Back to blog

Flashcards vs Practice Problems: What Each One Actually Trains

Flashcards train knowing, practice problems train doing. How to split your time by course type, and the one card to write whenever a problem gets you stuck.

If your exam is made of problems, practice problems are the main event and flashcards are support. If your exam is made of facts, it is the other way around. Most courses are a mix, and the useful question is not "which one is better" but "which part of this course is knowing and which part is doing," because flashcards are very good at the first and close to useless at the second.

Here is the one sentence version of the split. A flashcard can teach you which tool to pick up. Only a problem can teach you how to use the tool. If you have ever had a deck where you got every card right and then froze on the midterm, that is the split, and you were on the wrong side of it.

I work on Nebulearn, a flashcard app, so I have an obvious reason to tell you to make more cards. I am going to spend a good part of this page telling you where not to.

Knowing that versus knowing how

There are two different kinds of thing you can know, and they are stored and practiced differently.

The first kind is facts. The definition of enthalpy, the formula for the period of a pendulum, the conditions under which you are allowed to use a certain theorem. You either can say it or you cannot, and you can check in two seconds. This is what a flashcard is for. One cue, one answer, flip, grade.

The second kind is procedures. Setting up a free body diagram, pushing an integral through three substitutions, balancing a set of journal entries, tracing a pointer through a linked list. Knowing how is not a thing you can say. It is a thing you can do, and it has a lot of little decisions in it that you only meet when you are actually doing it.

The trap is that facts about a procedure feel like the procedure. You make a card that says "steps to solve a first order linear differential equation," and the back has five steps, and you learn the five steps. In review you read the front, you recite, "find the integrating factor, multiply through, recognize the product rule, integrate, solve for y," you flip, and you were right. Your brain is saying, "I know how to do these." Then the problem set gives you one where the equation is not in standard form yet, and you do not know what to do first, because step zero was never on the card. It was never on the card because you only find step zero by getting stuck on it.

That gap between reciting a method and running it is the same one I wrote about for coding in why your LeetCode cards feel easy but you still fail interviews. It is not special to programming. It shows up anywhere the exam asks you to produce work instead of a word.

What each one is actually good at

So you can see the division of labour in one place, here is how the two tools line up.

FlashcardsPractice problems
TrainsRecall of a single fact or conditionChoosing, sequencing and executing steps
FeedbackInstant, right or wrongSlow, and usually "wrong somewhere in the middle"
Time per itemSecondsMinutes to an hour
SpacingEasy, a scheduler does itHard, you have to plan it
Fails whenThe exam asks you to do somethingYou do not know the vocabulary or the formulas the problem assumes
Feels likeFast, clean progressSlow, messy, often discouraging

The last row is why people over-invest in cards. Twenty minutes of flashcards gives you eighty little wins. Twenty minutes of problems might give you one finished problem and one where you gave up. It is very natural to drift toward the thing that feels productive. I made a whole argument about that in stop tracking study hours, and the short version is that the comfortable activity is rarely the one the exam is going to grade.

The second-to-last row matters just as much in the other direction, though. Problems fail when you are missing the facts underneath them. If you sit down to a thermodynamics problem and have to look up three definitions and two formulas before you can start, most of your hour went to looking things up, and very little of it went to practicing the actual skill. That is where cards earn their place in a problem-based course. They clear the lookup work out of the way so the problem time is spent on problems.

Splitting your time by course type

Since nobody takes one pure kind of course, here is roughly how I would split a week of study time. These are starting points, my own rule of thumb, not something measured.

Course typeExamplesCardsProblems
Mostly factsAnatomy, pharmacology, vocabulary, history datesMost of the timeA little, as self-testing on old exams
MixedOrganic chemistry, physiology, accounting, statisticsAbout halfAbout half
Mostly proceduresCalculus, physics, circuits, algorithmsA small sliceMost of the time
Proof-basedAnalysis, abstract algebraDefinitions and theorem statements onlyWriting proofs

For the bottom two rows, "a small slice" really is small. Ten minutes a day of cards to keep the definitions, formulas and conditions warm, and then everything else goes to problems. If you are in a course like that and your card time is bigger than your problem time, something has gone sideways, and it usually means the cards have become a way of avoiding the problems.

For the mixed row, the split is not really about time. It is about order. You do the cards first so the terms are loaded, and then the problems, and the problems tell you which cards were missing. That feedback loop is the next section.

The card worth writing: the place you got stuck

This is the part that connects the two, and it is the one habit from this page I would keep if I could only keep one.

When you work a problem and get stuck, there is a specific moment where you stopped. You look at the solution, and you find the line where it does the thing you did not think of. You think, "oh, of course, you take the control volume around the whole thing instead of just the pipe." That moment, the moment of "oh, of course," is the most valuable thing the problem had to give you. And if you just nod and move on to the next problem, it is gone by Thursday.

So you write one card, right then. Not the whole solution. Just the trigger and the move:

Front: A flow problem gives you inlet and outlet conditions but nothing about what happens inside the device. What does that tell you about where to draw the boundary?

