The Sweet Spot: How Hard Should It Be?

Think about the last thing you quit not because it was too hard, but because it was too easy. The language app still drilling you on words you learned months ago. The training course that spent an hour on the thing you already do for a living. You didn't rage quit. You drifted, which is quieter and somehow worse, because there was never a moment to point at and call the end.

Now think about the opposite. The textbook that lost you by the second page. The pickup game where you dropped every point until you stopped turning up. That one stings differently, but the outcome is the same. You are gone.

Somewhere between those two failures is a band where you actually stick around and get better. The interesting claim, and the reason for the rest of this piece, is that the band is narrower and more precise than most of us assume. There may even be an optimal difficulty, a setting you can aim at.

Boredom on One Side, Panic on the Other

The two ways to lose a learner sit on opposite ends of the same dial. Turn the difficulty too low and you get boredom: nothing is at stake, attention wanders off, and the task becomes a chore you finish without really being there. Turn it too high and you get something closer to panic: the gap between what is being asked and what you can do is wide enough that effort feels pointless, so you protect yourself by checking out.

The comfortable middle, where challenge sits a notch above your current skill, is where you actually stay and improve. Most of us have felt it and filed it under a good day. The harder question is whether you can engineer it on purpose, and for that you would need to know where on the dial it lives.

Someone Calculated the Optimal Difficulty

In 2019, a group of researchers led by Robert Wilson tried to pin that band down with math rather than intuition. Writing in Nature Communications, they asked a clean question. If you could set the difficulty of practice anywhere you liked, what setting would make a learner improve the fastest? For a broad class of learning algorithms, the ones that inch toward the right answer by nudging themselves a little after each mistake, they derived an exact answer.

The answer was about 85 percent. Tune the difficulty so the learner is right roughly 85 percent of the time and wrong the other 15 or so, and it learns faster than at any other setting. They named it the Eighty Five Percent Rule.

There is something satisfying about a soft idea like "not too easy, not too hard" resolving into a specific figure. It is the difference between "season to taste" and an actual measurement. But the figure comes with fine print, and the fine print is the whole reason to trust it rather than to slap it on a poster.

A Result From Simulations, Not a Room of Students

Here is the part most people repeating the rule leave out. Wilson and colleagues did not sit a few hundred students down and vary how hard their quizzes were. They proved the result mathematically and then confirmed it in computer simulations: an artificial neural network learning to sort inputs into two boxes, and a more biologically plausible model of the way animals learn to tell signals apart. No human beings were trained in the study at all.

That matters, and the authors are the first to say so. In the paper they note that despite how intuitive the idea feels, no formal work had actually tested how training accuracy affects learning, and that running such a test is an important direction for future work. So the cleanest statement of the rule is this. For a certain family of learning systems, 85 percent is provably the fastest setting, and whether human learning obeys the same number in the messy real world is a question the paper deliberately leaves open.

This is more interesting than "science proves you learn best at 85 percent." It is a precise, testable formalization of an instinct that teachers and coaches have worked from for a very long time.

An Old Idea, Finally Pinned Down

The instinct is old, and the researchers knew it. Wilson and colleagues line their result up against the idea of flow: in their model, matching challenge to skill produces the fastest learning, challenge pitched well above skill behaves like anxiety, and challenge well below it behaves like boredom. Educational psychology arrived at the same place from a different direction. Janet Metcalfe and Nate Kornell called the sweet spot the region of proximal learning, the finding that people study most efficiently when they work on material just beyond what they already know, neither mastered nor hopeless. The idea traces back to the older notion of a zone of proximal development, the space a single step past what you can currently manage on your own.

It is worth separating this from a related idea it often gets tangled with. Robert Bjork argues for what he calls desirable difficulties: practice conditions that feel harder in the moment, like spacing sessions out or testing yourself instead of rereading, tend to build sturdier long term memory even though they slow you down day to day. That is about the kind of difficulty you introduce. The 85 percent rule is about the amount. Wilson and colleagues are careful to call these frameworks related but distinct, and the distinction is worth keeping: you can practice exactly the right kind of difficult thing and still have it pitched at the wrong level.

What the 2019 derivation adds is a number where there used to be a hand gesture. Long before it, labs studying perception routinely tuned their training tasks to somewhere around 80 to 85 percent accuracy, and people given a free choice of difficulty tend to drift toward the middle as they improve. The math did not invent the sweet spot. It gave a well worn intuition a precise address.

Fastest, Not Furthest

One more piece of fine print, and it is the one most likely to get mangled. The 85 percent figure maximizes the speed of learning, not the final level of it. A learner training at a different difficulty still improves. It just gets there more slowly. The rule is about the rate at which the needle moves, not the ceiling it eventually reaches.

This corrects the way the rule sometimes gets sold, as if being right 90 percent of the time means you are wasting your life. You are not. You are learning at a gentler pace, which is completely fine for something you want to enjoy rather than optimize. The rule earns its keep only when speed is the thing you actually care about.

Aim for Roughly Right, Then Raise the Bar

The practical version is simpler than the math. If you want to get better at something quickly, look for challenges you can clear most of the time but not quite all of it. Right around four out of five attempts, with a real chance of missing the fifth. If you are getting everything correct, the task has gone stale and you should make it harder. If you are missing most of it, you have overshot, and the honest move is to drop back to something you can mostly do before creeping the difficulty up again.

The other half of the rule is that the target keeps moving. As you improve, the difficulty that used to sit at 85 percent slides down toward easy, and staying in the sweet spot means deliberately walking the challenge back up to the edge. This is why a good coach keeps adding weight, and why a language app that never gets harder eventually bores you flat. Each clean rep should make the next one a little less clean.

This is also where clearing challenges quietly does a second job. Every problem you solve at the edge of your ability is a small piece of evidence that you can, which is exactly how self-efficacy gets built, by doing rather than by being told. Difficulty pitched right does not only teach the skill; it teaches you that you are someone who can pick it up, which is often what carries you through the long middle of a goal.

The catch, in practice, is having a challenge you can actually dial. A goal like "get fit" has no difficulty knob; a specific, hard but reachable target does, which is part of why specific, challenging goals beat vague ones in the first place. This is the small, unglamorous thing Future You is built to help with: breaking a goal into steps you can size to the right level and ratchet upward as you improve, free on iOS and Android.

Where This Leaves You

The honest summary has two halves. The 85 percent rule stops short of proving how you, a human being, learn best. What it offers instead is a clean mathematical result about a class of learning systems, sitting on top of an old and well supported intuition that points the same way. Hold it that lightly and it is still one of the most useful rules of thumb going.

Aim for challenges you can clear about 85 percent of the time. When they start feeling easy, do not settle in and coast; that is the signal to make them harder. The edge is not a fixed place you arrive at. It is a line you keep stepping over, on purpose, a little at a time.