Around week six of every prep calendar, the same search gets typed into Google at midnight: "what percentage of practice questions do I need to pass the bar?" For the legacy MBE there was a whole industry of answers — raw-score conversion charts, 58–67% targets keyed to your state's cut score. For the NextGen exam, here is the truthful starting point: anyone selling you a precise percentage is guessing. The 500–750 score is equated, which makes raw-to-scaled conversion unknowable outside NCBE by design — and the first administration's results don't even exist until this fall.
That's not the same as "nothing is knowable." There are honest anchors, a defensible working benchmark, and — more useful than either — a set of readiness signals that predict better than any single percentage. This post is the whole toolkit.
Why the old percentage charts don't transfer
Three structural reasons, so you can spot borrowed-chart advice on sight:
- Different scale, one integrated score. The legacy MBE reported its own scaled component; NextGen feeds MCQs, integrated question sets, and performance tasks into a single number. There is no "MBE subscore" to benchmark against a chart.
- Different raw-percent math. Standalone MCQs are only about 49% of the score — and the select-two format awards partial credit (one correct pick of two), which drags raw percentages in ways the old charts never modeled.
- No conversion data exists yet. The first NextGen exam runs July 28–29, 2026; scoring and equating follow; jurisdictions release in October. Until real distributions exist, precision is theater.
What the legacy charts do still teach is the shape of the thing: on the MBE, the passing zone for most jurisdictions sat in the low-to-mid 60s percent raw — roughly 58–65% depending on the state's cut. The bar was never "get 90%." Wrong answers at meaningful volume were always part of a passing performance. That shape carries over even though the decimals don't.
The anchors that are real
- NCBE's recommended passing band is 610–620, which it describes as corresponding to the familiar 260–270 on the legacy scale — the product of standard-setting studies with 80+ panelists and concordance analyses linking the two exams. Continuity was engineered, not assumed.
- Announced state cuts cluster at 610–620, mapping onto each state's old number — the full list is in passing scores by state. If your state is still TBA, plan against 620, the top of the band, and let reality only ever surprise you pleasantly.
- Because continuity was the design goal, legacy practice heuristics carry in spirit: a candidate consistently answering in the low-to-mid 60s on honest practice questions was in the passing conversation on the old exam, and there's no structural reason the new one moved that neighborhood dramatically.
The benchmark we use — labeled as what it is
On BARGO, the readiness line on practice MCQs is 60%, with 70%+ as the comfort zone — a heuristic, and we say so everywhere it appears. The reasoning is positional: the passing band sits under halfway up the 500–750 range, and practicing comfortably above the scale's midpoint on mixed-difficulty questions leaves headroom for exam-day variance. The scoring guide walks the full logic, along with the guardrails we build around estimates: no projected score until you've answered 50+ questions, your state's actual passing line drawn on the chart, and written components flagged as self-assessed rather than silently blended.
One benchmark, though, is worth less than a curve.
What "on track" looks like, week by week
Against a standard 10-week runway:
| Phase | Mixed timed accuracy | What matters more than the number |
|---|---|---|
| Weeks 1–2 | 40–55% is normal | This is diagnosis, not verdict — a 45% baseline is information you paid nothing for |
| Weeks 3–5 | Mid-50s, climbing | Single-subject sets right after studying that subject should clear 60% |
| Weeks 6–8 | 60%+ sustained | Weak subjects pulling within ~10 points of strong ones; timed-vs-untimed gap narrowing |
| Weeks 9–10 | 65%+ at full pace | Simulated sessions steady; error review producing repeat-miss rates near zero |
Three honesty rules make the curve mean something:
- Fresh questions only. Accuracy on questions you've seen before measures memory of the bank, not readiness. Repeating a small free question pool until it flatters you is the most common self-deception in bar prep.
- Honest difficulty mix. A bank of softballs makes everyone feel ready. Practice sets should include the select-two items — and expect your select-two percentage to run below your single-answer percentage. That's the format's partial-credit math, not a doctrine problem.
- Timed, at the real 1.8-minute pace. An unfinished section is a pacing problem wearing a knowledge problem's costume. By the final weeks, your timed and untimed accuracy should be nearly the same number.
The half a percentage can't see
Here is the uncomfortable arithmetic: MCQ accuracy — the number everyone tracks — maps to about 49% of the score. Written and applied work carries the rest, and it has its own readiness signals: rubric self-scores on performance tasks trending up, tasks finishing inside their 60 minutes, and your issue-hit-rate against model answers climbing across attempts. A candidate at 72% on MCQs who has never self-graded a written answer is half-ready and fully confident — the signature NextGen prep failure. If your practice week doesn't produce at least a few rubric scores alongside the accuracy stats, the dashboard is measuring the wrong exam.
When the number is bad: the three-autopsy rule
A weak percentage is a symptom with three different diseases, and the fix depends on which one:
- Ran out of time → pacing. Drill smaller timed blocks; practice triage (flag-and-move); stop reading explanations mid-set.
- Narrowed to two, picked wrong → recall precision. The rule is familiar but not produced — that's a spaced-repetition problem, and the miss should generate a flashcard the same day.
- Never knew the rule → coverage. Back to the chapter, one honest pass, then requestion. If a whole starred topic is dark, it goes into next week's rotation, not tonight's panic.
Tag ten misses this way and you'll usually find one mode dominating — which converts "I'm at 54% and scared" into "I have a pacing problem," a thing with a known fix.
Averages hide; spreads tell
The NextGen score is integrated — there are no per-subject minimums — but a 68% average concealing a 45% subject is still a risk profile, not a pass profile: that subject is where exam-day variance does its damage. Readiness means no subject stranded far below your line. It's why our analytics split accuracy by subject and topic, and why the weekly rotation should come from miss data, not from which subject feels scariest — feared and weak are usually different subjects.
What October changes
The first NextGen results land October 2 in Missouri, October 9 in Washington, with the rest of the first wave behind them — the first real pass rates, and possibly the first public score-distribution data, in the exam's history. When they publish, we'll fold anything usable back into this post and into the dashboard's estimates. Until then, distrust precision, trust trendlines.
Start yours with real data: the free 10-question diagnostic runs timed at the real pace with full explanations and a subject-by-subject readout — the first point on the only curve that matters. From there, the $99 lifetime platform tracks accuracy by subject and topic across 2,900+ questions, draws your state's passing line on the chart, and tells you — with labeled honesty — whether "on track" is what the evidence actually shows.