What a wrong answer costs you on WaniKani

What a miss costs

WaniKani publishes the exact rule. The drop depends on how many times you missed the item in the session and on the stage it was sitting at when you missed it.

Two consequences fall out of it. A single miss at Guru or above costs two stages, so a Guru II item lands back at Apprentice IV and owes you a fresh 47-hour wait plus a 7-day one to get back where it was. And a miss at Apprentice I costs nothing at all, because the floor catches it.

Why your queue never runs backwards

The going-backwards claim comes from a reasonable-looking piece of arithmetic. A correct answer is worth +1 stage and a miss at Guru costs 2, so the expected movement per review is $3a - 2$, which hits zero at an item accuracy of two thirds. Below that, the reasoning goes, you are losing ground.

The arithmetic is right and the conclusion does not follow, because that drift is not the same at every stage. Below Guru a miss costs one stage, not two, and at Apprentice I it costs nothing. At 60% item accuracy the picture looks like this:

Guru and above is losing ground. Everything below it is gaining. Items pile up in the region where the drift is positive, climb out of it, fall back into it, and climb out again, and every item still reaches Burned eventually.

There is a shorter argument that settles it without any simulation. Apprentice I to Burned is a gap of eight stages. Every promotion is worth exactly +1, and every demotion costs at least 1. So over the life of any item that finishes, the number of promotions has to exceed the number of demotions by at least eight, at every accuracy there is. Run it at 50% item accuracy and the average item takes 89 promotions and 67 demotions to burn. Lopsided, slow, and still forward.

The bill arrives as reviews

Eight consecutive correct reviews take an item from a fresh lesson to Burned. That is the floor, and it is the same floor for everyone. Accuracy decides how far above it you land.

The curve is gentle at the top and vicious at the bottom. Going from 90% to 85% item accuracy costs you about two extra reviews per item. Going from 70% to 60% costs you thirty.

What that does to a day

Reviews per item converts straight into reviews per day. A queue fed at some rate, whose members stay for some length of time, holds the product of the two. Each item you start eventually generates its full review count, so a steady lesson pace produces a daily load equal to the pace times the reviews each item costs.

At 15 lessons a day, held steady, that lands here:

This is the mechanism behind the wall people hit somewhere in the twenties. Nothing changed about the lessons. The accuracy slipped, every item started costing more reviews, and the same daily habit stopped being enough to clear the day's queue. The pile does the rest on its own.

Where two thirds is a real number

The two-thirds figure does describe something true, just not a reversal. Below it, items that reach Guru tend to fall back out of it before they climb further, so your Apprentice pile is being fed from above as well as from your lessons. Guru stops being a milestone and becomes a place items visit.

That is what treading water feels like from the inside, and it is why the fix is rarely "study harder". The items bouncing off Guru are a small, identifiable set. They have a name.

Leeches are the items doing this to you, and they are worth attacking individually rather than by grinding the whole queue harder.

Your accuracy percentage flatters you

One catch sits underneath all of the above. Everything so far has been about item accuracy, meaning the share of reviews you clear without a single miss. The percentage WaniKani reports is a different quantity: it counts questions, and most items ask two of them.

Retries make the gap wider than it first looks. Miss a question and WaniKani puts the item back in the session, so you answer it again, and every attempt lands in your review statistics. A completed review always ends with one correct answer per question, because that is what completing it means, which gives an exact relationship rather than an estimate.

Because $m$ varies by person, the conversion is a range rather than a single number. An item accuracy of two thirds reads as anywhere from 75% to 85% depending on how many wrong answers your misses generate. You may see $\sqrt{2/3} \approx 82\%$ quoted for this, from a model that assumes meaning and reading fail independently and ignores retries entirely. Both assumptions push the number the wrong way.

What to aim for

Reported accuracy in the high 80s is where the cost per item stops compounding. At 90% an item takes about 12 reviews to burn against a floor of eight, at 85% it is 17, at 80% it is 30. There is no cliff anywhere in that range, so treat a dip as a bill arriving rather than a failure state.

The lever is almost never accuracy itself. Accuracy is downstream of how many unfamiliar items you are holding in Apprentice at once, which is downstream of your lesson pace. Cutting lessons to 5 or 10 a day drains the pile, accuracy recovers because you are seeing fewer strangers per session, and the review cost per item falls with it. The lessons per day guide has the pacing side of this in detail.

The number worth watching is not the percentage. It is what your finished items cost. If your burns are averaging 12 reviews you are paying a 50% surcharge on every item you own; if they are averaging 25 you are paying triple, and that is worth acting on well before the queue makes the point for you.