Bratha

Part 2.1 — Deliberate Practice

17 min read

This is Part 2 of 5 in the Learning & Skill Acquisition Series (the depth on step 4 of the four-step method)

  • Part 1 — Foundation:
  • Part 2 — Mindset (2 sub-articles):
  • Part 3 — Behavioural (2 sub-articles):
  • Part 4 — System:
    • Part 4.0: The Learning System (Obsidian + Excalidraw, structure vs context, the publish workflow)
  • Part 5 — Chemistry:

Table of Contents

What this article is, in one sentence

Part 2.0 ended with one line: practise the specific sub-skills your map says are load-bearing, against an honest error signal, in conditions close enough to the real thing that the practice transfers. This article unpacks every word in that sentence. If Part 2.0 was how to build the map, Part 2.1 is how the map becomes a skill you can actually use.

Why deliberate practice needs its own article

The word “practice” is doing a lot of work in most learning advice, and almost all of the work is hidden. Two people who both “practise daily” can end up with completely different outcomes, because most of what people call “practice” is actually rehearsal.

Rehearsal is doing the thing again in the way you already know how. It’s running through the recital piece you’ve already memorised; it’s grinding through ACCA past papers on the topics you already understand; it’s running the same Zone 2 ride every week. Rehearsal feels productive (you’re “putting in the hours”), it produces a satisfying sense of fluency, and it produces almost no improvement after the first few weeks.

Practice is doing the thing at the edge of your current ability, getting an honest signal about where it broke, and adjusting. Practice feels worse than rehearsal (because you’re failing on purpose), it produces a less satisfying sense of fluency in the moment, and it’s the only thing that actually compounds.

Most people who plateau in a skill aren’t lazy and aren’t under-practising. They’re rehearsing, faithfully, every day, and confusing it with practice. The rest of this article is about what separates the two and how to stay on the right side of the line.

The definition: practice ≠ rehearsal

The working definition

Deliberate practice is repeated, focused attempts at a sub-skill at the edge of your current ability, structured to generate an honest error signal, followed by adjustment. The four required ingredients: edge of ability, honest error signal, adjustment, and repetition of that adjusted version.

Take any one of those out and you no longer have practice:

  • No edge of ability → rehearsal. (You’re staying inside the part you can already do.)
  • No honest error signal → flailing. (You’re trying things at the edge, but you can’t see what worked.)
  • No adjustment → mileage. (You’re getting the signal but not changing anything in response.)
  • No repetition of the adjusted version → one-off insight. (You changed it once and then went back to the old way.)

This definition also gives you a clean diagnostic for whether a given session was practice or not. If you finished a session and didn’t know more about your own weakness than when you started, you weren’t practising. You were doing something else (probably rehearsal, possibly entertainment).

The focus-hours reframe of the 10,000-hour rule

The “10,000 hours to mastery” claim is one of the most repeated numbers in the productivity world, and one of the most misleading. It originates with Ericsson’s research on world-class performers, was popularised by Gladwell in Outliers, and Ericsson himself spent years pushing back on the version Gladwell wrote.1

What the research actually says is closer to: world-class performers in highly developed fields (violin, chess) typically have accumulated about ten thousand hours of deliberate practice by the time they reach world-class. Two things are doing all the work in that sentence: the word “deliberate” and the qualifier “world-class.”

The honest reframe:

The reframe

==Mastery runs on focus hours, not calendar hours.== A focus hour is one full hour of deliberate practice as defined above (edge of ability, honest error signal, adjustment, repetition). It is not “an hour spent doing the thing.” Most people’s “10,000 hours” of guitar contain perhaps 1,000 focus hours and 9,000 hours of rehearsal-with-the-TV-on.

Two consequences:

  • Mastery often comes much sooner than 10,000 calendar hours, sometimes dramatically sooner, when the focus-hour density is high. The 10,000-hour figure is a marker, not a wall.
  • It also often never comes at all, no matter how many calendar hours you put in, if those hours are mostly rehearsal. This is why people plateau permanently in skills they “practise daily.”

For most non-world-class outcomes (functional competence, professional-level skill, “good enough to make a living from this”), the number is much lower. Kaufman’s First 20 Hours claims you can reach functional competence in most skills in about twenty focused hours, and the underlying claim (that the early steepness of the learning curve gets you most of the way to “useful” surprisingly fast) is robust. ==The compounding really happens after that, when you have enough skill to meaningfully practice.==

Trial-and-error is the engine; practice shrinks the time per trial

Here’s the principle that ties everything together:

The mechanism

==Trial-and-error is the engine of skill acquisition. Practice does not replace it; practice shrinks the time per trial, which multiplies the number of trials you can run per unit time.==

This reframes what “getting good at something” actually means. A novice in a domain spends most of a trial setting up the trial: working out what to do, second-guessing, executing slowly. An expert spends almost no time on setup and almost all of it on the actual attempt, plus the read-out of what happened. ==An expert isn’t doing better trials than a novice in the same amount of time; they’re doing many more trials, getting many more error signals, and adjusting many more times.==

This is also why the felt experience of practice changes with skill level:

  • Novice practice is exhausting because every trial is half setup. The error signal comes slowly, the adjustment is uncertain, and you can only manage a few trials before fatigue.
  • Intermediate practice is satisfying because trials get faster, the error signal sharpens, and you start seeing your own patterns.
  • Expert practice is frustrating because the trials are so fast and the gains so small that you have to deliberately re-engineer the difficulty to keep generating an error signal. (This is why elite performers spend so much time on weird-looking drills that isolate one specific sub-skill: at their level, the regular thing has stopped producing useful errors.)

The strategic implication is that the fastest route to better practice is whatever shrinks your time per trial. For a programmer, that’s better tooling, faster feedback loops, smaller test cases. For a lifter, that’s an exercise selection that lets you fail cleanly at the bottom of a rep instead of grinding through it. For a student, that’s actively recalling answers (a fast trial) instead of re-reading notes (a slow non-trial). Every improvement in the speed of the trial multiplies the rate of learning.

Generating the error signal early

The hardest part of practice is the part nobody enjoys: forcing yourself to fail visibly, on purpose, early.

You cannot correct what you can’t see. If the only error signal you get is the result at the end of a long session, you’re learning at the slowest possible rate, because every adjustment is based on a fuzzy aggregate of dozens of small mistakes you didn’t catch individually. The trick is to move the error signal earlier in the loop, so each adjustment is based on a clear, specific failure.

Three mechanisms do most of the work:

  • Produce before consuming. Before reading the textbook chapter, attempt the end-of-chapter problem. You will fail. The failure is the error signal; the failure tells your brain what to pay attention to when you do read the chapter. A chapter read after a failed attempt sticks an order of magnitude better than the same chapter read cold. (This is the cognitive-psychology effect called productive failure; it’s well-replicated.2)
  • Predict before checking. Before you look up the answer, write down what you expect it to be, and why. The gap between prediction and answer is the most efficient teacher in the world, because it bypasses the “I knew that already” illusion.
  • Write before reading. When you sit down to “learn” a topic, draft your current understanding of it first. The act of writing surfaces the holes. The holes are your reading list.

Each of these turns one slow trial into many fast trials. A two-hour study session built around prediction-then-check generates dozens of error signals; a two-hour session of passive reading generates one (“did the material feel familiar?”).

The discomfort budget

Part 1.0 introduced limbic friction: the unpleasant feeling at the start of focused effort that the brain uses to decide what to rewire. Practice has a related but distinct quality: the felt cost of generating error signals on purpose.

This is sometimes called the discomfort budget: the amount of unpleasant-but-productive failure you can tolerate in a given session before the brain starts pattern-matching the activity as “thing I want to avoid” and bailing.

The honest rule

If a practice session never felt uncomfortable, you weren't practising. You were rehearsing. Conversely, a session that felt uncomfortable the whole time, with no successes, is also not sustainable: the discomfort budget runs out and you start avoiding the activity.

A working ratio for most people, in most domains, is something like 60–80% success and 20–40% productive failure. The exact number varies (chess research suggests something closer to 85% success for stable engagement3), and the point isn’t the precise ratio. The point is that there’s a band where practice is hard enough to generate error signals but easy enough to sustain.

Two implications:

  • A session that’s “easy” the whole way through is a warning sign, not a win. You’re operating below the edge of your current ability and the work isn’t doing much.
  • A session that’s punishing the whole way through is also a warning sign. You’ve stepped past the productive edge into territory where the error signal is too noisy to extract anything from (think: a beginner sparring with a black belt — there’s plenty of failure, but the failure modes are too uniform to learn from).

The sweet spot is uncomfortable in a specific, productive way: you can name what you were trying to do that didn’t work, and you have a guess about what to try next.

The plateau diagnostic

Every skill produces plateaus. The amateur reading of a plateau is “I’ve stopped improving; I need to push harder/longer.” That reading is wrong almost every time. Here’s the correct one:

What a plateau actually means

==A plateau means the current technique is mined out. Not the skill, not your potential. The specific approach you’ve been practising has extracted everything it can extract; further repetitions of it return nothing.== The fix is a technique change, not an effort change.

This shows up in every skill:

  • Lifting. You stall on linear progression after several months. The fix isn’t more sets; it’s a different progression scheme (double progression, undulating, block periodisation). The body had adapted to the stimulus; the stimulus has to change.
  • Running. You stop getting faster despite increasing mileage. The fix isn’t more miles; it’s adding intensity work (intervals, tempo runs, the things the Athletic series covers). Mileage trained one capacity to its current ceiling.
  • Programming. You can ship features but feel stuck at “junior” indefinitely. The fix isn’t more features; it’s deliberately taking on the kind of work that requires the next skill (reading codebases, designing systems, code review). The current activities trained one set of skills to their ceiling.
  • Languages. You hit conversational competence and stop progressing despite still using the language daily. The fix isn’t more conversation; it’s stepping into harder territory (reading literature in the language, technical writing, formal contexts). Casual conversation had taught you everything casual conversation can teach you.

The pattern in all four: the activity that got you here will not get you to the next level. Plateaus aren’t a sign that you’ve hit your limit; they’re a sign that you’ve hit this method’s limit. The diagnostic is to ask, “what about my current practice is no longer producing a fresh error signal?” The answer points at what to change.

This is also why plateaus correlate so strongly with quitting. The natural response to “I’ve stopped improving” is “I must not be made for this,” when the real answer is “the practice I’ve been doing has run out of teaching.” A practitioner who internalises the plateau diagnostic doesn’t quit at plateaus; they treat them as scheduled prompts to reshape the practice.

The 90/10 of practice itself

The same 90/10 rule this blog uses everywhere (90% of the result is structural, 10% is the margin) applies recursively to practice:

The 90/10 of practice

==90% of the gain comes from what and how you practise; 10% comes from raw volume.== Doubling the hours of bad practice produces almost nothing; reshaping the content of the practice produces more than tripling the volume.

This is why two students with the same number of study hours can end up at wildly different levels: the structure of those hours is doing almost all the work. A focused 90-minute block of prediction-then-check on the topics your last mock exam said you were weakest on is worth perhaps three hours of re-reading notes you already understand.

The implication is brutal but liberating. If you’re already practising several hours a week and not seeing progress, the answer is almost never more hours. The answer is to look at the shape of the practice. Where’s the error signal coming from? Are you actually at the edge of ability, or just inside it? Is there an adjustment you haven’t tried? When did you last deliberately change the kind of practice you’re doing?

How the four steps support practice

Part 2.0’s four steps aren’t separate from practice; they’re what makes practice possible. Take any one out and the practice degrades:

  • Without step 1 (causation history), you don’t know which sub-skills are load-bearing, so you practise whatever’s in front of you, which is usually the wrong things.
  • Without step 2 (your own narrative), you can’t tell when the practice is producing structural learning versus when it’s just adding scar tissue to a wrong understanding.
  • Without step 3 (root note), every error signal floats free instead of updating your map. You “learned” the lesson and forgot it within a week because there was nothing to attach it to.
  • Without step 4 (this article), the map never becomes a skill. You can describe how to do the thing and you cannot actually do it.

