How to revise a topic you understand but keep missing

If you already understand a topic and keep missing problems on it, the topic is not what needs revising. You could explain modular arithmetic to a friend. You can follow every line of the official solution. And on the paper, under time, you still do not get there.

That gap is real, and rereading the chapter does not close it, because the chapter is the part you already have. Rereading feels productive for exactly that reason: everything on the page is familiar, nothing resists, and you finish believing you are ready. Then the next problem on the same topic goes the same way.

The fix is to stop revising the topic and start revising the step.

Understanding is not the same skill as producing

Following a solution and producing one are different acts. When you read a solution, someone has already made every decision for you: which modulus, which bound, which quantity to compare. You check each step and each step is fine. When you face a blank page, the decisions are the whole job, and none of them is written down anywhere in the chapter you keep rereading.

So "I understand it" usually means "I can verify it". The test that tells the two apart is cheap. Take a problem you missed last week, close the solution, and write it out from a blank page. If you stall, note where. That point is what you are actually missing, and it is almost never the definition.

Sort your misses before you revise anything

Go through the problems you missed on one topic and give each a one-word label for the step that broke. Nearly every miss lands in one of three bins.

• Recognition. You did not see that the problem belonged to this topic at all. You tried algebra for twenty minutes on something that falls to a residue table in two lines.
• Choice. You saw the topic and made the wrong first decision inside it. Right idea, wrong modulus. Right idea, a bound too loose to finish.
• Execution. You chose correctly and then dropped a case, or slipped an arithmetic step, or proved half of an "if and only if".

The bins need different revision, and that is the point of sorting. A recognition miss is trained by reading many problems and naming the idea without solving. A choice miss is trained by drilling the decision itself. An execution miss is trained by writing out full solutions slowly and checking each line against the claim it is meant to prove. Rereading the chapter trains none of them.

Most students who keep missing a topic they understand find that their misses cluster in one bin. That cluster is your revision plan for the week.

A choice miss, worked: which modulus

Take $x^2 + y^2 = 3z^2$ has no solutions in positive integers. A student who understands modular arithmetic knows exactly what to do in principle: reduce modulo something and find a contradiction. Then they try mod 2, mod 4, mod 5, run out of time, and walk away believing they need to revise modular arithmetic.

They do not. They need to revise one decision. The $3$ in the equation is the signal, and the fact that does the work is that squares modulo 3 are only $0$ or $1$. So $x^2 + y^2 \equiv 0 \pmod 3$ forces both $x$ and $y$ to be multiples of 3, and from there the equation shrinks to a smaller copy of itself.

Revising that miss means writing the decision down as a rule you can apply cold, not rereading the topic:

• A coefficient or constant in the equation is often the modulus. Try it first.
• Squares: mod 3 leaves $\{0, 1\}$, mod 4 leaves $\{0, 1\}$, mod 8 leaves $\{0, 1, 4\}$.
• Cubes: mod 7 leaves $\{0, 1, 6\}$, mod 9 leaves $\{0, 1, 8\}$.

That is three lines. It is the whole of what this student was missing, and it took five minutes to write. The modulus guide builds the full version of the table.

The same shape shows up in IMO 1959, Problem 1: prove $\frac{21n+4}{14n+3}$ is irreducible. Everyone who misses it knows what a gcd is. What they miss is the decision to look for a combination of numerator and denominator that is a constant, here $3(14n+3) - 2(21n+4) = 1$. Write that move down once as a rule ("to show a gcd is 1, find a combination equal to 1") and the next problem of this kind is a two-line problem.

A choice miss, worked: what to bound

Bounding misses look different but break at the same place. Take find all $n$ with $n+1$ dividing $n^2+1$. The student who misses it often tests small cases, spots that $n = 1$ works, and cannot prove nothing else does.

The missing decision is to shrink the expression until what is left is too small to be divisible by anything large. Since $n^2 + 1 = (n+1)(n-1) + 2$, the divisor $n+1$ must divide $2$, which caps it at $2$ and leaves only $n = 1$. The rule to write down: when a growing quantity divides another, reduce until the remainder is a constant, then the divisor is bounded by that constant.

Find all $x \ne y$ with $x^y = y^x$ is the other common bounding miss. The answer is the pair $2$ and $4$, and students find it by trial. What they cannot do is rule out everything else, because they never ask which side grows faster. Taking logarithms turns the equation into $\frac{\ln x}{x} = \frac{\ln y}{y}$, and that function strictly decreases from $3$ onwards, so two different values at or above $3$ can never match. One of the pair must be below $3$, and checking $1$ and $2$ by hand finishes it. The rule: when you can find an answer but cannot close the list, look for a monotone quantity and use it to bound one variable.

Test it cold, later

You have revised the step when you can do it without the notes, on a problem you have not seen, some days after you wrote the rule. Not before.

1. Write the rule for each miss in one line, in your own words.
2. Three or four days later, take a fresh problem on the same topic and do only the first move, timed. Do not finish it.
3. If the first move was right, the rule has stuck. If not, the rule was too vague: rewrite it more concretely and repeat.

The delay matters. Testing the same afternoon measures what is still in short-term memory, which is the feeling of understanding that started this whole problem.

Where to get the fresh problems

The problem archive lists each problem with its topic, difficulty and expected time, so you can pull a cold problem on the topic you are fixing without seeing its solution first. The technique pages state each idea once, which is enough: you already understand them. And if your misses turn out to be mostly recognition rather than choice, how many practice problems a day explains why a mixed set trains that and a blocked one does not.