The Skeptical Methodologist

Software, Rants and Management

The 10x Myth

There’s a myth abound in software development circles, and it needs some deconstructing. It’s probably one of the best indicators of how much further the industry has to go in regards to sexism, since it’s a patently masculine myth, evoking images of great Greek Heroes slaughtering thousands of men as they move forward.

The Myth of the 10x Developer

Now, this isn’t a myth because it hasn’t been researched. There is ample amounts of research on programmer productivity, at least from the 80’s, and if it is still to be believed we should assume that there is at least some difference in  programming ability between developers.

The real myth comes from the interpretation of these results, and that’s where the testosterone-fueled neck-bearded bias comes in. There’s the results of these studies, and then there are how the studies are understood by so-called “Rock Star” developers who always assume they are one of the 10x’ers and that’s their justification for why they shouldn’t be forced to get along with anyone.


The results are thus:

There is roughly an order of magnitude difference in productivity (measured as time to get code working for a toy problem) between the best and the worst programmers, with causes unknown.

However, here’s how it’s commonly repeated. See if you can spot the difference:

There is roughly an order of magnitude difference in general productivity between the best and average programmers, and it’s due entirely to innate talent.

So, let’s take this apart one by one.

We’ve solved this problem before…

The first flaw is somewhat methodological, however I don’t think the researchers ever claimed that their toy problem measured generalized productivity, so it’s also a flaw in how the general population of brogrammers have read the result. Think of it this way, if I took a random sample of programmers and gave them a test, even if their skills were all roughly the same, what kind of result would I see? I’d see some programmers doing better than others, because they’ve solved that or a similar problem before. I can compare a person who’s never written an SMTP server to someone who has, ask them to do so, and witness a miraculous 10x or more productivity benefit to the programmer who has built it before. Imagine that!

This is similar to what you might call the halo effect. Rock stars are identified by their ability to solve their specialized problem very well – perhaps they’ve built a few Rails apps from start to finish. They’re going to be great at that. But throw them at writing a compiler and watch them flounder.

Distribution of wealth…

The second issue is the confusion of the average for the worst programmer. Let me give an example of why this might be an issue. If I reported to you that the best programmers are 10x better than the worst, that’s one bit of information, but not enough to really say anything about the average (i.e., the majority of programmers). If I then said that average programmers are 9x better than worst, now we know something about the average, and the distribution. Unfortunately, we have no idea which way the distribution of these results is skewed, at least as it’s commonly reported.

First, it’s outright false to say that the best developers are 10x better than the average, even though that’s often what’s reported. Second, we don’t know if the productivity difference at hand is due to the best being that much better than everyone else or the worst being that much worse than everyone else. The issue here is that due to all the manliness in our industry, we of course all assumed it must be the former, and not the latter. Because we’re all magically that 10x’er, and that is why everyone else is jerks.

We point to this myth over and over again to justify why we don’t get along with others. It’s nearly always used to justify mistreatment of our colleagues – they’re all idiots, I’m brilliant, and that’s why I shouldn’t change and they should. I’m a 10x programmer. But clearly, the best you can claim if the distribution has a fat tail to the right (i.e., the worst are much worse than the average) is that you’re only a little bit more productive than the average, and dogonit, you ought to pay attention more to what your colleagues say because they’re all nice people and would it kill you to shower?

Cause and Effect…

The last issue that the studies say nothing about, at least as they’re repeated in their mythological form, is why? There’s an implied why, an implied cause for this difference, but it’s hardly ever stated. These genius gods-among-men programmers are so productive because they wield the magic of Zeus. What they do cannot be replicated, repeated, or taught to anyone. They’re entirely packaged up, not able to be distributed or copied. If you want the 10x programmer, you have to accept his ego, his arrogance, his complete lack of communication or emotional skills, and his tendency to shit all over everyone else. And yes, it’s almost always a he.

The issue here is that we have no idea what makes the 10x best programmers more productive than the worst programmers. Sheer numbers of years of experience don’t seem to play a role, but is that because there’s far too many enterprises where we can disappear and never have to code again? A year at a large enterprise curating UML documents is not the same as a year getting your own Rails site up for customers to use. What tools or techniques did these 10x programmers use that the worst ones didn’t? Were they more skilled with the debugger? Did they adopt more structured coding conventions (this was in the era of structured code)? Did they test their code any different?

There are many questions we can ask, and non of the productivity myths answer. It is almost always supposed to be magic, always supposed to be something innate to the neck beard itself that grants the brogrammer who dons it the powers to develop and deploy only the best code, and disregard everyone else’s opinion who may have something to add (or learn!).


There are 10x programmers – or at least there were, in the 80’s – who were ten times, roughly, more productive than the worst programmers. But given how bad some programmers can be, it’s probably safer to say that you should ensure you don’t hire (or at least train) the worst of your crew rather than try to always hire the best. Average programmers are, on average, pretty good in my experience. That informs me that the tail is fat to the right, not to the left. Moreover, invest in methods and tools that are shown to increase productivity: iterative methods, testing and peer review, static analysis tools, and training in your methods of source control and deployment. It isn’t magic – there is a way to turn average to great, and we can figure out what that way is if we use the methods of SCIENCE!

Finally, don’t fall for this machoismo myth that there are the great men and then there is everyone else. It keeps far too many ‘good’ average developers who don’t fit our implied mold of the great programmer – apparently an asshole white guy – out of organizations that sorely need them.


August 26, 2015 - Posted by | Uncategorized

1 Comment »

  1. […] gone into what’s wrong with the ‘common’ view of this adage in other posts. We’ll just sum up here and say that it’s a complete misunderstanding of the science, […]

    Pingback by The 4 Types of Programmers « The Skeptical Methodologist | August 21, 2016 | Reply

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: