Is your skill hard because it has lots of parts, or because the timing has to be just right?
The history and science of skill complexity, organization and practice design, made playable.
1 of 11 · Start here
Think of a skill you’re learning. Answer honestly, there’s no wrong answer.
A dance routine: yes. A single jump: not really.
In a jump, the dip powers the push: yes. In a slow yoga flow you can pause almost anywhere: not much.
Or try an example (swipe for more):
2 of 11 · Try it
Here’s a jump, frame by frame. Tap a dotted line to cut the jump there, as if you practiced the two halves separately.
tap a line
Some cuts cost nothing. Some quietly turn the jump into a different skill. Tap one to see which.
3 of 11 · The toolbox
Picture any skill as boxes joined by lines. Thick line: these two parts depend on each other a lot. Thin line: barely. Every way of practicing keeps some lines and cuts others.
4 of 11 · The story
Science doesn’t move forward by proving things right. It guesses, tests, and catches its own mistakes. That’s Karl Popper’s big idea. David Deutsch adds that the best explanations are hard to fiddle with: you can’t tweak them to fit any result. Watch both happen here.
swipe
5 of 11 · Your turn
A good theory risks being wrong. So should you. Predict each result before you see it.
6 of 11 · The evidence
Compared with always practicing the whole skill. Dots show the direction of the evidence, not exact sizes.
Flight trainers answered “which is better?” with a smarter question: better per hour, at what cost? Play with the numbers. The starting values are made up.
Science name: transfer effectiveness ratio, hours saved per hour of piece practice ● Roscoe, 1971. In acrobatics, “cost” includes fatigue, injury risk and needing a spotter.
Practice devices for parts of flying, then simulators, then the real plane. The research started here, and transfer and cost are still measured. ●
Keyhole-surgery skills are drilled on box trainers and simulators before real operations, often until a set standard is reached. ◐
Whole tasks from simple to complex, with separate drills only for routine bits that must become automatic (the 4C/ID model). ● van Merriënboer, 1997
Therapists teach daily tasks like dressing by building them up step by step, sometimes starting from the last step. ◐
In one study of pianists, the best practice involved finding exactly where mistakes happened and repeating that passage until it was stable. ◐ Duke, Simmons & Cash, 2009
Progressions, drills and “whole, then part, then whole” are everywhere. Careful tests of specific progressions are rare.
7 of 11 · Fair warning
Much of this research comes from pilots, video games and lab tasks. How much carries over to acrobatics depends on how similar the task is: how fast it happens, how much of the body it uses, and how much thinking it involves.
Even within movement science, lessons learned on simple lab tasks sometimes reverse for complex skills ● Wulf & Shea, 2002. That is the best reason to test any of this on your own skill.
8 of 11 · Acrobatics
Is a handstand more timing-sensitive than a backflip? Yes and no. It depends which kind of timing you mean.
One moment decides the next. Get the take-off of a backflip wrong and the whole flight is wrong. This limits where you can cut a skill.
Science name: organization
Small errors grow by themselves unless you catch them early, again and again. A handstand is the extreme case. This sets how fresh and precise your practice must be.
Closest science name: continuous control. Splitting “timing” in two is this page’s own framing.
Tap start. After a moment the handstand begins to tip. Tap catch the instant you see it move.
A handstand behaves like a pencil balanced on its tip. Any lean lets gravity pull you further, so the lean grows on its own: x(t) = x₀ · cosh(t/τ), with τ = √(h/g) ● physics. This is the simplest model. A real body falls a bit slower, and the 6 cm margin is illustrative.
Try lowering the balance point. In this model, a lower balance point falls away faster, though it needs less force to stop. The same physics makes a short pencil harder to balance than a long one.
Why it matters: the corrections can’t wait for your eyes. Skilled hand balancers catch the lean early and small, mostly through pressure in the wrists and fingers, with shoulders and hips joining in for bigger wobbles ◐ Kerwin & Trewartha, 2001; Blenkinsop et al., 2017. Your nervous system needs roughly a tenth to a fifth of a second to respond ◐, and that’s a big slice of how fast the lean runs away.
