I find the claim that typing at 120 WPM "borders on the physically impossible" very funny.
The study they are quoting is based on data from 1993-1997. While the numbers feel bit low even for that era, I would expect noticeable improvement these days.
The big error rates are less significant than what the author thinks. Backspace was disabled in the test, and I imagine vast majority of the errors would have been corrected on the by the typist had they access to backspace. Of course one might also raise the question how much did the disabling of backspace disrupt the typists flow?
One only need look at sites like Typeracer to see that there's plenty of people doing "the physically impossible" now. The funny thing is, that site has a CAPTCHA for keeping out script cheaters, and I often score much higher on that than the races --- I've achieved 240WPM with 95% accuracy on that, whereas my usual speed is in the 140-150WPM range.
The keyboard has a big effect on speed too --- in the past decade or so, low-travel and low-force keyboards have become significantly more popular because they're less tiring to type on.
120 WPM (at even 5 characters per word) would put you with the very highest actions per minute gamers in the world in terms of APM - and they are just spamming keystrokes and clicks. Now, with both hands on the keyboard you could enter more keystrokes, but you would have to be accurate as well. World record is over 200 wpm, but if you're over 120, you're amongst the fastest. Here's a competitive typist only hitting a little over 120 on some exceptionally difficult text: https://www.youtube.com/watch?v=4GDusA21cEA
Except that each "action" is one click. Whereas a word corresponds to ~5 clicks, assuming 5-letters-per-word average. This means that 120 WPM is analogous to 600 APM.
Yes, but that word is a sequence of actions in muscle memory, so you can't really think of it as being the same as an "action" in the sense of clicking on a unit or issuing a build command. It's like the difference between streaming I/O and synchronous I/O.
You could say the same thing about issuing gaming actions. Selecting a hatchery, then selecting larvae, and then building 3 drones, are all sequences of actions in muscle memory, analogous to typing one word.
> IMO, any errors on a typing test should fail it. Though this is likely somewhat biased from an older software dev.
My IDE will flag my error and offer a correction if I mistype a function or variable name; it'll offer autocompletion options from the first character. It'll warn of syntax errors in real time and provide niceties like syntax highlighting, automatic indentation and automatic bracket closure. Typos do occasionally cause grief even in a modern development environment, but they're a relatively minor issue. I don't think I'd notice the difference if somehow I could reduce my typing error rate by an order of magnitude.
Text-to-speech is horribly cumbersome for code, especially if you're using a language with a profusion of curly braces and semicolons.
Look up tables can sometimes just be a huge array of numbers. 137295.276 vs 137395.276 is the kind of issue I am talking about.
But really any kind of error even if the IDE catches it is still a distraction. As to errors and speech recognition I am more talking about text messages or online posts not code.
The study they are quoting is based on data from 1993-1997. While the numbers feel bit low even for that era, I would expect noticeable improvement these days.
The big error rates are less significant than what the author thinks. Backspace was disabled in the test, and I imagine vast majority of the errors would have been corrected on the by the typist had they access to backspace. Of course one might also raise the question how much did the disabling of backspace disrupt the typists flow?