anishtro is a layout for the letters of the English alphabet, made for symmetrical keyboards with at least one main key per thumb.
qlucj kpmw·
anisv bhtro
··ygx zfd··
e ·
The dots above represent keys that anishtro doesn’t say anything about. Here’s a suggestion, though (␣ means space):
qlucj kpmw-
anisv bhtro
/;ygx zfd.,
e ␣
The layout is suitable if:
-
You agree on the following rating of the keys. The lower the number, the better the key. Mirror it for the right hand.
64336 22225 66447 1In other words, the thumb keys are the absolute best, closely followed by the home keys. The index fingers’ and middle fingers’ keys are quite good in general, while the ring fingers’ and pinkies’ are worse. Worst of all are the center colums operated by the index fingers by reaching towards the middle of the keyboard.
-
You want a letter on a thumb key (one of the absolute best keys, as seen above).
-
You like to type punctuation with your ring fingers and pinkies. The bad ratings for their keys above are mainly for typing letters, which are part of a “flow”—as opposed to punctuation characters, which are typed more or less one by one.
-
You agree on that the single most important characteristic of a layout is to have all the most common letters on the home keys.
-
You agree on that the second most important characteristic of a layout is to avoid using the same finger twice in a row.
-
You are looking for a layout that is “fully optimized,” not a compromise between comfort and QWERTY similarity. This also means that you are ready to re-learn your typing from scratch, since only two letters remain on their QWERTY positions: A and Q. (You can swap G and V if you want yet two, though.)
-
You like “rolling” (horizontally) between the keys marked with uppercase letters, in either direction. (Again, mirror it for the right hand.)
qLUCj ANISv ··YGx -
You don’t care about other metrics that some other keyboard analysers use, such as “finger travel distance,” “hand balancing” and “bottom row penalties.” (“Hand balancing” is especially unimportant, in my opinion, since to compare the workload between the hands also requires taking all non-letter keys and keyboard shortcuts into account. The only way to measure that is to use a key logger for a long period of time. It is better to adjust other keys than letter keys if needed, that are not part of “typing flow”.)
-
You are looking for a more symmetrical keyboard layout; where both hands are used for punctuation, and the same fingers on both hands do that job.
-
You think it’d be fun!
The assumption that you can mirror the above diagrams for the right hand is the reason that a symmetrical keyboard should be used.
I strongly recommend putting space on a thumb key, which is why at least one key per thumb is needed.
anishtro was made in the Spring of 2014. In the Summer of 2015, I found an even more optimal layout, which I call anishtro2:
·lucj kdgwq
anisv bhtro
··yfx zpm··
e ·
However, the differece is so small that I cannot even be certain that it actually is better. It exists mostly for the sake of completeness. Between the Spring of 2014 and the Summer of 2015 I’ve used anishtro and never looked back. I don’t think it’s worth re-learning.
Shortly after that, I used two different data sources and came up with yet a variation (called anishtro3):
·pucj kldwq
anisb vhtro
··yfz xmg··
e ·
It seems almost as if endless variations can be made, all with the same base:
· u c j k w q
a n i s bv bv h t r o
· · y fg xz xz · ·
e ·
The keys with two letters should have either of those two (depending on how the others are placed). The empty keys should have one of D, F, G, L, M and P.
Which one of all of those variations is the best is difficult to say, perhaps even impossible.
There are a couple of changes that you could make without reducing the amount of “rolls” or increasing same finger usage.
- Mirror the entire layout. (If you’re looking for as much QWERTY difference as possible, this is the way to go!)
- Swap E and Space. If you’re used to typing space with your left thumb since before, why re-learn?
- Swap G and V.
- Swap J and X, as well as K and Z, if you don’t agree on the relative rating of those keys.
- Swap the top and bottom row.
- Move Q to above O. That’s actually what the layout should be by default, since AQ is more common than OQ. However, that difference is so marginal that I saw no point in re-learing the position of q from the QWERTY one. I know that I claimed that this layout isn’t a compromise between comfort and QWERTY similarity, but hey, no rules without exceptions! (Note that anishtro2 includes this change.)