Back: Draw it around the whole device. Unknown internals are a hint that you are not supposed to need them.

That card is small enough to grade honestly, and it is about a decision, not a calculation. Next time you see a problem with that shape, the card is what fires. This is the same idea as the post-exam audit, where you card the questions you got wrong, just done continuously on your problem sets instead of once after the midterm.

A few rules that keep these cards useful:

  1. One card per stuck point, at most. If you got stuck in three places, the problem was too far above you, and you need an easier problem more than you need three cards.
  2. Card the condition, not the numbers. "When do I use this" is a fact. "What is the answer to problem 4.17" is not worth anything.
  3. Write it in your own words, at the moment you understand it. The textbook's phrasing is the phrasing that did not work on you the first time.

What to put on cards in a problem course

Even in a course that is ninety percent procedures, a short list of things does belong in a deck, and it is worth being specific, because "make cards for the formulas" is where it usually goes wrong.

Conditions of use. When is the formula valid? What has to be true before you can apply the theorem? This is the highest value card in any quantitative course, because picking the wrong tool is a far more common way to lose marks than arithmetic.

Definitions, stated precisely. In a proof-based course, a fuzzy definition makes every proof impossible. Card them exactly.

The formula itself, with meaning attached. Not just the symbols. What each term is, and what happens to the result when one of them grows. I went through that card shape in encoding math formulas.

Recognition cues. "A problem that mentions steady state is telling you which term is zero." These are the stuck-point cards from the last section, and over a term they become the most useful part of the deck.

Units and standard values. Small, boring, and they save real time.

And the things that do not belong: full worked solutions, multi-step derivations on one card, and anything where the back is a paragraph. If you find yourself wanting to put a worked example on a card, what you actually want is to redo that example on paper in a week. Put a reminder in your calendar and leave it out of the deck.

Problems need spacing too, and nobody does it for you

Here is the thing people miss when they move from a card-heavy course to a problem-heavy one. The reason flashcards work so well long term is not the cards. It is the spacing, which I covered in what spaced repetition is. A scheduler brings each card back just as you are about to lose it. Practice problems benefit from exactly the same treatment, and there is no scheduler doing it for you.

What most people do is finish problem set three, hand it in, and never look at that kind of problem again until the night before the final. Then they discover that the method from week three is gone.

The low effort version of spacing problems looks like this. Every time you sit down to do the current problem set, start with one problem from two weeks ago and one from a month ago. Pick ones you found hard. Do them cold, on paper, without the solution open. It costs fifteen minutes, and it also mixes old problem types in with new ones, so you have to work out which method applies instead of assuming it is whatever this week's chapter was about. That mixing is the idea behind interleaving two subjects, applied inside one course.

If you want to track it, a card can help here too, as a pointer and not as content. A card that just says "redo problem 3.12, the one with the two-stage compressor" is a perfectly good use of a scheduler. You are borrowing the scheduling without pretending the card is the practice.

Where other tools are the better call

Since I have told you to spend most of a problem course away from flashcards, it is only fair to say which tools are actually good for the problem side.

For standardized exams, a real question bank beats anything you can build. The questions are written in the exam's style, the explanations are written by people who know what the exam is testing, and the bank tracks which topics you are weak on. Students studying for medical boards, the CPA, the bar or the NCLEX are right to treat the qbank as the centre of their prep and cards as the thing that catches what the qbank reveals.

For coding, the judge on a practice site is the feedback, and no card can replace running the code. For math and physics, the old exams from your own course are the best material that exists, because they were written by the person writing your final. Use them.

And if your whole course is thirty definitions and a quiz on Friday, you do not need a system at all. Quizlet or a stack of paper cards will get you there, and setting up anything more is procrastination with extra steps.

When the balance tips the wrong way

Two failure modes are worth watching for, one on each side.

The first is memorizing the problems. You redo the same problem set three times, and by the third time you are getting everything right, and you think, "okay, I have got this." What you have got is those twelve problems. You recognize problem seven by its diagram and you remember the answer has a square root of two in it. Change the diagram and it falls apart. The check is simple: can you do a problem you have never seen, from a different textbook or a different year's exam? If not, you were reviewing answers, which is just passive review wearing a harder costume.

The second is the deck growing until it eats the problem time. In a course like physics, a deck that takes forty minutes a day is too big. Go through it and delete everything that is really a procedure pretending to be a fact. What should be left is conditions, definitions and stuck points, and that should fit in ten minutes.

The split I started with is the whole thing. Cards are for knowing, problems are for doing, and the stuck-point card is the piece that lets one feed the other. That is the workflow I had in mind for the capture side of Nebulearn: the PDF of the solution sits next to the card editor, so the moment you see the step you missed, you write the one card and go back to the problem. FSRS is on by default, so that card comes back about when you would otherwise have forgotten the trick.

Practice ProblemsFlashcardsSTEMStudy Workflow