GOLEM keyboard project

What's wrong with QWERTY?

QWERTY is the default logical layout for most people using Latin characters. However, it was designed for the mechanical typewriter of the late 19th century and the considerations its inventor had to make are simply no longer necessary or valid. Other than that, it's not only inefficient but can cause health issues too.

What was the purpose of QWERTY?

Most of the time you'll hear it was designed to prevent jamming of mechanical typewriters. While this may be true, I didn't find any evidence for this. Surprisingly, this concept seems to be quite new and was not even mentioned in early books about Christopher Latham Sholes and his work.

Nevertheless, neighbouring metal levers of the typewriter (also called hammers, type bars or strikers) were prone to jam, that's why it's said Sholes moved frequent character pairs away from each other. Is this true? See the table below.

Pic: Typewriter with metal levers

Since touch typing wasn't common in these early days of typewriting, we should probably look for slow/fast key pairs rather than close/far away keys. Thus, the goal could have been the prevention of rolls, the quick consecutive keypresses we love today.

th 1.52       en 0.55       ng 0.18
he 1.28       ed 0.53 (!)   of 0.16
in 0.94 (?)   to 0.52       al 0.09
er 0.94 (?!)  it 0.50       de 0.09 (!)
an 0.82       ou 0.50 (?)   se 0.08 (?!)
re 0.68 (?!)  ea 0.47 (?)   le 0.08
nd 0.63       hi 0.46 (?)   sa 0.06 (?!)
at 0.59 (?)   is 0.46       si 0.05
on 0.57 (?)   or 0.43       ar 0.04 (?)
nt 0.56       ti 0.34       ve 0.04 (?)
ha 0.56       as 0.33 (?!)  ra 0.04 (?)
es 0.56 (?!)  te 0.27 (?)   ld 0.02
st 0.55 (?)   et 0.19 (?)   ur 0.02

Taking a look at the list above, where I've marked the most frequent English bigrams I would type with rolls, I'm not convinced at all that QWERTY was designed to slow down these moves.

Almost half of the most frequent character pairs are still close enough to type them rolling your fingers.

Anyways, it's fine either way. If QWERTY can't even do what it was designed for, it's a strong argument for looking for something better.

The point is, we still have the obsolete QWERTY and horizontal staggering on common keyboards because of the long gone levers. (Btw, there were never any levers in computer keyboards I guess.)

It seems the main idea behind the most common (and default) arrangement of characters lost its reason of justification decades ago. The very problem QWERTY was meant to solve doesn't exist anymore.

On the other hand, we have problems and development goals which should be addressed: speed, finger travel, ergonomics, etc.

How bad/good is QWERTY?

Now that we know that QWERTY is obsolete, there's one more important factor we should know before embarking on the arduous mission to design or learn something new: how bad it is actually?

Of course, it depends on your language, the national variant you use, your typing habits, etc.

I made a simulation with the QWERTZ layout (a common QWERTY-variant in Germany and Central-Europe swapping Z and Y), a large Hungarian corpus and my Golem evolution model, and the performance of QWERTY was good for the top 1.82%.

Wait, what?! QWERTY is in the top 2%? That doesn't look so bad at all. It beats 98.2% of all the layouts. Why the hassle then to come up with alternative layouts?

Well, just imagine the number of potential layouts. I optimized for 38 chars on the base layer. The number of possible layouts in this case is 38 factorial, which equals to 5,23*10^44.

1.82% of this is 5.13*10^42. That's the number of layouts performing better than QWERTZ.

Let me help you imagine that:

5 130 000 000 000 000 000 000 000 000 000 000 000 000 000

So QWERTY beats 98% of RANDOM logical layouts, but I'm sure we can find a much better one if we put some effort in the optimization instead of just trying random ones.

Intuitive & genetic optimization

The aim of making my Golem evolution model was exactly this. To find an optimal logical layout for myself (typing mainly in Hungarian and English).

Again, the exact numbers will depend on your language and many other factors, but FYI:

To type the corpus I used for the calculation above, your fingers would travel 314 km. With some optimization this finger travel can be reduced to 197 km. Hmm, not bad.

Similarly, with QWERTZ the number of row changes is 22% vs. 10% optimized. Base row keypresses: 33% vs. 53%. Home row jumps: 6.6% vs 1.3%.

Of course, you don't have to write your own modeling environment. You can swap characters by simply looking at a heat map based on your letter frequencies. This is called intuitive logical layout design and it may work as a first step.

(Just try these three swaps as an initiation: E-D, T-F and N-J.)

You can easily move the most frequent letters on your home row this way. However, intuitive optimization probably won't work for more complex rule sets.

Alternative layouts

Even Sholes, inventor of the QWERTY layout, knew that he can do better. And he did! He approached Remington, the company popularizing his layout with a revised key arrangement. (They were not interested.)

Many more people tried to invent better logical layouts, the most notable being probably August Dvorak and his simplified keyboard.

Other well-known contemporary alternatives are Colemak, Workman, etc. for English, but alternative layouts are also available for other languages: e.g. NEO for German, BEPO for French typers – or the impressive Turkish F-keyboard from 1955!

Custom layouts

Whichever layout you choose, consider finetuning it. You will find that large corpora used to compare layout performance (mostly classical literature) are not representative of your personal typing habits at all.

Other than that, unless you are a writer churning out novels all day, very little of your working hours is spent with continuous typing. E.g. programming is rather editing. Or optimizing for hotkeys (graphical design, audio/video editing) may require a whole different approach too.

Keep this in mind.

Also, tweaking your layout is addictive! So be very careful. :)


QWERTY is obsolete. Whether you choose one of the popular alternatives like Dvorak, Colemak, Workman etc. or find your own unique layout, you better do it as soon as possible, because it will boost your typing efficiency.

Mr. Sholes would be the one most surprised by the fact that we still use his age-old layout.

Btw, I'm planning to publish my tools and to make a comprehensive video series about using them for layout optimization. The first step for this is raising funds for a proper audio setup on Buymeacoffee. Consider helping me with this project if you can.