There is a piece of received wisdom about past papers that almost everyone shares. It goes like this: past papers are something you solve once your ability has matured. Touch them before your foundations are solid and you'll get crushed and lose confidence — and you'll have "used up" that precious first-sight attempt. So you save them carefully for the final stretch, and solve them as a last test of your ability.
This column replaces that wisdom. A past paper is something you open on day one of your studies. Saving it for a test of ability is, I believe, the single most wasteful way to use one — it throws away most of the value a past paper holds.
The reason is simple: before a past paper is a scoring device, it is the blueprint of the exam. What comes up, in what format, with what point weighting, and in how many minutes you'll be made to process it. Read that information first and then study, versus studying comprehensively without reading it — the destination differs even with identical study hours. It's the difference between walking with a map and walking without one.
Let me draw the boundary with the neighboring columns first. How to cycle through a problem set to make knowledge stick, I covered in the column on working through problem sets — that one is about comprehensive drilling to get knowledge in. How to spend the 60 to 180 minutes of the actual exam, I covered in the column on time allocation and marking mistakes — that one is about execution on the day itself. This column stands between them. It deals only with reading the examiner's habits and the point weighting, and converting the knowledge you've built into points on the day — the use of past papers during the preparation period. When to open them. How many years to work through. How many passes, and with what division of roles.
Open the most recent year as a "map" on day one — you don't need to solve it
The first step is almost disappointingly light.
- On the first day of your studies, put the most recent year's questions and answers in front of you.
- Don't try to solve them. Look through them, and write out four things on a single sheet of paper: the section structure (what comes up, and how many questions), the point weighting (which sections are heavy), the time limit (overall, and a rough estimate of what each section can afford), and the question format (multiple choice or written; testing knowledge or testing thinking).
- Pin that sheet where you can see it from your desk. Everything you study from here on, you check against this sheet.
While you're at it, if the organization running the exam publishes an exam outline or statement of scope, read it once on day one. It's the "legend" that goes with the blueprint.
→ Why it works: You're opening the paper for information, not for a score, so there is nothing to lose by failing to solve it. The received wisdom that "past papers before foundations will destroy your confidence" assumes you're using the paper as a scoring device. Used as a map, day one — when you still know nothing — is exactly when it's worth the most. Most of the waste in exam preparation comes from the mismatch of "I studied it but it never came up" and "it came up but I never studied it." Reading the blueprint first is one of the surest ways to shrink that mismatch before studying even begins.
How many years — benchmarks by exam type, cut off at the year the format changed
"How many years should I do?" has different benchmarks by exam type.
- University entrance exams: use 5–10 years for your first-choice school and 3 years for backup schools as the baseline. For many universities, commercial past-paper collections give you several years to a decade or so.
- TOEIC and Eiken: count in "administrations," not years. For TOEIC, each volume of the official practice-test books contains two full tests, so the number of volumes you own times two is your drill volume. For Eiken, the official site publishes the past several administrations. With official material, use every test you have.
- Professional certifications: use the most recent five administrations as the baseline. Many national qualifications and certifications publish their past questions, and lining up the last five makes the recurring patterns visible.
And one proviso common to all types: cut off at the year the format changed. Past papers from before a major change in system or format drop one rank in priority. The shift from Japan's Center Test to the Common Test (from 2021) and the TOEIC format change (2016) are the classic examples, and Eiken also changed its question format for some grades in the 2024 academic year. Check when your exam's "change year" was, as part of your day-one map reading.
→ Why it works: The core value of a past paper lies entirely in being "the same examiner, the same format as the real thing." An old year from across a format change has lost much of that value — still usable as material for checking knowledge, but thin as a map or a rehearsal. Weighting your first choice heavily and backups lightly is because the payoff of past-paper practice lies in adapting to that specific exam's habits, and investment in adaptation should scale with how much the school matters to you. Five years is the floor because with only one or two years you can't tell "it just happened to come up that year" from "it comes up every year." That floor, though, applies to your first choice only: for backup schools, three years is enough, because the goal there stops at confirming the format and the feel of the clock rather than full adaptation to the examiner's habits. Ten years is the ceiling guideline because going further back often means crossing a format change.
Three passes, three roles — the same "solving," but a different purpose each time
Before anything else, settle the inventory of the years in your hands. The most recent year is your day-one map. The one or two years just below it are set aside, out of the rotation, as the "unseen" stock for the dress rehearsal described later. The rotation covers everything that remains. The year you used as your map belongs in the rotation too — on day one you only read its structure, you never solved it. So if you can get seven years, the rotation is five — including the map year — with two ring-fenced for the rehearsal. The "five years" in the backward-planning example below refers to this rotation count.
Work through past papers on a baseline of three passes. But the three passes are not the same activity repeated. Switch the purpose each time.
- Pass one = measurement. Solve with no time limit. After the day-one map reading, digest one year at a time in parallel with your foundational studies. The goal is to measure the distance between where you are and the passing line, and your score rate per section. Not timing yourself is itself the point.
- Pass two = execution. Solve under the real time limit. This is the first time the stopwatch comes out. The order you solve in, the skip-calls, the review time you set aside up front — the exam-day designs from the time-allocation column get practiced here, before the day, until they're in your hands.
- Pass three = recovery. Solve only what you got wrong. Don't redo everything. Sort the questions you missed through pass two into three types — "knowledge gap," "misreading or jumping to conclusions," "ran out of time" — and crush them type by type.
→ Why it works: Run all three passes with the same purpose and, from pass two onward, "I can solve it because I remember the answer" contaminates everything — it stops being measurement and stops being practice. Switch the purpose each pass and the memory of the answers stops mattering: the value of pass two lies not in correct answers but in practicing how you spend time, and the value of pass three lies not in correct answers but in classifying how you go wrong. And pass one is untimed in order to separate "I could solve it given time" from "I couldn't solve it even with time." These two demand completely different fixes, yet in a single timed attempt they collapse into one undifferentiated "wrong."
"When to start" — decide by subtraction from exam day
When to begin serious past-paper work is decided by subtraction, not by mood.
Start date of pass two = exam day − (2 weeks of rehearsal) − (weeks needed for pass three) − (weeks needed for pass two)
Let's plug in numbers. Say you're doing five years. Pass two takes, per year, "the real time limit plus scoring and review (about the same again, up to double)" — half a day's work. If two years per week is a realistic pace, five years takes about three weeks. Pass three is errors only, one to two hours per year — about one week in total. Rehearsal takes two. Add it up and you need to be into pass two six weeks before the exam. Pass one sits outside this window — front-load it, starting from the day-one map reading, in parallel with your foundational studies.
And keep unseen years for the final two weeks. The one or two years just below the most recent one (which you used as your map) stay out of the rotation. But not for a "test of ability." They're for a full dress rehearsal of the day itself — solved start to finish at the same start time, on the same schedule as the real exam. Not a session for celebrating or mourning the score, but the final check that the time management you've built holds up on questions you've never seen.
→ Why it works: Without subtracting from the deadline, passes two and three become "if there's time left" slots — and they will almost certainly vanish. It's exactly the same logic as setting review time aside up front, from the exam-day column. And the rehearsal needs unseen years because on a year you've already solved, your reading speed and decision speed are inflated, and the load of the real exam can't be reproduced. First-sight freshness is the one resource in a past paper that is genuinely consumed by use — so ring-fence it for these two weeks alone.
The year-by-section score matrix — locate exactly where to go back to
Don't look at your pass-one and pass-two results as a total score. Make a table.
Years down the side, sections (or topics) across the top, score rates in the cells. Five years by five sections is a tiny table of 25 cells. Then read it by column — the same section, down across the years.
- One column is consistently low → a knowledge gap in that topic. Go back to the problem set. How to cycle through it once you're back there is exactly as written in the problem-set column: narrow down to the units you missed and rotate those.
- Columns look random, but only the later sections are low → not knowledge; you're running out of time. The place to go back to isn't the problem set — it's the time-allocation design of pass two.
- Only one year's row is low → most likely that year was simply hard. Don't mistake it for a topic weakness.
Keeping the destination concrete to the very end is what removes the hesitation. If, say, on an English exam the grammar-and-usage column runs low across the years, the place to go back to is unit-level grammar drilling. passed.jp's English problem set has question pages organized by grammar unit — from the be-verb through the passive voice, the present perfect, and relative pronouns — so you can narrow down to just the units behind the low column and drill them intensively. From the low column, to the unit, to focused unit-level drilling: the table connects in one straight line to the place you return to.
→ Why it works: A single sitting's score carries the noise of "how you happened to do that day." Line up the same section across years and the noise averages out, leaving only the structural weaknesses in view. And whether the weakness is really "knowledge" or really "time" completely changes where you should go back to. Return by feel, without the matrix, and you get misdirected effort — stacking knowledge work onto a section you were losing to the clock. The five minutes the table takes is five minutes that prevents that entire misfire.
The three things you must not do
To close, three classic failures in past-paper use, each with its reason.
- Saving them until the final stretch. The most wasteful of all. The longer you study without a map, the more investment piles up in material that won't come up — and if you open the paper late and discover the format doesn't suit you, no time remains to change course. Besides, a past paper doesn't disappear when you solve it. The feeling of "using it up" comes from seeing it only as a scoring device. The one thing that is consumed is first-sight freshness, and that's protected by ring-fencing one or two years for the rehearsal.
- Cycling the same year until you've memorized the answers. The purpose of the passes is recovering your error types, not memorizing correct answers. A high score with the answers memorized is indistinguishable from real growth — meaning the past paper is broken as a measuring instrument. Once pass three has crushed the error types, that year can graduate. From the fourth pass on, I believe the only thing that grows is déjà vu, not your score.
- Solving from the oldest year forward. Driven by the urge to "save the newest for later," working oldest-first means you hit the years closest to the real exam's current tendencies at your most exhausted, most time-poor stage. Flip the order. Start from the newest year and work backward, stopping at the format-change year. The only exception is the one or two unseen years ring-fenced for the rehearsal.
Closing: the waste isn't in solving them — it's in starting without reading them
Let me compress the whole design into one line each. Open the most recent year as a map on day one. Decide the year count by exam type and the format-change year. Split the three passes into measurement, execution, and recovery. Set the start date by subtracting from exam day. Locate where to go back to with the matrix. Don't hoard, don't memorize, and solve newest-first.
What's wasteful about past papers is not solving them early. It's starting to study without reading the blueprint. An examiner's habits are like terrain — they persist across the years — and a past paper is the one map surveyed directly from that terrain. The earlier you read the map, the more readily the same study hours convert into points.
The name passed.jp has "passed" built into it. I hope your day one begins not at page one of a problem set, but at the moment you open a past paper as your map. From that day on, the total amount of wandering in your studies goes down.