Standing balance works the same way. Instead of smooth, constant correction, the body makes small, quick “throw and catch” adjustments ● Loram & Lakie, 2002. A handstand is the same problem on a smaller base, run by smaller muscles.
Once your feet leave the floor, nothing can add spin. The rotation you took off with is the rotation you have. You can only spin faster by tucking tighter, or slower by opening up ● physics. Errors don’t grow in the air. The spin is frozen at take-off.
What you can still steer is when you open. Gymnasts use their eyes during the flip to time the opening and the landing ◐ Lee, Young & Rewt, 1992; Bardy & Laurent, 1998. So a backflip has one big decision at take-off and one small, steerable one in the air.
Schematic. Tucked bodies spin several times faster than open ones; the 3× here is a ballpark ◐. Skilled performers really do fix take-off errors by timing when they open ◐ Yeadon & Mikulcik, 1996.
A cartwheel touches the floor four times: hand, hand, foot, foot. Every touch is a chance to push and correct, so it forgives a lot. Slow it right down and it changes character, becoming a chain of still balances, so a very slow cartwheel is partly a different skill.
An aerial is a cartwheel with the hands taken away, and that’s the whole difficulty. The corrections the hands used to make must now be built into the take-off, because nothing in the air can change your spin except your shape. It leaves the forgiving family and joins the backflip’s.
Where the timing pressure sits
Schematic, not to scale.
Where I’d put six acrobatic skills on the classic two-question map. These are coaching judgments, not measurements. Argue with them.
9 of 11 · Fatigue
One more guess, on the lab experiment closest to handstand balance.
That lab task is a precise squeeze between thumb and finger, steered by tiny changes in pressure. Handstand balance runs on the same kind of control in the same area: small, well-timed pressure changes through the fingers and wrists. The match is close, but no one seems to have run this study on handstands yet.
Tired muscles produce shakier force and can blur your sense of where your joints are ◐ Enoka & Duchateau, 2008. Reviews of balance research agree that tiring the muscles doing the balancing makes balance worse ● Paillard, 2012. In a handstand, that eats straight into a margin of centimeters and fractions of a second.
Whether tiredness hurts learning, not just the day’s performance, was debated for decades, and older studies gave mixed answers ◐ Godwin & Schmidt, 1971 ○ Carron, 1972. The 2019 study is one of the cleanest tests so far, because it measured learning after the tiredness was gone.
In jumping skills, fatigue mostly costs power. A lower take-off means less time in the air to finish the spin. That’s less a learning problem than a safety problem, and the more dangerous one.
Two caveats: the lab fatigue was strong, and a few moderate holds may not reach that level. And if long holds are your goal, some practice while tired is part of that skill.
For example: freestanding holds of 10 seconds or more, out of 10 tries.
Plan A: endurance holds before balance practice. Plan B: the same holds after it. Keep everything else the same.
Count at the start of the next session, rested. That measures what you learned, not how you did while tired.
If plan A wins, the idea is in trouble. One person can’t settle the science, but you can settle it for yourself.
10 of 11 · Be skeptical
Each with the best reply. A criticism without a reply is only half a test.
11 of 11 · Where it lands
Pulled together from everything above. It’s a proposal, not settled science, and it’s written so it could be proven wrong.
Cut where parts don’t depend on each other. Don’t cut where they do, especially when the link happens too fast to fix on purpose. And whatever you cut apart, you’ll later have to learn to put back together.
Film skilled performers. If the timing of one phase shifts every time the other does, they’re tightly linked. Decide this before the experiment, so nobody can relabel the skill afterwards.
Practicing tightly linked parts separately should transfer poorly. Loosely linked parts should split cheaply.
If tightly linked parts practiced separately work as well as whole practice, at the same cost, the idea is in trouble.
It fits the results on this page: the jump’s dip, the arm swing, the costs of splitting body parts, the backward-built landings, the moving spotlight and the bowling machine. But fitting known results is easy. Its value is in predictions that haven’t been tested yet.
Glossary
How sure is each claim? ● solid well established. ◐ gist right in outline; check details before quoting. ○ check needs verifying. Spot a mistake? Tell me at hi@softacroaaron.com. That’s how this page gets better.