Hacker News new | past | comments | ask | show | jobs | submit login
When You’re Hot, You’re Hot: Career Successes Come in Clusters (northwestern.edu)
134 points by barry-cotter 8 months ago | hide | past | web | favorite | 42 comments



I'm a bit skeptical.

They looked at scientists (citations), artists (painting prices), and film directors (imdb ratings). The question is whether the works that they're looking at are actually independent.

For scientists, it's rather unlikely that, having discovered a rich vein of research, a scientist would put out a single perfect paper and then never revisit the topic. Instead, I'd expect several papers exploring aspects of the research, which would get cited where those aspects are applicable; that pattern would produce several highly cited papers which look like a 'hot streak' but are actually just one piece of work.

For film directors, the Peter Jackson example the article gives is similar: the three Lord of the Rings movies are a 'hot streak' consisting of the director repeatedly putting together the same actors playing the same characters in the same story.

For artists, they're using the art market, in which prices are determined by the opinions of critics, gallerists, and other experts herding towards a consensus about importance and quality, as a proxy for quality. A conclusion that "the stuff artist X was doing during this period was important" would naturally produce a cluster of works escalating in price.

My expectation would be that after such effects are accounted for Einstein's anno mirablis that produced four transformative and unrelated papers would once again appear to be extremely atypical.


Whatever may be the contribution of actors in addition doesn't invalidate the results though. Whatever environment forms around the scientist may lead to a streak of great works. The streak remains.


But if we're looking at things like a trilogy, or a series of papers in the same small area, it only appears to be a streak due to a (somewhat artificial) decision about divvying up the results. These sorts of "streaks" could be interpreted instead as single events, and the whole premise of the article goes out the window.


I think your analysis is correct but you've missed the point of this article and the research cited.

The research says these hot streaks exist, you've given some examples of explanations for their existence, the two aren't contradictory. I don't think that article is making out that they're not caused by your proposed explanations.


What the comment points out is that the fields might be streaky ones, and that that isn't necessarily something that can be generalized. It's the choice of those specific examples that's being criticized.


These things happen about half a century. Lagrange, Gauss and Laplace outdid Einstein a few times over.

Wiener comes close but is not as famous as he did the groundwork for others.

Probably more groundbreaking research you haven't heard about.


I wanted to dismiss this finding as some kind of reputation adder, whereby after a hit a professional's work would receive extra positive attention that might not have been warranted. However, but the authors appear to extend their result to artists such as Van Gogh, who was not, to my understanding, famous during his lifetime. According to the authors, he managed to produce his best works in a clustered manner while the feedback from these works presumably remained about constant.


The research is using prices to determine the "best" paintings for an artist, but that means that their data for art is basically determined by the not-exactly-quality-based forces involved in pricing art. "Van Gogh painted this one at the same time as X that sold for $$$$Y" is an effective selling point to certain segments of that market.


In general I think the clustering of success comes from recognition of successful bits during/after repeated practice or iterative production. Once you recognize that you are onto something you can pursue that until you either become bored with it or the market does.

I have had two successes that kicked off hot-streaks and they each came after several years of thrashing around, learning the space and material with which to work. I am in the middle (hopefully) of a new thrashing phase, lots of core competency being built up, (waxonwaxoff) some minor successes, and a whole lot of dissatisfaction.


I have noticed this as well. Insights don't come out of nowhere (although they often appear to). In my experience, they are the result of focus and effort over a period of time.


This is a very under-appreciated article.

The more I look around the more I feel every one has at least the following phases in their career- Fall, Struggle, Rise, Plateau. The thing is the order is not guaranteed. But the phases come in parcels of decades.

The best is the order I mentioned. Sometimes the order is brutal. You go like Rise, Plateau, Struggle, Fall. The best years happen in early life- Like 20's. You go through it assuming this is how its going to be all life, don't invest or save up. And then reach 40's and realize you have another 40 to go and regardless of how hard you work, You have to suffer through the remainder.

In some people's case the big success comes in the last decade of their working career.

Either way the most important to note is, that luck goes around. No one remains lucky forever. One must be worried if start tossing a coin, and head comes up straight 100 times.


The more I look around the more I feel every one has at least the following phases in their career- Fall, Struggle, Rise, Plateau. The thing is the order is not guaranteed. But the phases come in parcels of decades.

That describes my career perfectly.

(this started in the mid 90s)

Fail - first pass of my first major development project out of college. I was the only developer at my company. I knew how to program - I done complicated side projects in assembly and C for years before and during college.

Rise - Asked a former professor for help, came out with version 2.0 of the project. Between two raises, a basically guaranteed 20^ bonus, and changing jobs, I was making $40K more in 4 years.

Plateau - next 8 years, took my eye off the ball, stayed at the one job too long and between bonuses being cut, and 3% raises, only made $7K more in year 9 than I made in year 2.

Plateau #2 - I changed jobs, learned a lot and muddled my way through the recession for 3 years.

Rise - over the next 5 years, changed jobs 4 times and made $55K more at the end.

For context, because of the mistake made by staying at a job too long up until 2008 and focuses on always getting jobs where I’m barely qualified to learn and build my resume, I’ve been under the median salary for my market for my years of experience since then. So I’m definitely not trying to brag. I’m also not in Silicon Valley. I am in a major metropolitan city.


For me I lived my childhood on a plateaux then kept falling all throughout my teens and early adulthood then for the past 10 years it's been all struggle. I'm desperately trying to get back to the plateaux level which is what I need to feel normal.

When I see colleagues at work who were born in a difficult place and now they have the same job as me, I feel bad because they've never been happier but I'm having the worst time of my life.

At the same time, I find that people who come from a difficult place tend to have their own distinct type of mental scars.


Can confirm.

Got 4000 stars on GitHub and suddendly a publisher offered me a book deal.

Had some interviews and people told me, they already knew me from the news, where I got to via the 4k GitHub stars.


Please do elaborate if you can.


I wrote a React tutorial in 2015.

A React core dev starred it on GitHub.

I got about 4k stars for this and some news coverage (front-end/JS/React news)

Some publisher asked me if I want to turn it into a book, so he can sell it for/with me.

Later I had a few interviews where I was asked to show my GitHub account. When the interviewers saw the 4k repo they told me they read about it a few days ago in the news and now they were impressed because they talking to the creator of it in person.


I wonder how much of it is having one breakout hit drives more attention towards their later ones and helps them have better resources (budget, staff, equipment, etc).

Works that might have ignored or simply missed earlier now get more attention.

And works that might have failed due to lack of resources now have a better shot.


i don’t think it’s so much that people become “hot” as people tend to work on the same thing for a while. So of course your biggest hits will be in close proximity to your next biggest hits — they are variations on the same theme! Hot topics, not hot people.

I would be quite surprised if patterns like Einstein’s annus mirabelus are actually the norm, with four largely unrelated breakthroughs in the same year.


It would be interesting to know how these "hot streaks" coincide with personal events occurring in people's lives.



… is the first paper cited in the abstract of the paper in the article: https://www.nature.com/articles/s41586-018-0315-8


Link to the actual paper https://arxiv.org/abs/1712.01804


Exciting research for someone who did their best work recently. But maybe scary for someone who think their best work happened few years back.


What about Jeff Dean?


Between 2000-2004: GFS, GroupVarInts, Snappy, Protocol Buffers, Google's cluster-management system, MapReduce, AdSense, BigTable, & a critical part of the Search serving stack that doesn't appear to be public.

Between 2011-2015: LevelDB, Google Brain, another important internal project that I don't think is public, and TensorFlow.

Between 2004-2011: Spanner, some work on Google Translate, and consultation for the replacements for GFS, Proto1, and MapReduce.

