I should clarify that by referring to it as a myth, I'm not denying in my post that there are people who are 10x as productive as most people.
The "myth" is that there was a formal study done and these performers were identified. The reality is that the study was looking at the difference best and worst performers, not best vs average, which is an important clarification.
This has since evolved to a further myth that 10x programmers exist, but are somehow always harmful and should be avoided. There are plenty of articles examining "The 10x programmer myth" which, with no more evidence than the original assertion, put forward that 10x programmers should be avoided.
And here in this article, it gets re-written again, that they aren't even 10x as productive, and just produce ten times as much in terms of LOC but it's all terrible code!
> The reality is that the study was looking at the difference best and worst performers, not best vs average, which is an important clarification.
That is an important, often misunderstood, clarification, but I think it’s also subtly incorrect.
It’s 10x higher than some minimally competent level, not 10x higher than the worst in the field.
I think the original studies (certainly later follow-up studies did this) excluded results from participants who did not complete the assignment at all. That makes some sense from a data analysis convenience standpoint, but truncating the left tail and then saying some are 10x better than the absolute worst (that you truncated in the previous step) isn’t fully representative of what we see in the field.
In any case, people on here tend to be extremely sceptical of social science results, and demand much better standards and reproducibility.
Except that particular 1968 paper, which is held up and treated as gospel, because it helps to confirm rather than challenges their personal experience.
If we are going to demand better, we should do so consistently. Reproducing that result with more rigour would be a useful start.
I group the skepticism that I see around this topic into two buckets.
One is "it's not 10x, but is maybe 4 or 5x"; the other is "if differences exist at all, it's definitely less than 2x".
I get/have sympathy and agreement for the first type on a technically-precise level, but it also doesn't really change any of my behavior as a technologist or people/org leader.
The second type of skepticism I just don't understand at all from people who have worked more than a couple years in industry.
> The reality is that the study was looking at the difference best and worst performers, not best vs average
This is a myth, the worst performers didn't complete the task or delivered a faulty solution, those were removed from the study so you didn't see the worst.
Best to worst would be much much larger than 10x, however finding anything larger than 10x would take a very long time as it means waiting around for the slowest to finish.
Edit: And if you add back all of those worst performers, best vs average would also shift significantly towars 10x.
The "myth" is that there was a formal study done and these performers were identified. The reality is that the study was looking at the difference best and worst performers, not best vs average, which is an important clarification.
This has since evolved to a further myth that 10x programmers exist, but are somehow always harmful and should be avoided. There are plenty of articles examining "The 10x programmer myth" which, with no more evidence than the original assertion, put forward that 10x programmers should be avoided.
And here in this article, it gets re-written again, that they aren't even 10x as productive, and just produce ten times as much in terms of LOC but it's all terrible code!