 1.08 | ~9 | ~14 |
| UWSA 2 (Step 2 CK) | 0.85 | ~12 | ~18 |
Fourteen more correct answers out of 200 is a different proposition from "I need 15 more points". It is roughly seven percentage points of accuracy, which is a real amount of work but a bounded one.
Run your own raw result through the Step 1 form converter or the Step 2 CK form converter to get the cost per question for the form you actually sat. Do not borrow another form's number: across the seven Step 1 forms the cheapest and steepest curves differ by about 10%.
A worked example
You sat CCSSA Form 13 and missed 78 of 200. This site's model converts that to 215, with a pass reading of 60% and a realistic exam range of 220–227. Form 13 is one of the forms whose model corrects upward.
Your target is 230. Three different gaps are hiding in that sentence:
- From the converted estimate: 15 points, which at 1.07 points per question is about 14 more correct answers, or 64 wrong instead of 78.
- From the top of the realistic range: 3 points, roughly three questions.
- From the passing standard: you are already above it on both readings.
Which gap you are looking at changes the decision completely. A student who needs three more questions to reach the middle of their target range is in a very different position from one who needs fourteen, and both are in a different position again from someone whose estimate sits below 218.
Do this decomposition before you decide anything. It takes two minutes and it regularly turns "I am miles away" into "I am four questions and a fortnight away".
Step two: check how much evidence you actually have
One assessment is a data point, not a verdict. Two assessments a fortnight apart are a direction. Three, spaced across your preparation, are the closest thing to a trend you will get.
Before deciding anything, ask:
- How recent is the low result? A form from three weeks ago describes a student who has since done three weeks of work.
- Did you sit it under exam conditions: timed, full length, one day?
- Was it a form that costs more per wrong answer than the one before it?
- Does the low reading fall outside the band the model already admits to?
If the disappointing number is your only number, the fastest useful move is usually to collect a second one under proper conditions rather than to reschedule on the strength of the first.
A weighted forecast across your sittings is more stable than any single result, because it lets recency count for more without throwing the older data away.
Step three: separate "below target" from "below passing"
These are different decisions and they deserve different thresholds.
Step 1 is pass/fail. The only number that matters is 196 on this site's models, and the models are least reliable exactly where the decision is tightest, which is near the line. If your estimate sits within a few points of 196, you are inside the zone where the model cannot distinguish you from someone several points either side.
Step 2 CK is scored, so a below-target result is a trade-off rather than a binary. A delay that pushes your score release past an application deadline can cost more than the points you gain. That calculation depends on your timeline, not on the model.
Write down which of the two situations you are in before you look at any more numbers. The reasoning is different and mixing them produces bad decisions.
Step four: the constraints no calculator sees
A conversion model knows your raw score and nothing else. The decision depends on things it cannot observe:
- Your eligibility period. Registration windows have end dates, and rescheduling rules and fees vary. Check the current terms on usmle.org rather than on a forum.
- Your calendar. Rotations, research blocks, and application timelines each impose their own hard edges.
- Whether more dedicated time is actually usable. Extra weeks help if you have a specific plan for them. Extra weeks spent re-reading material you already know mostly buy fatigue.
- Cost. Rescheduling fees, another self-assessment, another month of question-bank access, and lost income all count.
- The asymmetry. A failed attempt is recorded on your USMLE transcript. A delay is not. That asymmetry is the strongest argument for postponing when the estimate genuinely sits near the passing standard, and it is a weak argument when the estimate is comfortably clear but short of an ambitious target.
A defensible default
If your most recent, exam-conditions estimate sits below the passing standard, and your exam is within two weeks, postponing is usually the conservative choice, provided you have a concrete plan for the extra time.
If the estimate is clear of the standard but under a personal target, the question becomes what the target is for. Delay buys points only at the rate the table above implies, and only if the study time is genuinely productive.
If your assessments disagree, resolve that before you touch the date. Convert them on a common scale, weight the recent ones more heavily, and see whether the disagreement survives.
Whatever you decide, convert the decision into a schedule immediately. The study plan builder takes a current estimate and an exam date and lays out the weeks, including where the remaining assessments go. A postponement without a plan tends to become the same three weeks you just had.
The checklist, before you touch the reschedule button
Answer these in writing. If you cannot answer one of them, that is the thing to go and find out, not a reason to decide faster.
- What is my most recent estimate, from a full-length sitting under exam conditions?
- How many additional correct answers does my gap represent on that specific form?
- Do I have a second, independent result that agrees with the first?
- Is the gap larger than the band this model already carries?
- What exactly would I do with the extra weeks that I am not doing now?
- When does my eligibility period end, and what does rescheduling cost?
- What downstream dates does a delay push: applications, rotations, a second exam?
- Who else needs to know, and have I spoken to my school's advisor?
A student who can answer all eight rarely needs anyone to tell them the answer. A student who can answer two is not yet at the point of deciding.
Quick answers
How many points can I gain in two weeks? There is no verifiable answer, and any site quoting a points-per-week figure is inventing it. What you can quantify is the target: the table above tells you how many additional correct answers your gap represents.
My estimate is a few points below the passing standard. Is that decisive? No, but it is the situation where the model is least able to tell you anything, and where the downside of being wrong is largest. Treat it as a prompt to collect another data point quickly.
Should I postpone because my score dropped on the latest form? Check the form's cost per wrong answer first. Some of the drop may be the conversion line rather than your performance.
Does postponing look bad to programs? An appointment you move but never sit leaves no record of its own; the attempts you actually take, including failures, appear on your USMLE transcript. That is the asymmetry to weigh, alongside whatever your own programme requires.
What if I cannot postpone? Then the decision is made, and the useful work is triage: your weakest two systems, timing discipline, and protecting sleep in the final days.
The limits of everything above
These estimates come from conversion lines built on community-reported data, not official equating, and NBME publishes no wrong-answers-to-score formula for its self-assessments. They cannot forecast your exam day, and they say nothing about your health, your finances, or your programme's requirements.
This is a framework for thinking, not advice about your particular situation. For that, talk to your school's academic support office or advisor, who can see the constraints a calculator cannot.