Actually, that question is rather pointless. This project is not about creating the “best” layout ever, that obsoletes all other. The reason it was made was because I couldn’t find any layout that fulfilled the things mentioned in the “Who should use it?” section. Comparing anishtro to other layouts from that point of view obviously makes anishtro the winner. Using another analyser, with different opinions, might make anishtro look really crappy.
If you like numbers, though, here is a table that compares a few well-known layout using my analyzer (see the “Tools” section below, and Analyzing a full layout).
English Letter Frequency Counts: Mayzner Revisited or ETAOIN SRHLDCU by Peter Norvig is an analysis of English letter frequencies using the Google Corpus Data.
Michael Dickens has a repository that contains a program to optimize keyboard layouts, as well as some English letter frequency data.
For comparison with my own data, I also used their calculated bigram frequencies. Having three different data sets made by three different people in different ways all giving about the same results makes me confident that the numbers are reliable.
| Layout | Same finger | Rolls | S.f. Norvig | R. Norvig | S.f. Dickens | R. Dickens |
|---|---|---|---|---|---|---|
| QWERTY | 7.78% | 6.67% | 6.94% | 6.68% | 6.87% | 6.54% |
| QGMLWB | 5.42% | 4.04% | 4.62% | 3.94% | 4.86% | 4.16% |
| Dvorak | 2.85% | 5.30% | 2.50% | 5.88% | 2.61% | 5.63% |
| Workman | 2.86% | 8.54% | 2.82% | 8.97% | 2.94% | 8.75% |
| Colemak | 1.35% | 7.99% | 1.07% | 7.68% | 1.25% | 7.44% |
| Arensito | 1.28% | 13.64% | 0.95% | 13.85% | 1.05% | 13.27% |
| Maltron | 1.04% | 13.67% | 0.73% | 13.33% | 0.85% | 12.83% |
| anishtro | 0.89% | 14.65% | 0.70% | 14.38% | 0.80% | 13.73% |
| anishtro2 | 0.88% | 14.66% | 0.69% | 14.39% | 0.76% | 13.74% |
| anishtro3 | 0.94% | 14.50% | 0.68% | 14.17% | 0.74% | 13.66% |
“Same finger” (abbreviated “S.f.” in some columns above) is the percentage of all your keypresses that will be performed by the same finger (excluding typing the same letter twice in a row), which tells how bad a layout is.
“Rolls” (abbreviated “R.” in some columns above) is the percentage of all your keypresses that will be part of a roll, as described in the “Who should use it?” section. This is not essential, but a big bonus. (When creating anishtro, I expected having to choose between less same finger usage and more rolls, but it turned out that anishtro both got the least same finger usage and the most rolls. How convenient.)
Remember that the above table is made from my ananlyzer, with my opinions (as detailed in the “Who should use it?” section). For example, QGMLWB takes loads of other things into account, Dvorak favours hand alternation (which some like and some don’t) and Colemak is an optimization between optimal layout and QWERTY similarity. The table is only useful if you agree with my opinions and wonder how well one of the above layouts would fit you.
Also note that Arensito seems to be almost as good as Maltron and anishtro, but it does not put a letter on a thumb key, making it have one less home key. In the above list, only Maltron and anishtro do. That’s their big advantage.
First off, you need Node.js.
See scripts/repl.js for how to interactively play with data to analyse layouts.
Before using that file you need to run npm install to get all dependencies.
You also need some data to work with. You may download the data I used, or see ./create-data for how to create your own. Norvig’s data and Dickens’ data are also available. Or do something entirely different.
It might be a good idea to have a look at the two dependencies as well: wikipedia-gather and text-frequencies-analysis.
scripts/layout.js is the file that does the actual analysis of layouts. Have a
look at it for more information, and run npm test to test it.
If you have the time and interest, you may read a very long explaination in How anishtro was made.
I also recommend reading the Most Ergonomic Keyboard ; Most Efficient Keyboard article by “Joe,” since it was a big inpiration.
The layout and all files used to build it is in the public domain (unless otherwise stated).