Every page that promises the worst time to visit Japan seems to lose its conviction on the way down. The warnings come first, the dates come vaguely or not at all, and the final section takes it all back: every season has something, there’s no truly bad time, go whenever you can. That ending is why you had to keep searching. This page is built not to do that. Six windows in the Japanese year can do real damage to a first trip. They have real dates, the dates are just below, and this page still means them at the last line.
One honest boundary before the list, because half the people who land on a page like this actually want its mirror image. This is the avoidance half of the timing decision, at full depth — nothing here will tell you where to aim. The aiming half — which months to target, how the seasons compare, what to do when your dates are already fixed — is the job of the companion guide, When to Visit Japan: The Months to Aim For — and the Months to Avoid. If that’s what you came for, go there now. If you want the danger map alone, with nothing softened at the end: keep reading.
The six windows, dated
No build-up. Here is the entire answer in one box, in calendar order; the rest of the page explains each line and then ranks how bad they really are.
The six avoid-windows — basis marked per line (statute, JMA 1991–2020 normals, or custom); each is treated in full below.
- December 29 – January 3, the statutory closure core (government and administrative offices, by law), with customary commercial closures spreading wider on both sides — the week the country shuts.
- ~March 22 – April 4 by the JMA city normals (Fukuoka-to-Osaka flowering through full bloom; recent years running earlier) — the blossom crush on the first-trip cities.
- April 29 – May 5, the statutory Golden Week frame — the days the whole country travels at once.
- ~June 7 – July 19 by the Tokyo-area rainy-season normals — regionally staggered: Okinawa ~May 10 – June 21, and Hokkaido has no official rainy season — the quality-tax weeks.
- Around August 13 – 16, customary Obon, anchored by the statutory August 11 holiday — the homecoming peak.
- August – September, the typhoon formation and landfall peak by the JMA normals, inside a wider May–October landfall-risk window — the one window that can rewrite an itinerary for you.
Why these weeks go wrong: one calendar, everyone on it
Four of the six windows run on the same engine, and it isn’t tourism. Japan’s time off is synchronized. National holidays aren’t spread across the year for different people — they land on fixed dates for essentially everyone at once, so the country doesn’t take its breaks in shifts. It takes them together.
Anyone who lives on this calendar plans by it without thinking about it. Whether the big weeks will be crowded is not a question a resident asks; which days the whole country will be moving is simply known, the way weekdays are known, and personal plans get built around that certainty — book far ahead, or sit those days out. What reads from outside as a travel warning is, from inside, just the shape of the year.
Two consequences follow for a visitor. First, movement: when the same days off arrive for everyone, long-distance travel compresses into them — the same departures out, the same returns back. Second, closure: the same synchronization that moves the country also shuts parts of it, on fixed dates, at the year’s turn. And there is no separate country reserved for tourists. The trains a first trip rides, the rooms it sleeps in, the famous places it queues for — these are the same trains, rooms and places the national load lands on. A first-time visitor inside a synchronized week isn’t sharing Japan with other tourists; they’re sharing it with Japan.
The remaining two windows — the rainy weeks and the typhoon window — run on a different engine entirely: weather structure, not law. Their failure modes are different too, and their entries below treat them differently, because pretending all six windows are the same kind of bad is its own way of being wrong.
The calendar of windows to avoid — in full
Each window below follows the same template, deliberately: the dates first, with their basis; then the mechanics of what actually happens; then what it does to costs; a verdict; and the window’s edges — where it genuinely begins and ends, which is where most planning mistakes live. Calendar order, starting at the turn of the year.
Late December to early January — the country closes
The dates (basis: statute, plus custom): Government and administrative offices close December 29 – January 3, fixed by law (Act No. 91 of 1988) and recurring every year; January 1 is itself a national holiday. Commercial closures — shops, restaurants, museums, attractions — are customary, not statutory: each operator chooses its own days, and the customary spread reaches wider than the legal core on both sides. Only the statutory core has exact, guaranteed dates.
What makes this window unlike every other on the page is that its central mechanism is closure — and closure with a blur. The statutory core tells you exactly when the offices shut; it cannot tell you which restaurants, which museums, which small places you crossed the world for will be dark, or for how many days, because those decisions belong to each operator and are published nowhere as a single list you could check. A trip routed through this window becomes a sequence of doors that can’t be verified in advance, in the one week when the closed ones cluster.
The cost consequence is quieter than the spring windows’ but real: you pay full trip costs — the flights, the nights, the days of leave — for a country running at partial capacity. Nothing discounts itself for being closed. Structurally, this window sells you less Japan for the same trip.
Verdict: the easiest no on this page. A first trip runs on the assumption that the country is open for it, and this is the one week where that assumption fails — by law at the core, unverifiably at the edges.
The edges are asymmetrical in certainty rather than in date: the core is exact, and the customary spread around it is not, and nothing will firm it up for you before you arrive. The honest planning posture is to treat the window as wider than its six statutory days, and to assume the closures thin — not vanish — as you move away from the core in either direction.
Late March to early April — the blossom crush
The window (basis: JMA station normals, 1991–2020 — city-anchored; there is no national date): flowering opens Fukuoka ~March 22, Tokyo ~March 24, Kyoto ~March 26, Osaka ~March 27, with full bloom about a week behind — Tokyo ~March 31, Kyoto ~April 4, Osaka ~April 4. The wave keeps moving north after the first-trip cities: Sendai April 8 → April 13, Sapporo May 1 → May 6. (Naha’s flowering normal is ~January 16 — a different species and a different season.) Recent years have run earlier than the normals: Tokyo’s 2026 flowering came on March 19, full bloom March 28. Normals are thirty-year averages, not promises for your year.
The mechanics here are demand mechanics, and they’re brutal in a specific way: a worldwide, year-long anticipation is aimed at roughly two weeks, in a handful of cities, on dates that cannot be guaranteed in advance. Because nobody can book the bloom itself, everyone books the window — defensively, early, and at the same time. Domestic and international demand land on the same few cities on the same few days, and the uncertainty of the exact dates makes the crowding worse, not better, because people widen their bookings to cover the risk.
The cost consequence follows without a single price needing to be quoted: a short, fixed supply of days meeting the year’s most concentrated demand produces the year’s earliest sell-outs and its heaviest booking pressure. This is the window where planning late doesn’t mean compromise; it means exclusion.
Verdict: if the blossoms are the point of your trip, this window isn’t a trap — it’s a bill, and you pay it with open eyes and early bookings. If they’re not the point, these are the most crowded and most contested days you can choose by accident.
The edges are city-edges, and the dated block above is boundary precision, not viewing advice: the crush on the first-trip cities has a beginning near the flowering normals and decays after full bloom — and because recent years run early, the dangerous edge is the front one, where a booking aimed at the “safe” start of the window can land inside it. The northern dates mark where the first-trip cities’ window ends; they are not a recommendation to chase the wave.
April 29 to May 5 — Golden Week, when the whole country travels at once
The frame (basis: statutory holiday law, recurring annually): four national holidays anchor the window — April 29 (Showa Day), May 3 (Constitution Memorial Day), May 4 (Greenery Day) and May 5 (Children’s Day). May 1 and May 2 are not holidays. The frame is April 29 – May 5, every year.
The 2027 layout (basis: the officially fixed 2027 calendar): April 29 falls on a Thursday, isolated — April 30 is a regular workday — and May 1 – 5 form five consecutive days off (Saturday, Sunday, then the Monday-to-Wednesday holidays), with no leave required and no substitute holiday arising in the window.
Golden Week is what happens when the synchronized calendar reaches its maximum. Four legal holidays inside seven days convert a working spring week into the country’s largest shared vacation, and the earlier section about everyone sharing one calendar stops being an abstraction here: this is the week it describes.
The mechanics are total in a way no other window matches. This is the closest the year comes to every demand stream arriving at once, and the whole movement points at the same finite set of long-distance trains, city rooms and famous places that a first-trip itinerary is made of. Inside this frame you are not a traveler among travelers; you are a guest inserting yourself into a nation’s own vacation, at its peak, competing for its infrastructure on its biggest travel days.
The frame never moves; the shape inside it does. How the fixed holidays fall against the weekends decides, each year, whether the week runs as one long unbroken block or as separated pieces — and that layout, not the frame, determines where the heaviest travel days land. This is why a generic “late April to early May” warning isn’t enough for real planning: if your candidate year is 2027, the layout in the dated block above — not the generic frame — is the planning fact.
The cost consequence is the purest demand pricing on the calendar: dates fixed by law, known years ahead, met by the country’s own maximum willingness to travel. Everything a trip consumes is at its most contested exactly here, and no fare table is needed to say so — the mechanism guarantees it. Booking early doesn’t defuse this window; it only decides whether you compete with a reservation in hand or without one.
Verdict: the worst week of the year to be a first-time visitor to Japan — not the most dangerous, not the most uncertain: the most completely taken. Whatever a first trip is for, none of it is easier, cheaper or better inside this frame.
The edges are the hardest on the page, and that is the one mercy: the frame is bounded by statute, workdays resume on schedule, and the country’s week snaps back quickly on both sides of it. The soft part isn’t the boundary — it’s the interior, where each year’s weekday layout decides which days inside the frame carry the heaviest load. Treat the frame’s edges as real, and check the layout for your actual year.
June to mid-July on the main-island spine — the rainy weeks (regionally staggered)
The windows (basis: JMA rainy-season normals, 1991–2020; entries and exits are declared retrospectively, to about five days’ precision): Okinawa ~May 10 – June 21 · northern Kyushu ~June 4 – July 19 · Tokyo area ~June 7 – July 19 · northern Tohoku ~June 15 – July 28. Hokkaido has no official rainy season. Individual years wander around these averages — northern Kyushu’s 2025 entry arrived about three weeks early, on May 16. There is no single national rainy season; there are regional windows, and they don’t line up.
This entry earns its place on the list by a different route, so it deserves precision about what it is and isn’t. Nothing statutory happens in these weeks. No synchronized movement, no closures, no demand spike — of the six windows, this is the only one whose problem contains no other people. The trap is the character of the weeks themselves: a structural rain window sits over the main-island spine for roughly six weeks, and the only plannable fact is that the window exists. Which of its days actually rain, and how hard, is not knowable at booking distance — even the window’s own boundaries are only declared after the fact, to about five days’ precision.
And here the honest cost statement is a deflation: there isn’t much of one. The demand convergence that punishes the holiday windows is absent, which structurally removes the pricing pressure this page keeps warning about elsewhere. What this window takes from a trip isn’t money; it’s quality — outdoor hours that count for less, days bent around rain.
Verdict: the mildest window on this page, kept on the page anyway. It is a tax on trip quality, not a barrier to a trip — and calling it more than that would be exactly the exaggeration this article exists to refuse.
The edges are regional before they are calendrical, which is why the dated block lists places instead of one range. The same calendar week can sit inside the window on the spine, past it in Okinawa, and outside it entirely in Hokkaido — route changes what these weeks mean more than dates do. And because entries and exits are declared retrospectively, every boundary in the block is soft by nature: this is the only window on the page whose own edges are officially fuzzy.
Mid-August — Obon, the homecoming peak
The dates (basis: statute for the anchor, custom for the window): the statutory holiday is August 11 (Mountain Day). Obon itself — the mid-August homecoming — appears nowhere in the holiday law: it is customary, typically taken around August 13 – 16, and company shutdown windows vary by employer and by year. Treat August 13 – 16 as the customary center, never as official dates.
Obon is the movement trap in its purest form. Where the New Year window closes the country and Golden Week vacations it, Obon relocates it: on the customary days the country travels toward family and then back again, and in combination with the statutory anchor just ahead of them, those days produce Japan’s mid-August domestic travel peak. The load lands hardest on exactly the thing an itinerary can least reroute around — long-distance transport and the accommodation along it, on days the whole country has privately agreed on.
Unlike Golden Week, none of this is law. Workplaces aren’t legally shut; the country has simply converged, employer by employer, on approximately the same days. That makes the window’s center dependable and its edges genuinely unknowable from outside — there is no calendar you can consult to learn where this year’s edge falls.
The cost consequence tracks the custom: demand converges on the customary center and thins with distance from it — peak-demand behavior on transport and rooms around those days, easing at the margins. Structural, predictable, and centered on dates the law never printed.
Verdict: treat the center of this window as being as hard as Golden Week, because while it lasts, it is. The difference is all in the edges — and the edges are precisely the part you cannot verify from abroad, so plan against the wider window, not the four famous days.
Soft edges, hard center — that is the shape to plan by. The customary core is dependable; where a given year’s shoulder days fall depends on employers, not statutes, and shifts year to year. The margin of this window is real but unmappable in advance: give it room on both sides rather than betting on a clean boundary.
August and September — the typhoon window
The numbers (basis: JMA typhoon normals, 1991–2020): about 25 typhoons form in the region per year; about 3 make landfall in Japan; formation and landfall both peak in August–September. May through October is the landfall-risk window — a description of when landfall risk exists, not a forecast that a storm will meet your dates.
This window belongs to a different category from everything above it, and the category matters more than the months. The other five windows crowd a trip, price it, close it out of places, or dampen it; this is the only one that can take scheduling authority away from you. When a typhoon crosses a route, the days it touches stop being plannable — movement across the affected region stops being reliable, and the itinerary you wrote is suspended in favor of the one the weather writes.
The frame has to stay probabilistic, in both directions. Around three landfalls in an average year — even concentrated into the two peak months — is a country-level normal, not a trip-level fate: the verified numbers describe Japan’s year, not your week, and nothing in them says a storm will find you. What they do say is that choosing these two months means choosing the part of the year where that possibility sits at its annual maximum. This is a different kind of decision from every other window on this page, where the damage is certain but bounded; here it is unlikely but unbounded by your preferences.
The cost consequence is shaped the same way: not a pricing window at all, but a contingency cost — the one entry on this page whose bill, if it arrives, arrives all at once, as rebooked days and rerouted plans rather than as higher prices. The odds favor never paying it. The mechanism means you don’t get to choose when you do.
Verdict: the only window on this page that can overrule you outright — and also the only one that usually doesn’t fire at all. Read it as a probability with force behind it: not a reason for fear, but a reason for slack in the plan and respect for the two peak months.
The edges here are gradients, not lines. Risk exists across the wider window in the block, concentrates hard in the two peak months, and thins on both sides of them — the tail of the window still carries the tapering remainder. No single date inside it is safe or unsafe by itself; the window describes odds, and odds move smoothly, not at a boundary.
Not all six are equal — how bad each one really is
Strip the dates away — they’re all above — and what’s left is the question the compressed box couldn’t answer: how bad is each window, really? A list that scores all six the same is quietly lying in both directions at once — overselling the mild ones, underselling the brutal ones. Here is the severity map this page actually stands behind, in four tiers.
| Tier | Window(s) | What the tier means |
|---|---|---|
| Certain and total | New Year · Golden Week · Obon’s center | Fires every year, on schedule, at full force. No booking strategy changes what the week is. |
| Certain but purchasable | The blossom crush | Fires every year — but the damage is contention and cost, so enough lead time and budget can buy a path through it. |
| Probabilistic but forcible | The typhoon window | Usually doesn’t fire on any one trip — but it is the only window that can take days from you by force when it does. |
| Certain but survivable | The rainy weeks | Fires every year and wrecks nothing. It degrades a trip; it doesn’t deny one. |
On the top tier: these three are what the word “avoid” was coined for. They are certainties, they are total while they last, and a first trip spent inside them buys less country for more money with more friction at every step. Nothing a first-timer wants from Japan gets better during these weeks, and almost all of it gets worse.
On the blossom crush: the one window whose severity depends on why you’re coming. For the traveler who merely wandered into those weeks, it behaves like the top tier in everything but name. For the traveler who chose them for the bloom, it downgrades to an expensive ticket — still crowded, still contested, but bought deliberately, which changes what it is without making it cheap.
On the typhoon window: refuse to average it. Calling it fine because it usually misses, and calling it terrible because it sometimes doesn’t, are both wrong in the same lazy way. It is a small chance of a forced rewrite, sitting at its annual peak — hold both halves of that sentence at the same time.
On the rainy weeks: the bottom of the map, and deliberately so. Putting them in the same breath as Golden Week is how worst-time pages turn into clickbait, and the deflation here is the honesty: a quality tax is worth avoiding when you can choose, and survivable when you can’t — both true at once, neither one a reason to move it up a tier.
That’s the severity map, and this is where the page stops — with all six windows still standing and none of them talked back down. What you’re holding is half of a decision: the half that says where not to land. The aiming half — which months actually deserve a first trip, how the seasons trade off, and what to do when your dates are locked — is the companion guide’s job: When to Visit Japan: The Months to Aim For — and the Months to Avoid. Take the map with you. It keeps exactly this shape when you leave.