How (not?) to practice typing
For a couple months last year, I spent 20 minutes a day deliberately practicing my typing to see how much faster I could go.
Why?
Having spent most of my life online, I was already a very good typer, averaging about 95 words per minute at the start of this exercise. But anecdotally, I've known people who type around 120 wpm (normal people, not professional typists or people who modify their keyboards to chase record times). Since I haven't actually practiced typing since Type to Learn in third grade, it seemed possible that I could improve.
And typing faster would be useful, because I type a lot—especially for work, but also for hobbies like this one. How useful? Let's say 50% of my work involves typing, whether that's coding, writing design docs, or messaging on Slack. And of that time, let's say only 20% is spent actively typing, leaving 80% for thinking about what I should be typing next or reviewing what I've already written. If that's the case, going from 100 to 120 wpm would permanently increase my productivity by 2%, and in a very general way that would translate across most possible career paths.1
There are countless ways I could make myself 2% more productive, but typing is uncommonly well-suited for deliberate practice:
There are proven techniques to learn: others already type measurably faster than I do.
There is immediate feedback: online tools can tell you how fast you type while you're typing.
Improvement is quantifiable: all that matters is typing more correct words per minute.
You can increase difficulty over time, such as by practicing harder words or words you’ve mistyped before.
So improving my typing seemed both important and achievable. And at a meta level, I wanted an opportunity to practice the discipline of "deliberate practice."
How?
I tested a few online tools, but I settled on Monkeytype for having the nicest and most customizable interface. Every Friday, I benchmarked my progress with a fixed series of tests: how many words could I type in 60 seconds, how many errors did I make when trying to maximize accuracy, and how fast could I create a 100-word essay from scratch. For the rest of the week, I did some general practice along those lines but also mixed in more specific experiments: how did my accuracy change at different speeds, how did my speed change if I used different fingers, what happened if I changed keyboards, etc.
I also researched advice on typing faster, but that well ran dry after only a couple days: there wasn't much credible advice out there at the level I was looking for (most advice was how to go from slow to average, or from super-fast to record-breaking-fast). So I mainly focused on learning from my own practice—noting the errors I made, looking for patterns among them, reflecting after each session, coming up with new hypotheses and testing those.
Did it work?
Not really:
If you fit a straight line through the data, it slopes upward, showing an improvement of 5wpm over a month and a half. But most of that is driven by low performance on day one when I was still figuring out how to use the tests—take that away and the increase is only 2wpm, and quite possibly just noise.
Trying new techniques also didn’t really help. One suggestion was to use “alternate fingerings” for consecutive letters you would normally type with the same finger. (For example, instead of typing "ed" entirely with your middle finger, use your index finger for that d instead.) This allows you to move your second finger into position while you're typing the first letter, saving a fraction of a second each time. The theory sounded good, but in practice it was really hard to learn.
To focus my practice as much as possible, I focused on just the "ed" pair, and each day practiced this a few times on a list of words that all contained those two letters in order. And on the surface, it seemed like I was getting better: over a month and a half, my speed on those words increased by 10 wpm! But as a control, I also kept testing myself on the "baseline" strategy of always using the same finger for each key ... and it turned out my performance there was also increasing at the same rate:
I was surprised by this on a couple of levels. First, given that my overall typing didn't speed up that much (in the first chart), why did my baseline go up in this test? My best guess is that because this was a smaller set of words, the same ones kept popping up over and over, so I became much better at typing those words in particular.2
Second, it sure felt like the alternate fingering was faster—when I switched back to normal, the tiny pauses when my middle finger jumped from the E to the D felt eternal. And when I play the piano, I use different fingers for the same note based on what’s fastest, so why wouldn't the same technique translate to typing? I bet the alternate fingerings would eventually catch up and surpass the baseline given enough practice, but my 25-year head start practicing the normal way was too much to overcome.
Another failed hypothesis: I briefly became obsessed with the keyboard itself, because I did much better the first time I tried a new one with a different weight. But subsequent tests showed that was a fluke, and I was pretty much the same no matter what keyboard I used. (Even my posture didn't really matter, contra the advice I read; I could type about as fast on the couch as at my desk.)
What did I learn about typing?
The first letter of each word was especially hard. A neat feature of Monkeytype is that it shows you each time you make an error and go back to correct it, so for a while I tracked and classified each of these errors. The rough breakdown was: 50% of the time I just typed a completely wrong letter; 25% of the time I missed a letter I should have typed; and 25% of the time I swapped two letters out of order. Within each of the first two categories, at least 1/3 of the errors came on the first letter of the word—much more than you'd expect from chance.
However, I'm not sure how much to make of this, because...
It's really hard to test typing in the wild. On the “type random words” test, the gating factor quickly became not finger speed but reading comprehension—I caught myself starting to type the next word before I actually understood what it was, leading to a lot of errors (and especially a lot of first-letter errors). If I was trying to master a random-word-typing test, this would be a great thing to work on. But for the real world it’s not very important—I already know the words I want to type.
I eventually switched to Monkeytype's other tests that give you real quotations (which are easier to process because each word logically follows from the last). These are harder to benchmark, though, because they have a lot more variance: some have lots of short and easy words, others lots of long words or proper nouns.
My typing output is pretty variable. For a skill that feels pretty automatic, it was surprising how much my typing speed varied from day to day. There were three or four days on which I was running a solid 10wpm below average and felt like I just didn't have it at all.
There are really internet communities for everything. I didn’t find it very useful, but in an anthropological sense I was fascinated by the Monkeytype Discord server: there are 40,000 members and the main channel alone easily has multiple thousands of messages a day, all just on the subject of typing fast. You can even read several Google Docs with insane formatting that span dozens of pages about one person's strong opinions on how to type faster, although I don't recommend it.3
What did I learn about deliberate practice?
Getting better really takes time. I would have probably needed at least an order of magnitude more than the 15 minutes a day and/or the two months I spent on this to make real progress. (Deep down I probably knew that going in, but I wanted to see the results). The upward trajectory on "ed" words suggests that if I'd spent that much time focusing on every other type of word as well, I might have made similar progress.
You need to focus on weak points. Just paying attention to my typing didn't really seem to move the needle very much; testing specific hypotheses like fingerings and keyboards seemed more productive, even if they didn't ultimately lead to any silver bullets. If I was continuing onward, I'd try to come up with tests that really worked on the places where I made the most errors (e.g. skipping a letter).
This is hard! I knew going in that deliberate practice takes effort, but experiencing it firsthand felt different. Just doing a benchmark test felt easy, but every time I practiced the alternate fingerings really felt like a chore. So this type of practice takes real dedication—you need the mental bandwidth to actually focus on it, and that bandwidth comes at the expense of other priorities.
Ultimately, that's why I stopped practicing after two months: my improvement was slower than I’d hoped (perhaps because I'm closer to the practical ceiling than I expected, or because 25 years of typing have made too stuck in my ways already), and it didn’t seem worth the level of effort it required.
As Steve Yegge once wrote (quoted here), "the most important course a computer science student can take is Typing 101."
You may also notice that the starting baseline wpm for this test was much lower—that’s because the “ed” words tended to be longer and/or trickier.
Substacks that span thousands of words about one person’s data-driven opinions on typing are highly recommended, of course.