Preliminary Results of Typeracer’s “Typo Analysis” Project
As I’ve written before, part of our vision is that Typeracer be more than just a game; we hope to make it an effective and fun tool for improving your typing speed. This is one of the main ideas that inspired Typeracer founder Alex Epshteyn, who decided to create Typeracer because he thought he typed too slowly!
We are confident that just playing Typeracer will help you speed up your typing while having fun. But, just saying “play Typeracer more” is not a satisfying answer for some people who are serious about improving their typing speed, and of course that’s not the perfect answer
So, we are currently working on developing the best typing tutorial available on the internet. That’s the goal, at least. One of our first steps, aside from talking to some of the best typists in the world, is to try to understand what slows people down. One of the biggest problems for most people, of course, is that they make typos. I am determined to figure out why. In order to answer this question, I first need to know what mistakes people are making.
So, I did some research. Dozens of Typeracer users participated in my experiment, in which they basically did some typing without fixing their typos, and sent me the results. I’m a bit slow, and haven’t finished yet, but here are some of the initial conclusions/results.
1) About 15-25% of all mistakes are on short words (3 characters or less). Words like “the” and “of” and “is” are words we’ve typed a million times before, but we still make a lot of mistakes on them. This is probably a result of people trying to type faster than they can actually handle. Also, for the fastest typists (approximately 90wpm and above), mistakes on short words made up a larger percentage of their total errors (25%, instead of 16% and 17% for the other two groups).
2) About 10% of mistakes occur on long words (8 or more characters.) Of course, long words are less frequent, so we might expect fewer mistakes here, but I still thought this was a bit surprising.
3) About 15-30% of typos are omissions of a single letter, such as typing “mjor” instead of “major”. Typists in my middle speed group (approximately 70-90wpm) made the most mistakes here (30%).
4) Nobody had a big problem with punctuation or capitalization. Each of these problems accounted for just 2-5% of typos for all users.
5) About 10% of the mistyped words are existing English words. For example, some people typed “to” instead of “too”, or “this” instead of “his”, or “build” instead of “built”. This seemed surprisingly high to me, and it suggests that people’s mistakes are sometimes more mental than mechanical.
6) Slower typists are much more likely to accidentally hit a neighboring key. For example, instead of typing “civilization”, one typist typed “cibilization”. For my slowest group (around 40-60wpm), 21% of their errors were of this kind. For the medium group, this number was 9%, and for the fastest group, it was only 5%. Not surprisingly, I suppose, these slower typists have more mechanical problems, and the fastest typists almost never have these problems.
7) The biggest mistake for my fastest typist group (by far) was adding extra letters to an otherwise correctly-typed word. I participated in the experiment myself, and typed “shasll” instead of “shall”, and “arazor” instead of “razor”. For my fastest typists, 46% of their mistakes were of this kind! Also, I think I know where these extra letters came from: 63% of the time, the added letter was a letter that also appeared elsewhere in the word (most common), or in the next word (also very common). For example, instead of “through”, one user typed “throught”, so that letter “t” got typed twice. Or, instead of typing “spin and”, one user typed “spind”, so the “d” from the second word got typed twice. It was not common for typists (especially the fast ones) to incorrectly add a totally random letter.
8) Slower typists are more likely to have multiple, consecutive errors. Ten percent of the errors of my slowest typists occurred in a series of incorrect words, but for my fastest typists the number was half. Fast typists seem to catch their mistakes more quickly, not letting themselves get caught up in “bad momentum”.
9) Fast typists are much more likely to make mistakes on rarer words (this was kind of subjective, but I used my best judgment.) My medium and slower typists made only 1% of their mistakes on these less commonly-used words like “etiology” and “mozzarella”, but 7% of the mistakes made by the fastest typists were made on these words. This provides support for what I have always suspected, which is that one reason fast typists are fast is because they have familiarized themselves with the common letter combinations and words–maybe just by typing a lot. So, when they type a word they might have never typed before, they have more problems and can’t keep up their super speed.
10) Fast typists are most likely to have problems with single letter reversals. For example, they might type “teh” instead of “the”, or “refelctions” instead of “reflections”. This kind of mistake accounted for 34% of the mistakes made by my fastest typists, but for the other typists, it was just above 20%.
These are just some of my tentative conclusions. I need to do more analysis, and probably collect more data (if I have time). If there are any statistics buffs out there, or anybody wants to see my data, let me know.
Our next step is to try and see how this insight can translate into practical suggestions that can be implemented, in order to help people type faster.
David Pritts (valikor)