3 ms·
A sports analogy appeared in yesterday's HN thread on this topic: the difference between an average runner and the fastest man alive is less than 50%. The count
by alkonaut 12y ago
A sports analogy appeared in yesterday's HN thread on this topic: the difference between an average runner and the fastest man alive is less than 50%. The counter argument here is that the measurable variable here isn't really the skill of the programmer but the success of the company (or equivalent). You only have to be best by a small margin to win. If you count wins between me and Usain Bolt he would be more than 10x as successful. Now you could argue that a lot of programming is just factory work; you can complete it in 20% longer time and your salary is lower by some factor. The truth is somewhere in between: a company with developers and other staff that are better than their competition by some factor, will go on to be a successful company a lot more often than others. The measurable business result could be thousands of times better.
So what is being measured? We can't measure "skill" separate from experience, but even as a compound the interesting measure is the success of what you produce. If your product or service is earlier on the market or slightly better, you might have 10x the return of the "average".
- bluedino 12y ago>> the difference between an average runner and the fastest man alive is less than 50%. So that means the average man can run the 100m in 15 seconds? It's a much better comparison to stretch that distance out. How fast (slow?) does your average human take to complete a marathon?
- raverbashing 12y agoI think it's an ok comparison but remember this is not a normal distribution (I'm not sure what's the actual distribution) How many people do it in less than 10s? A lot of people probably do it in more than 20s
- alkonaut 12y agoPrecisely: the distribution of 100m sprint times is a curve with a small number of people below 10s. The question in programming is perhaps: are there people who should be paid 10x the average developers? Maybe such a claim is where the myth of the 10x programmer comes from? Comparing with sports: it's economically sound to pay a footballer who is marginally better than another 100x more because the wins count, not their performance! There are many arguments as to why someone with 4x problem solving speed doesn't actually provide 4x the business value, but in fierce competition a tiny individual performance advantage can mean a huge business advantage.
- DougBTX 12y agoFor track runners, 12-13 seems to be regarded as slowish, so 15 might not be too far off. http://training.fitness.com/interval-training/what-average-100m-sprint-time-21720.html http://training.fitness.com/interval-training/what-average-1... Also note that the parent said "average runner" but you replied with "average man", most men are not "runners", so that is shifting the goals quite a bit.
- alkonaut 12y agoYes, the definition of "runner" is somewhat blurry but you could argue that you need at least some training to qualify, just like the 10x programmer is compared to an average programmer, not the entire population.
- ebiester 12y agoA relatively fit person ran it in approximately 15 seconds without training. High school training in track brings it down to 12, from when I looked it up in the last thread. So the difference between someone in high school and a professional runner is 12 to 10 seconds, approximately, and the difference between professional and elite is .5 seconds.
- clarkm 12y agoExactly. Just like not everyone can run an ultramarathon, not everyone can program fizz buzz.
- snowwrestler 12y agoThe parent says "average runner" and you said "average man." There might a significant difference there.
- VLM 12y agoIn the spirit of providing reasonably good analogies, there's a game some people play where given a programming language the scissors are unleashed and a statement or construct disappears from the language temporarily and now the players try to implement "whatever" in the now slightly neutered language. The general spirit of ... write me a fizzbuzz without the mod operator or a function reimplementing mod. Now there's plenty of space for jokes like do this to C++ and that language is so fat most people won't notice if a random component disappears. Or do this to PDP-8 assembly language or a turing tarpit and the language is so small that one component removal might make it no longer turing complete, maybe impossible to complete the challenge. But generally speaking, yeah, remove a big part or enough small parts and you'll take a 10x performance hit, sure. Now there's "know" and then there's kyu/dan ranks from go. Too much binary thinking means people who have a correctly applied checkbox don't really know what they're doing, or maybe they do. Thats why its hard to predict.