Hacker News new | past | comments | ask | show | jobs | submit login
The Role of Deliberate Practice in the Acquisition of Expert Performance (acesounderglass.com)
125 points by luu 3 months ago | hide | past | web | favorite | 30 comments

David Epstein's book "Range" offers a different view of expert performance -- that it doesn't work very well in domains with ambiguity (such as a hidden state), or when a variety of skills are required.

In other words, expert performance applies to piano and chess very well, but a diversity of backgrounds is more important for things like management or R&D more broadly.

NOTE: Most of my comments on HN are about "practice", so feel free to just read them.

I'd be curious what qualifies as "ambiguity" and "diversity of background" however. Is that simply a collection of smaller expertises that you can pull from? Many domains have a degree of ambiguity to them - how do you create a musical masterpiece? how do you beat your chess opponent? Knowing "the moves" only ensures you can implement a plan.

It seems to me possible to improve at the skill of discovery - but I'm not sure how, and perseverance, creativity and luck swamp everything else. Those aren't skills.

Ah, but "chance favors the prepared mind", so it IS possible to improve your luck.

But is it a "skill"?

It is better called "ability". And "luck" is better called "recognizing an opportunity".

If you're unable to utilize an opportunity, which takes skill, or are unable to recognize an opportunity, which also takes skill, you're out of luck.

Ericsson's popular-market book Peak has much more than the paper (there are many citations in the back).

I remember in particular a duscussion about music students practicing for an optimum, or maybe maximum, number of hours per day. I'm in math and there are a number of people in math who also say four is the limit; a quick google found this for example https://theweek.com/articles/696644/why-should-work-4-hours-.... FWIW.

Also well-konwn are the first two paragraphs of Littlewood's Miscellany, https://www.gwern.net/docs/math/1986-littlewood-littlewoodsm..., p 189.

FWIW one might be able to find their own optimum by carefully listening to their body. In college when I overdid math I had a very distinctive buzz in my head (like the brain "shorting out"), that signified I am done for the day.

Remarkably, the same material that stupefied me just prior to "shorting out" would become crystal clear the next morning. I think the brain needs to be pushed up to a certain limit before it considers the material important enough to be worth of serious "structural changes". The pay-off is non-linear.

The difficult part is that you might be tempted to quit because you're "tired", even before you actually hit the wall, but then blowing past the limit in a fit of zealotry just destroys motivation without any extra benefit. The balance is hard to find. The best I came up with is try to observe my performance diminishing and use that as an "objective" signal. Easier in some areas than others.

> Remarkably, the same material that stupefied me just prior to "shorting out" would become crystal clear the next morning.

Very much in line with what one reads in Barbara Oakley's "A Mind for Numbers," and her free Coursera course based on the book (both recommended).

The Dunning-Kruger paper is really about the value and difficulty of self evaluation, not the "stupid people don't know they are stupid" that the internet memes it about.

And this is nicely actionable. When learning a new task, it can be very valuable to consider how to evaluate success.

Four per day per domain ? or is it four hours of thinking/growth and more hours of rest ?

By that I mean can we do more than one thing per day.

i would think it is 4 hours per domain.

like crop rotation (or your analogy of choice), the brain needs to be engaged in different ways at different times.

after hours of programming i am exhausted, but i can easily play an instrument for a few more hours, or do housework, etc...


And one can use overlapping synergies. Like hiking/running to let your brain creates solutions on the go :D

which is what charles darwin in his schedule seems to be doing

I know of no hard information either way.

All the studies focus on single-task performance. Indeed, typing random characters for several hours a day would lose my attention.

I don’t know of a study that estimates the maximum number of productive hours over a varied workload such as running a startup. There clearly is a limit, but it’s high — maybe 90 hrs/wk for motivated people.

In practice, it’s the neglecting of other things like relationships and exercise that kicks in before mental energy limits.

There’s a difference between productive and deep focus.

Even if it’s a factoid, based on personal experience and anecdotal evidence from colleagues, 4h is a good approximation.

I will admit though, that there’s a confounding variable here, that is how much your coworkers value your time. When I worked 4h or 6h workdays, everyone considered my time to carry a premium, so they avoided interrupting me if they could.

I do know people that can work extremely long hours and still be productive but they are quite rare. For a large majority of people a ninety hour work week, would be extremely counter productive at best but most likely dangerous for your mental health.


I'm curious about the nature of the 90hours per week too.

Is it ~slow paced work, no rushing but constantly walking ideas and attempts ?

Can they endure 90 hours of hard dead ends / frustration / disappointment ? Or is it 90hours of tangible progress (lets include small dead ends that provide nice references for more ideas) and pleasure ?

Newport's book puts the distinction between deep work and shallow work.

Calling up restaurants and trying to get them to switch to your booking software is shallow work. Needing to sit down and program the complete back end of that booking software including testing by Friday is deep work.

You can only do so much deep work before your reserves of energy are depleted.

The assumption that programming is deep work seems like a self-serving prejudice. Which parts of it are automatic will differ from person to person, but I imagine that a substantial majority of programmers will spend a substantial chunk of their programming work doing things that are shallow to them.

One of the challenges of deep work is getting in the zone. Does the book go over that?

If by "zone" you mean "flow," then yes; here's a relevant quote from Deep Work:

> The connection between deep work and flow should be clear: Deep work is an activity well suited to generate a flow state (the phrases used by Csikszentmihalyi to describe what generates flow include notions of stretching your mind to its limits, concentrating, and losing yourself in an activity—all of which also describe deep work). And as we just learned, flow generates happiness. Combining these two ideas we get a powerful argument from psychology in favor of depth. Decades of research stemming from Csikszentmihalyi’s original ESM experiments validate that the act of going deep orders the consciousness in a way that makes life worthwhile."

Newport, Cal. Deep Work: Rules for Focused Success in a Distracted World (p. 85). Grand Central Publishing. Kindle Edition.

Since the OP starts off quoting Cal Newport's book "Deep Work", we need to differentiate between shallow work and deep work.

Running a startup is a whole bunch of shallow work most of the time.

Work where you have to get into "the zone" to get it done, that's not 90 hours per week. Or likely even 19.

One thing I commonly hear in response to a question of the form “How can I effectively learn X?” Is “through experience” or “you just have to do it”. One frustrating thing about that response is that it takes skill to know what to focus on while practicing. In some domains like sports, people recognize that coaches are useful for learning what to focus on as you practice. It would be great if there were coaches for skills such as a software engineer’s prioritization, planning, and estimation of day-to-day tasks.

I wonder if theres a difference between straight 'studying' and practice that involves mechanical movement. What I mean is reading a book on a subject vs. doing gesture drawing or piano playing etc.

I can definitely believe theres rapid diminishing returns after an hour when practicing things that require physical activity.

In any case, I still believe that Cal Newport's Deep Work is a good reading for someone who wants practical advice on Ericsson, Krampe, and Tesch-Römer’s 1993 paper. I wrote a review in detail of Deep Work that you can see here: https://alvaroduran.me/deep-work

There is deliberate practice and there are geniuses who also do deliberate practice...

If you have not yet seen the extremely talented composer and piano player Emily Bear perform the “Bumble Bear Boogie”, you are in for a treat.


I'd love to find meta analyses of problem solving skills learning.

I'm thinking a review of spaced repitition software, like Anki, logs would give more information on efficient learning.You could map hours spent and schedules to outcomes.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact