Why you keep running out of time on the AMC

Everybody describes this the same way: I know how to do the questions, I just run out of time. That sentence contains a diagnosis, and the diagnosis is usually wrong. It says the problem is speed, and speed is almost never what happened.

Three quite different failures all end with the same symptom, and they need three different fixes. The good news is that a single past paper tells you which one is yours, because each leaves a distinct fingerprint in the arithmetic.

The clock is a budget, and it is exactly this big

The paper is 25 questions in 75 minutes. That is $75 \times 60 = 4500$ seconds, so the average question is allowed 180 seconds. Three minutes, and not one of them is spare.

Read that as a loan rather than as a speed limit. Every question you spend four minutes on has borrowed sixty seconds from a question you have not read yet, and the loan is fine as long as something repays it. Running out of time means a loan went bad. The whole diagnosis is finding out which one.

Cause one: a single stall

Go and get your last timed paper. Count how many questions you actually answered before the clock beat you. Suppose it was 18.

Eighteen questions at the budgeted 180 seconds is 3240 seconds, or 54 minutes. You had 4500. So 1260 seconds, 21 minutes, went somewhere that is not on your answer sheet. That is 28% of the entire paper spent on questions you did not end up answering.

Nobody spends 21 minutes evenly. It is one question, occasionally two. You saw a route, the route needed one more step, and each next step felt closer than the last. That is the fingerprint: a large unexplained remainder and a clear memory of which question ate it.

Cause two: a uniform drag

The second fingerprint is the absence of the first. You answered 20, you remember no disaster, and every question felt normal.

Take a fifth longer than budget on everything and you spend $180 \times \tfrac{6}{5} = 216$ seconds a question. Divide 4500 by 216 and you get through 20 of them. Five questions never read, and no single moment where anything went wrong.

This one hides because it feels like working carefully. It is usually three habits stacked: solving forwards when the answer choices were right there, computing an exact value when the question only needed a size, and re-checking arithmetic you already checked. Fixes two and three below are aimed squarely at it.

Cause three: you meet the paper in its order instead of yours

The questions rise in difficulty, but that ordering is the average over everybody who sits the paper, not over you. Your personal wall is somewhere in the middle and it is not where the paper thinks it is.

Price the difference. Suppose question 9 is your wall. You give it fifteen minutes, you solve it, and as a direct result you never read questions 16 to 20. You banked 6 points for the wall and $5 \times 1.5 = 7.5$ for five blanks, so 13.5 from those six questions.

Now skip it. Question 9 goes blank for 1.5, you reach the last band with time in hand, you answer three of them and leave two: $1.5 + 3 \times 6 + 2 \times 1.5 = 22.5$.

Nine points, and note that in the first version you won the fight. Stalling costs you nine points even when the stall succeeds.

Fix one: an abort rule with a number in it

Every question gets two checkpoints, and both are clock times rather than feelings. At 45 seconds, ask whether you have a route; if you do not, you are guessing at a route, which is the expensive kind of guessing. At 6 minutes, twice the budget, you leave, no negotiation.

The cap is affordable, which is the point. Three questions run all the way to it and you have spent $3 \times 360 = 1080$ seconds, leaving 3420 for the other 22, which is 155 seconds each. Two and a half minutes a question, and you finish the paper.

Compare the uncapped version of the same day. Two questions run 14 and 11 minutes rather than 6 and 6, so 25 minutes instead of 12. The 13 minutes of overrun, at the 90 seconds an early question actually takes, is 8 questions of reading time you no longer have.

To make the cap real you have to write the start time of each question in the margin. Nothing else works, because the whole failure mode is that time does not feel like it is passing.

Fix two: bound the answer instead of computing it

Half the drag in cause two comes from answering a harder question than the one on the paper. You are not asked for a value. You are asked which of five.

Take the greatest integer below $\sqrt{2026}$. There is nothing to compute: $45^2 = 2025$ and $46^2 = 2116$, so the root sits between 45 and 46 and the answer is 45. That is ten seconds, and it never produces a decimal.

The move generalises further than estimation questions. Parity, a rough size, an order of magnitude, one crude inequality: each of them kills options without ever producing the value. Bounding and squeezing is the habit of trapping a quantity between two walls, and the walls are usually far cheaper to build than the quantity.

Fix three: work backwards from the five options

A two-digit number has digits summing to 11, and reversing it makes it 27 larger. Forwards, that is two equations in two unknowns and a minute of algebra.

Backwards, it barely qualifies as work. Only eight two-digit numbers have digits summing to 11, and 47 is the only one where the reversal gains exactly 27, since $74 - 47 = 27$. If the paper hands you five options, you never need the other three. Working backwards from a testable option is the version of this exam that the format is actually asking you to sit.

The two fixes are the same idea twice, incidentally. Reversing the number swaps its digits, so the gain is $9(b - a) = 27$, giving $b - a = 3$ immediately. Bounding and substitution both work because a multiple-choice paper rewards eliminating, not deriving.

The pacing you are aiming at

Three minutes a question is an average, not a plan. A workable plan spends the early questions fast and buys the middle with the change: ten questions at 90 seconds is 15 minutes, ten more at 4 minutes is 40, and that leaves 1200 seconds, 20 minutes, for the last five.

Two checkpoints are worth memorising, because they are the only two you can check without arithmetic under pressure: question 10 by minute 15, question 20 by minute 55. If you are behind at the first one you are dragging, and if you were on time at the first and behind at the second, you stalled.

If your target does not need the last five questions at all, that 20 minutes redistributes across the ones you are answering, and the one-month AMC 12 plan works that trade out in full.

What to do on the next paper

Sit one paper with a pen and the margin discipline: the start time of every question, written down, all 25. Then mark it twice, once for correctness and once for time, and read the second marking first. You will find one of the three fingerprints above, and you will stop guessing which one you have.

The diagnostic is the fastest way to find the band where your wall actually sits, which is the number cause three depends on. The daily problem posts a new one every day in three tiers and needs no account, which is enough to keep an abort habit warm between full papers. And the technique pages and the problem archive are where the two fixes above stop being advice and start being practice.