The four steps and deliberate practice are the same loop, not two separate tools. The map you build in steps 1–3 tells you what to practise; the practice in step 4 updates the map. Each pass through the loop sharpens both at once. This is what compounding looks like in skill acquisition: each cycle gives you both a better map and a better skill, and each makes the next cycle more efficient.

Part 2 Takeaways

What to carry forward

  • Most “practice” is rehearsal. The diagnostic: did you end the session knowing more about your own weakness than when you started? If no, it wasn’t practice.
  • The four ingredients of deliberate practice: edge of ability, honest error signal, adjustment, repetition of the adjusted version. Miss any one and it stops compounding.
  • The 10,000-hour rule is about focus hours, not calendar hours. Functional competence in most skills is closer to 20 focused hours than 10,000 calendar ones.
  • Trial-and-error is the engine; practice shrinks time per trial. The fastest route to better practice is whatever tightens your feedback loop.
  • Move the error signal earlier in the loop. Produce before consuming, predict before checking, write before reading.
  • Honour the discomfort budget. Too easy = rehearsal; too hard = unsustainable. Aim for 60–80% success with productive failure on the rest.
  • Plateaus mean the current technique is mined out, not your potential. Change the technique, not the volume.
  • 90% of practice gain is what and how; 10% is volume. If you’re stuck, the answer is almost never more hours.
  • The four-step method and deliberate practice are the same loop. Each pass sharpens both the map and the skill.

Your Practice Task List

Restructure one current practice

  • Pick one skill you’ve been practising and feel stuck on. Be honest: is it actually stuck, or are you rehearsing?
  • Name the sub-skill you’ll target this week. Not “get better at X.” Specifically: which sub-skill of X, at the edge of your current ability, are you going to practise?
  • Engineer the error signal earlier. What’s one thing you can do before reading/watching/asking — a prediction, a draft, an attempt — that turns one slow trial into many fast ones?
  • Set the discomfort budget. What does a sustainable hard session look like for this skill? Roughly how often should you succeed vs. productively fail?
  • Check for a plateau. When did this practice last produce a fresh error signal? If it’s been a while, the technique is probably mined out and needs a structural change.
  • Read Part 3.0 next. The four steps and deliberate practice tell you what to do; Part 3.0 tells you when in your day to do it, and how to set up the conditions where focused practice is actually possible.

Sources & references

Disclaimer

The frameworks here describe averages in well-studied domains. Real skills have their own internal structure, and your mileage will vary. The honest answer to "how long will this take?" is almost always "longer than you think for mastery, shorter than you think for usefulness." Use the principles as a way to spend your hours better, not as a promise about how few hours you’ll need.

Footnotes

  1. Anders Ericsson and Robert Pool, Peak: Secrets from the New Science of Expertise (2016). Ericsson’s foundational paper is Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). “The role of deliberate practice in the acquisition of expert performance.” Psychological Review, 100(3), 363–406. Ericsson’s central pushback on Gladwell is that the 10,000-hour figure is (a) the average among world-class performers in specific domains (not a threshold), and (b) about deliberate practice, not just time spent. The variation around that average is huge. ↩

  2. Productive failure: Kapur, M. (2008). “Productive failure.” Cognition and Instruction, 26(3), 379–424. The robust finding is that attempting a problem before instruction produces better learning of the relevant concepts than instruction-then-attempt, provided the attempt is genuine and the instruction follows it. ↩

  3. The 85% success rate for optimal learning: Wilson, R. C., Shenhav, A., Straccia, M., & Cohen, J. D. (2019). “The Eighty Five Percent Rule for optimal learning.” Nature Communications, 10(1), 4646. The specific number applies to certain classes of perceptual learning tasks; the direction (there’s a sweet spot of difficulty, neither too easy nor too hard) generalises broadly. ↩

Graph View

Backlinks