When You Have Memorised the Question Bank Instead of the Law

Six weeks out, your practice scores are good and getting better. You have been through the question bank twice. On the third pass you are hitting 85%.

Then a friend sends you a question from a different course, on a subject you have supposedly mastered, and you cannot do it.

Nothing has gone wrong with your memory. You have simply learned the question bank rather than the law, and the two feel identical from the inside right up until the exam.

Why the third pass teaches you nothing

A practice question does its work the first time you see it, when you have to find the rule, apply it to the facts and choose. That is the whole exercise.

The second time, you half-remember the answer. The third time, you recognise the fact pattern from its first sentence, and what you are practising is recall of a previous answer — a completely different skill from the one being examined.

This is the structural limit of every commercial bar course: the bank is finite. Two to three thousand MBE questions is a lot, and it is still a number you exhaust. What happens after that is that your scores keep rising while your actual readiness plateaus, which is the most dangerous shape a study plan can take. You feel more prepared precisely as you stop improving.

The same applies to essays, and worse. Most courses offer a handful of graded essays, if any, because a human has to read them. Ten essays across a nine-week preparation is not deliberate practice; it is sampling.

The two things that actually run out

Be precise about what the constraint is, because it determines the fix.

Novel questions on demand. Not more questions — unseen ones, on the subject you are weak in, today. A fixed bank cannot do this by definition; once you have seen it, it is spent.

Feedback on written answers. The MBE marks itself. Essays and performance tests do not, and the gap between writing an essay and finding out what was wrong with it is where most bar candidates stop improving on the written portion. Why that gap dominates everything else is its own argument.

Everything else a commercial course gives you — the outlines, the lectures, the schedule, the discipline of a fixed structure — remains genuinely valuable. This is not an argument for dropping it. It is an argument about what to do in the last third, when the bank is spent and there are still five weeks.

Generating questions instead of re-reading old ones

Given your own outlines and the rules you are working from, a model can produce new MBE-style questions on any subject, at the difficulty you ask for, concentrated on what you keep failing. The question is not whether it can — it is whether the questions are good.

Two honest constraints, and they matter more here than in any other subject.

Verify every rule statement. Generated questions can misstate a rule, cite the wrong standard or apply a minority position as if it were majority. For the bar, where the whole exercise is precision about rules, a confidently wrong explanation learned in week seven is worse than no practice at all. Cross-check anything that surprises you against your outline or the source.

Generate from your material, not from thin air. Questions built from your own outlines inherit your jurisdiction and your course's framing. Questions generated from a bare prompt inherit whatever the model averaged over, which may be the wrong state's rule. In ExamFlow this means uploading the outlines you are already using and generating questions from them, rather than asking for "bar questions on contracts".

Used that way, the failure mode changes from exhausting the bank to needing to check the output, and checking is itself a form of study.

A workflow for the last five weeks

This assumes you still have your commercial outlines and are past the first pass.

Morning, 2-3 hours — learn and condense. Work through the day's subject in your outlines, then write a condensed outline from memory. From memory is the whole point: it is retrieval, not review.

Afternoon, 2-3 hours — practice. A 50-question set on the day's subject, timed. Then review every wrong answer and, crucially, every right answer you were unsure of — those are the ones that will flip under pressure. Then two essays on the same subject, marked immediately, with the feedback read before you write the second.

Evening, 1 hour — spaced review. Cards from earlier subjects, plus a mixed 25-question set spanning everything covered so far. Mixed sets are uncomfortable and they are the closest thing to the real exam, where nothing tells you which subject a question belongs to.

Saturday — full simulation. 200 questions, three hours morning and three hours afternoon. Not for the score; for the stamina and for finding out what your accuracy does in hour five.

Sunday — six timed essays, 30 minutes each. Then read all six pieces of feedback together and look for the repeated error rather than the individual ones. The pattern is the thing to fix.

What this is worth, honestly

The last two rows are the trade. You gain volume and targeting; you take on the job of verification and you keep needing the course for structure. Anyone telling you it replaces a bar course is selling something.

Reading your practice scores honestly

The last weeks of bar prep are governed by numbers that are easy to misread, and misreading them is how people arrive confident and unprepared.

A rising score on a reused bank means nothing. If the questions are not new, the trend measures familiarity. The only score worth tracking is on questions you have never seen. Everything else is noise dressed as progress.

Mixed sets are the real measure. Accuracy on a single-subject set is inflated, because knowing the subject does half the work of the question — you are not being asked to identify what area of law this is. Your mixed-set accuracy is typically several points below your subject-set accuracy, and mixed is what the exam is.

Track uncertainty, not just errors. Mark every question you were unsure about, including the ones you got right. Those are the ones that flip in hour five when you are tired. A candidate at 80% with a quarter of the answers guessed is in a very different position from one at 75% with none.

Watch accuracy by hour, not overall. In a full simulation, compare your first fifty questions with your last fifty. If it drops sharply, the problem is stamina, and the fix is more full-length simulations rather than more questions.

Essays: count repeated errors, not scores. An AI essay score is orientative and should not be tracked as a number. What is worth tracking is the list of errors that keep recurring. When the same structural mistake appears in three essays, that is the thing to fix, and it is worth more than another ten essays with the same flaw.

The same shape, other exams

This pattern — finite bank, exhausted early, written portion unmarked — is not specific to the bar. It applies to the CPA exam, medical boards with clinical vignettes, the CFA, and the LSAT. Any exam that tests applying knowledge rather than recognising facts hits the same wall at the same point in preparation.

The diagnostic is always identical: cover the answer options, state the rule out loud, then look. If the rule does not come but the right answer does, you have learned the bank.

If the bank is spent and the exam is not: generate new questions from your own outlines and have your essays marked tonight.