You aced the chapter 3 problems. The exam mixes them.

I could do every problem at the end of chapter 3. Every single one, quickly, correctly.

Then the exam put a chapter 3 problem, a chapter 5 problem, and a chapter 2 problem next to each other with no labels, and I sat there for two minutes on the first one not knowing where to start.

The material wasn't the issue. I knew the material. What I'd never once practiced was deciding which thing this was.

The hidden hint in the problem set

When you work through chapter 3's problems in order, you already know the answer to the hardest question before you start.

The chapter told you. The section heading told you. The fact that you just did four of these told you. So every problem begins from "apply the chain rule, correctly" instead of "what kind of problem is this."

On the exam, that hint is gone. Problems arrive unlabeled and out of order, and the first thing you have to do — the thing that determines whether you get anywhere at all — is classify. If you've done a hundred practice problems and zero classification reps, you're going in with one skill and being tested on two.

The fix takes about four minutes

Mix the practice. That's the whole technique.

Instead of doing ten chapter 3 problems, do three from chapter 3, two from chapter 5, three from chapter 2, two from chapter 4, shuffled. Now every problem starts with a decision.

Concretely:

• Write problem numbers on index cards or in a list, from several chapters, and draw them at random
• Or work backward through a chapter, then jump to a different one after every second problem
• Or, easiest of all, use an old exam, which is already mixed by construction

It's going to feel worse and be better

This is the same trap as retrieval practice. Blocked practice feels great — you get faster within a set, you get more right, you finish more problems per hour. Mixed practice feels clumsy. You get more wrong. You go slower.

You're going slower because you're doing an extra step, and the extra step is the one you're actually going to need. The confidence you built by doing ten chapter 3 problems in a row was confidence about a task that will never be on an exam.

The rate at which you finish practice problems is not a measure of anything. It's just a number that goes up when you make the practice easier.

A middle path that works better than either extreme

I'm not going to pretend I mix from the first day. You need some blocked practice when a method is brand new — the first two or three problems of a new technique should be blocked, because you're still learning the mechanics and there's nothing to classify yet.

So:

Learning a new method: block it. Three or four in a row. Get the procedure into your hands.

Everything after that: mix it. New material plus a couple from two weeks ago, every session.

Two weeks before the exam: mix everything, all the time, and stop doing single-chapter sets entirely.

Where this shows up outside of math

Anywhere a course teaches you a set of tools that all look reasonable in isolation.

Statistics. Every test looks obvious when it's in the chapter about that test. Given data and a question, choosing between a t-test, a chi-square, and an ANOVA is the actual skill, and it's not practiced by the chapter exercises.

Chemistry. Reaction mechanisms are recognizable in their own chapter and much less so when someone hands you a synthesis problem and asks how to get from one molecule to another.

Grammar or language. Drilling one tense at a time makes you good at that tense. Mixing tenses makes you good at speaking.

Case-based courses. A case in the chapter on one framework will always look like an application of that framework. That's a fact about the textbook, not about the case.

The one-sentence version

Practice the way the exam asks, not the way the book is organized.

The book is organized for teaching, one idea at a time, and that's correct — it's how you should learn it the first time. But the exam is organized like the world, unlabeled and out of order, and somewhere between those two things you have to do the practice that bridges them. Almost nobody does, which is why "I understood everything and then blanked" is such a universal experience.