Certainly seems like clusters to me.


Isn’t that just one big cluster from 2000-2015? Or, was that your point?


It's two clusters, one centered around big data from 2000-2004, and another on large-scale machine-learning from 2011-2015. The middle period from 2004-2011 mostly just featured Spanner and helping other people pick up the torch for his accomplishments of 2000-2004. While Spanner is great and last I heard had one big client inside Google, it's hardly as impactful as GFS, Protobufs, MapReduce, Google's serving system, and BigTable were, nor as Brain, TensorFlow, and that other internal project are now.


I think it may have to do more with one idea triggering another. For instance, Einsteins papers were all pretty much the "logical" conclusion of a single idea - mearly separated by publication date.

In my own life, I can witness similar events in collegues and myself. It may also be that others take a "hot streak" train of thought, style, premise, and once it becomes common place, that hot streak or cluster ceases to be relevant.


Einstein's annus mirabilis ("miracle year") in 1905 involved publishing 4 excellent papers, covering some very different ideas: the photoelectric effect, brownian motion, mass-energy equivalence, and special relativity. It's true that he later focused on relativity, but the photoelectric effect and brownian motion are still quite different.


Agreed. The photoelectric effect paper revolutionized quantum physics, the brownian-motion paper revolutionized statistical mechanics, and relativity changed everything.

They are really different papers, each of which most physicists would be extremely blessed to have just one in their lifetimes.


> For instance, Einsteins papers were all pretty much the "logical" conclusion of a single idea - mearly separated by publication date.

What was the root idea that all of his papers were the "logical" the conclusion of?


> Einsteins papers were all pretty much the "logical" conclusion of a single idea

What single idea are you referring to?


Damn. Could it mean if you can reflect and pinpoint your glory years being a while in the past then you are screwed?


I mean, is your career finished? Reading this, I wondered if there were any other clusters that would have been the glory years, but later on the person’s accomplishments were so much better it completely eclipses their past glory.

It seems the researchers concentrated on creative careers. Wonder if it is applicable to more corporate standardized careers and how would they would measure it. The title implies careers in general, but I don’t find what the article says to support the careers in general.

Would also warn against using this as self fulling prophecy.


If you were to study this for corporate standardized careers, how would you go about getting data?

I can't think of a reliable, detailed, readily available source of info like that. Maybe LinkedIn for profiles that explicitly state education years (HS and B.S. to estimate age) and then somehow filter through and classify job titles?

Several issues not including users who state themselves "Founder", "CEO", for MLM schemes. Maybe you could cut the population down to users who have had software titles.


As another person said, large well run HR departments have this data. Level employees come in at, performance reviews, compensation, and job titles within the company throughout employees tenure.

It would be hard to standardize it across multiple companies in a person's career. Maybe for people that stayed at one company their entire lives. Govt/military likely have higher rates of people working and retiring withinn their system for 20+ years. And their job titles should be standardized as well.


I expect it would be effectively impossible. The meaning of titles vary markedly between companies, a lot of companies don't even have well-standardized titles, and people often don't list a succession of titles within a given company. And I'm not sure how any of that maps to success anyway. Even if you had salary info, which you don't, that's only one metric and, for many, not even the most relevant one.


Perhaps data from large HR departments would be useful. Although that would limit it to people who stay at the same company.

Brings another question, is it possible to find out which two companies in the world have the most overlap between employees? To clarify, two companies which a majority of knowledge workers jump between.


No, the paper says you can multiple streaks as well.


It does. What I meant was if jt would be easy to get depressed with "oh damn there goes my streak so why bother as il never know when the next one will be". Now you could eqully go "cool there should be another one so il just keep trying".

If that is the case, is this just not another angle at "never quit,you never know things will pop up"?


If you say that too loudly on HN you'll get downvoted for promoting survivorship bias


Can’t wait for my first cluster!




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

Search: