When to reset on WaniKani (and how)
Level 17, back from a break. Level 27, two years away and 2,000 reviews waiting. Level 24, reading other people's reset stories before deciding. The forum has produced this same thread on a steady clock since the reset feature shipped in 2017, and everyone asking treats their case as unusual.
The replies treat it as settled, and they converge on two tests: how much still comes back in a cold review session, and whether you can face opening the app at all.
Start with the tree
- If you have been away less than 2 weeks: just catch up gradually.
- If you have been away 2 or more weeks but can recall 70% or more of items at sight: push through and lower the pace.
- If you cannot recall 70% at sight but feel confident 5 to 10 levels back: reset 5 to 10 levels back.
- Otherwise: reset to level 1 or 5.
What a reset deletes, and what survives
A reset rolls your account back to any level at or below your current one. The API reference states the effect in one sentence: assignments and review statistics "at that level or higher are set back to their default state." At that level or higher, so the target level is wiped too. Reset to 17 and your finished level 17 lessons go back in the queue. Guides that promise your target level survives, including an earlier version of this one, have it backwards.
Every level below the target keeps its SRS timing untouched. User synonyms and custom notes survive as well, per the knowledge base; they are just out of reach until the items unlock again. Your subscription is unaffected either way. Everything at or above the target is gone for good, and the staff announcement that shipped the feature is blunt about it: "no one can get your data back for you. It's gone forever, even to us."
One claim in reset threads does not hold up: that the second pass re-burns items in a fraction of the time. Lesson to burn is about 174 days of waits at minimum, and knowing an item already shortens none of them. What moves fast the second time is you, not the schedule. Returning learners routinely re-clear ten levels in two months at 95% accuracy, and their burn dates still arrive when the table says.
The giant-backlog fork
Months or years away, and a queue in four digits: this is the one place the community splits for real, and both camps have documented wins.
The no-reset playbook is How to kill a giant review backlog: turn on vacation mode so the pile stops growing, drill the material outside the SRS until it tests well again, then turn vacation mode off and clear the queue. Its author cleared a seven-month backlog in about two and a half weeks that way and was back to full speed in six. The case for it is that you probably have not forgotten as much as the queue size implies. The reset camp's wins are real too, and tend to be framed as morale decisions rather than memory ones.
What we would do: sit 50 reviews cold before deciding. If well over half come back, a few weeks of self-study is cheaper than re-earning levels you still own. If the sample comes back mostly blank, or if dread of the pile is what has kept you away for months, reset. What still tests well decides it; morale breaks the tie.
When the answer is no
A rough week is not a reset reason. A 200-300 review pile from a busy stretch clears in days: cap each sitting at 25 reviews no matter how many are waiting, take no new lessons until the queue reads zero, and accept one bad-accuracy week on the way through.
Being high level and scared of forgetting is not one either. Resetting from level 17 costs at least four months during which you learn nothing new, and the deeper you are the longer that dead zone runs. Missing some Enlightened items and relearning them through failed burn reviews is the system working as designed.
Accuracy panic needs calibrating before it needs a reset. WaniKani publishes no accuracy target, and the honest range is wide: 60-80% is common, 80-90% is solid, 90%+ is aspirational. If your sessions sit in the 50s and 60s on current material, pause lessons and self-study first. Reset is the fallback for when weeks of that change nothing.
How deep to go
There is no standard depth. Real resets run from a single level to all the way back to 1, from starting points as high as 43, and how far you go tracks how much you still recognise and nothing else. The most concrete rule in circulation is to walk back through any level where more than 70% of the items are unrecognisable. Ours is stricter, and is opinion rather than consensus: reset to the deepest level where you would trust yourself to pass a cold session at 90% or better.
Err deeper rather than shallower, but only slightly. Reset too short and you rebuild on shaky levels, then face a second reset two weeks later. Overshoot and you spend months waiting on material that was never in doubt. One or two levels below your comfort line is margin; ten is a waiting room. If you have review history, let it decide: Wanilog graphs your accuracy by level, and a visible cliff usually answers the depth question outright.
Pressing the button
- On wanikani.com, open Settings and find the Danger Zone.
- Pick the level to reset to.
- WaniKani emails you a confirmation. The reset runs when you press the button in that email, not before.
The two-step flow is deliberate, built that way "so nobody ends up resetting their account by accident." The API even records the two moments separately, a created_at when you ask and a confirmed_at when you mean it.
The second climb
A reset pays off only if the second climb runs differently. Cap lessons for the first month; five a day is plenty. Spend the freed attention on never missing a day, since the 4-hour and 8-hour Apprentice waits make a morning session plus an evening one the natural shape. And the next time life interrupts, turn on vacation mode before the pile forms rather than after.
After the reset, Wanilog files your previous runs under historic resets on the level-up page, so your finish-date forecast follows your current pace instead of averaging in the era you just deleted.