What is the most optimized layout recently?

[knowsnokb]

With programmable keyboards,and the ability to put any key where you want, what layout/s are the most optimized for...

1) english writing (novelist, bloggers)
2) programmers

Have they been scientifically tested? or at least have a lot of supporters?

And let's ignore the time it takes switching from qwerty or whatever layout you have gotten used to.

Just wondering what layout, for english writers and programmers have been shown, and studied to be the most optimal.

[davkol]

That's simple: Dvorak Simplified Keyboard and Maltron layouts.

Colemak isn't interesting, if you don't care about transition from QWERTY or shortcut compatibility; its typing properties are very similar to Maltron THOR, except for vowel separation (and usage of a thumb for a letter). The closest thing is MTGAP or carpalx, but that's still the same in principle, only with slightly different (arbitrary) design criteria and even less usage in practice.

[Snarfangel]

That's simple: Dvorak Simplified Keyboard and Maltron layouts.

Colemak isn't interesting, if you don't care about transition from QWERTY or shortcut compatibility; its typing properties are very similar to Maltron THOR, except for vowel separation (and usage of a thumb for a letter). The closest thing is MTGAP or carpalx, but that's still the same in principle, only with slightly different (arbitrary) design criteria and even less usage in practice.

I tweaked MTGAP software a bit (well, cheated by taking space and e out of the equation and putting them on the thumbs), and am currently using the following at home on a Kinesis:



I may move the Q under the I on the left, then tweak the punctuation a little (move double quote where Q was, and hyphen where double quote was, plus a few other minor changes). I'm also not completely sold on the numbers, specifically the shift 6789, though I like having all of them on the right. I've gotten so I can type without my cheat sheet, though slowly, and the rolls from one letter to another are fun.

[bcredbottle]

Ergodox. I have shift, ctrl, and space on the thumb keys so I don't have to leave home row/contort pinky every time I start a new sentence or want to copy paste.

[jacobolus]

[...] Have they been scientifically tested? or at least have a lot of supporters? [...] Just wondering what layout, for english writers and programmers have been shown, and studied to be the most optimal.

There has been no scientific testing, and not much informal study either, any time in the past ~30 years. It’s all just people’s personal preferences/biases, and their computer optimization programs built on ad-hoc untested heuristics.

[oneproduct]

[...] Have they been scientifically tested? or at least have a lot of supporters? [...] Just wondering what layout, for english writers and programmers have been shown, and studied to be the most optimal.

There has been no scientific testing, and not much informal study either, any time in the past ~30 years. It’s all just people’s personal preferences/biases, and their computer optimization programs built on ad-hoc untested heuristics.

Heuristics are indeed the name of the game here. There is no such thing as optimality in terms of keyboard layouts as it's impossible to say what the most important factors are. For example, Dvorak favors alternation, Colemak favors inward rolls and Workman disfavors lateral movements.

Certain people have different preferences as to what is comfortable, and some of it is based on their physiology, i.e. the size of their hands and the dexterity of their fingers. Rolling motions are also debated in terms of what is and what is not comfortable in terms of which particular fingers on the same hand perform them.

A lot of people (myself included https://geekhack.org/index.php?topic=67604.0 ), have tried making their own layouts based on what heuristics they personally deem to be important, but it's very difficult to prove that a layout is better or worse, and what works for one person doesn't work for others.

Personally speaking, I started with QWERTY, tried Dvorak and then ended with Colemak (not even really trying my own proposed layout seriously). For me Dvorak just didn't click. I find that too much alternation is a bad thing as it requires stricter organization between the left and right hand, resulting in transposition errors (two letters in a word swapped). Colemak's rolls are quite comfortable to me. And even if it turns out that Colemak is slightly less optimal in terms of what I could achieve in terms of typing speed, there's a human comfort factor which may be more or less valuable than sheer typing speed.

If you just want to look at numbers, I could argue (but I wouldn't do so overly seriously) that the layout I came up with in the thread I linked is "the best" because it has a higher alternation % than MTGAP and Dvorak as well as a higher inward roll % and low outward roll %. It's also almost exactly 50/50 in terms of left/right balance. And I have that one other reference which rates MTGAP and my own layout as close contenders for the top spot. MTGAP probably has lower travel distance than my layout, which gave it a tiny edge. However, PieterGen runs my layout through another layout analyzer which looks at bigram and trigram optimization (trying to optimize for rolls of size 2/3) and rates my layout slightly lower than ADNW, a layout I'm not too familiar with.

There's also some funny things to consider which most people don't, such as the optimum placement of the backspace key, which is dependent on each person. My error % when typing is around 5-8%, which is more common than quite a few letters, so it would make sense to give the backspace key a prime position relative to a lot of letters.

And then there's human factors such as having number keys in ascending/descending order rather than by frequency (0 and 9 for example are likely unnaturally common relative to other numbers due to things such as 1000 or $19.99) and having all the letters be adjacent to each other rather than mixed in with modifier keys (shift, ctrl, alt). Most people have this as a requirement when generating their layouts.

The physical hardware you have available is also a deciding factor. Above someone has MTGAP on a Kinesis Advantage split keyboard which most people won't have, which grants two thumb clusters. The Workman layout also gets much of its charm due to the fact that on a standard keyboard there's a lot of lateral movement on the index fingers, but that could be solved with a different physical keyboard layout rather than a different key layout.

[tuffff]

There is the ADNW project (adnw.de), which was created for german but may still be interesting for you. They have a few alternative layouts (http://adnw.de/index.php?n=Main.Varianten), but most importantly two pieces of software:

- A very nice and configurable optimizer (http://509.ch/opt.7z). You can configure your own hardware, set preference parameters for each finger, hand alterations, rolls and much more and insert your own corpus. You could even change the letter list. Let it run for a few hours and you got a personal optimized layout.
- An impressive AHK script (http://wiki.neo-layout.org/export/2475/windows/neo-vars/out/neo20.exe) that includes 6 layers. You can personalise layers 1 and 2 (shift), layer 3 is basically made for programmers (“\/{}*?()-:” on the homerow!), layer 4 puts all navigation symbols in the 60% area and layer 5 and 6 include greek letters and some lesser used symbols. You can see the layers at http://neo-layout.org, just hover over the “Ebene x”-buttons.

[jacobolus]

The physical hardware you have available is also a deciding factor.

Indeed. The QWERTY layout was designed for the original Remington No. 1 typewriter, on which neighboring type bars in the circular “basket” would jam when typed in quick succession, so the keys are distributed so that common English digraphs are separated on that physical machine. (The arrangement of type bars doesn’t directly/obviously correspond to key location on the keyboard.)

The Dvorak layout is designed for later typewriters from the 1930s, on which the type bars hit the front of the platen, and weren’t quite as prone to jamming, but still could jam if nearby type bars were activated at the same time. The design criterion for the Dvorak layout was to improve comfort by moving common letters to the “home row”, encourage alternation between hands, and put only uncommon letters on the bottom row, because the bottom row on a mechanical typewriter is really annoying to reach, compared to the home row or top row.

The Malt layout is designed for the Maltron keyboard, which has split halves, curved keywells, and extra thumb keys.

Some later keyboard layouts are optimized for flat laptop keyboards. (QWERTY, Dvorak, and Malt are all decidedly suboptimal in the context of a modern laptop board.)

[bocahgundul]

The physical hardware you have available is also a deciding factor.

Indeed. The QWERTY layout was designed for the original Remington No. 1 typewriter, on which neighboring type bars in the circular “basket” would jam when typed in quick succession, so the keys are distributed so that common English digraphs are separated on that physical machine. (The arrangement of type bars doesn’t directly/obviously correspond to key location on the keyboard.)

The Dvorak layout is designed for later typewriters from the 1930s, on which the type bars hit the front of the platen, and weren’t quite as prone to jamming, but still could jam if nearby type bars were activated at the same time. The design criterion for the Dvorak layout was to improve comfort by moving common letters to the “home row”, encourage alternation between hands, and put only uncommon letters on the bottom row, because the bottom row on a mechanical typewriter is really annoying to reach, compared to the home row or top row.

The Malt layout is designed for the Maltron keyboard, which has split halves, curved keywells, and extra thumb keys.

Some later keyboard layouts are optimized for flat laptop keyboards. (QWERTY, Dvorak, and Malt are all decidedly suboptimal in the context of a modern laptop board.)

Jacobolous may I ask what keyboard layout did you use?

[PieterGen]

- That's simple: Dvorak Simplified Keyboard and Maltron layouts.
- Colemak isn't interesting, if you don't care about transition from QWERTY or shortcut compatibility; its typing properties are very similar to Maltron THOR, except for vowel separation (and usage of a thumb for a letter).
- The closest thing is MTGAP or carpalx, but that's still the same in principle, only with slightly different (arbitrary) design criteria and even less usage in practice.

- Dvorak is indeed very good (I admit, my subjective opinion).
- Maltron, I have no experiece with. The E on a thumb seems like a good idea. Looking at the letters per hand, I have doubts about the alternation, might be too low. But what do I know.
- Colemak seems to be quite good for English prose. I must say that the idea of maintaining one-hand key combos such as ctrl-x or ctrl-z is ergonomically not good, it is better to press ctrl with one hand and the letter (such as x, or z) with the other hand.
- I used MTGAP for some weeks, very nice layout, nice optimizer too for making your own layouts. MTGAP felt like a better Colemak to me. That is, if you like rolls but also some sane alternation, MTGAP might be the ticket for you.
- Carpalx: the analyses look nice (although a bit opaque), but in use it did not feel nice to my hands. Maybe because letters (R, I) are not on the home position, but on the querty (G, H) spots? Somehow the layout felt as though it was overly optimized on one criterium (location), at the expense of alternation, same finger use, row jumps etc. Try for yourself though, it might be me and perhaps I didn' t give it enough time.

There is the ADNW project (adnw.de), which was created for german but may still be interesting for you. They have a few alternative layouts (http://adnw.de/index.php?n=Main.Varianten), but most importantly two pieces of software.

- That's right, and some of you may know that I also use a home cooked AdNW version, made for my use case: typing Dutch (90%) and English (10%).  AdNW can be seen as a modern version of Dvorak. The project's website is in German, but the layout can be adapted to any language or combination of languages. There is an  article about it in the Deskthority wiki, written in English to make this layout better known to non-German speaking keyboard geeks.


Image: an AdNW layout, modified for this split, minimal, no-staggering keyboard. This is Suka's Red Tilt keyboard.

The physical hardware you have available is also a deciding factor.

(QWERTY, Dvorak, and Malt are all decidedly suboptimal in the context of a modern laptop board.)

- I suppose you mean that on a flat kayboard the bottom row is easy to reach but the top rows are harder? At least, that is what I feel on a laptop. Qwerty Y for instance, is killing on a flat keyboard, whereas Qwerty B is very easy to hit.

Certain people have different preferences as to what is comfortable, and some of it is based on their physiology, i.e. the size of their hands and the dexterity of their fingers. Rolling motions are also debated in terms of what is and what is not comfortable in terms of which particular fingers on the same hand perform them.

+1  For instance, I prefer typing with the ring finger over typing with the index finger.

[PieterGen]

Personally speaking, I started with QWERTY, tried Dvorak and then ended with Colemak (not even really trying my own proposed layout seriously).

A shame! It looked really promising, your layout 

For me Dvorak just didn't click. I find that too much alternation is a bad thing as it requires stricter organization between the left and right hand, resulting in transposition errors (two letters in a word swapped). Colemak's rolls are quite comfortable to me. And even if it turns out that Colemak is slightly less optimal in terms of what I could achieve in terms of typing speed, there's a human comfort factor which may be more or less valuable than sheer typing speed.

This is indeed very personal! To me a high alternation layout feels more even, more flowing, whereas a rolly layout such as MTGAP or Colemak after a while feels like roll  - switch  - roll - switch , that is: a bunch of letters on one hand, then switch over, then a bunch of letters on the other hand. But, feel free to have a completely different opinion 

Not mentioned: the learning curves involved. My guess is that rolls are easier to learn and therefore that a on rolly layout, such as Colemak, it takes less time to achieve a certain speed and accuracy. On the longer run however, alternating layouts may be faster/ more accurate / more pleasant. As said, this is all opinion 

There are so many aspects that play a role in layouts...... For instance: what is your "penalty function"? To make a comparison to real life: for a holiday, would you prefer a nice hotel with nice beach, all week long; or rather a superduper great hotel with a fanastic beach, BUT with 1 day of ugly weather and 1 night of noisy, hard party-ing neighbors that keep you awake all night? What is "optimal" ? Coming back to keyboards: would you prefer a layout that does everything OK, or one that is *super* but on which some words are really hard to type on?

[davkol]

It's a matter of opinion only because there isn't sufficient data.

For example, L. Malt wrote on the topic of alternation:

Experiments on transmission of information in the nervous system (Efron 1963) show a 2 - 6 m.sec longer delay to the right hemisphere stimuli. If such a delay exists it appears reasonable to conjecture that transmissions that cross the corpus callosum, which must occur with contra-lateral sequences, will each be subject to the 2 - 6 m.sec delay.

We're stuck with such extrapolations.

[LuX]

Personally speaking, I started with QWERTY, tried Dvorak and then ended with Colemak (not even really trying my own proposed layout seriously). For me Dvorak just didn't click. I find that too much alternation is a bad thing as it requires stricter organization between the left and right hand, resulting in transposition errors (two letters in a word swapped). Colemak's rolls are quite comfortable to me. And even if it turns out that Colemak is slightly less optimal in terms of what I could achieve in terms of typing speed, there's a human comfort factor which may be more or less valuable than sheer typing speed.

I can absolutely second this. I tried Dvorak for a couple of weeks but ended up not using it because of the uncomfortable hand alternation. Colemak has a lot of 'combo' words that allow you to just roll your fingers on the keyboard to type them out.
From what I hear, a big design criteria of Dvorak was hand alternation, because the heavier spring-retracting levers on the later typewriters you needed more force to press the keys, so Dvorak essentially allows you to hammer those keys more easily through hand alternation. But now in the day and age of ultralight switches, I find Colemak much better. Not the best perhaps, but better enough that I will stick to it.

As for programming, I find the most important part is to have your most used symbols at an easy to reach position. This will change from language to language. What I currently use it this:

That capslock to backspace is the best thing since sliced bread. I put capslock on to the delete key, and delete to left win. I like having delete nearby, and don't miss that winkey at all.
If you have a programmable keyboard with the extra keys for effortless layer switching, it might be a good idea to put those symbols on the home road as well. Much of it is trial and error.

[PieterGen]

Nice

[yellowfour]

Of all the well-known layouts, evidence suggests that MTGAP takes the lead in most tests for both prose and code. Not surprising, since it was created from algorithms for finding the best layout. Colemak and BuTeck-(AdNW) are slightly behind.

However, a major flaw I see in all the current crop of layouts and tests is the emphasis and overworking of the pinky, including the home pinky. All known layouts have anywhere between 5% - 13% of work done by each pinky. Together that's 10% - 26% of all key presses, which is way too high. Pinky is not built to work that hard, and nor do they type that fast. So by putting a common letter on a home pinky, you actually type slower and tire out your pinkies.

Ideally pinky usage should max out at around 5% each and 8% total, but preferably even lower. Same with columns inside the index fingers. Altogether these 4 columns should do no more than 16% of the workload. Conversely, the 3 middle columns are much stronger and faster. Even the top and bottom rows in these 3 columns are typed much faster than the home pinky--sometimes 2x to 3x faster. Thus, by prioritizing common letters in the middle, you can achieve faster speeds, less fatigue, and better rolls.

As for rolls vs. hand alternation, each must be proportionate. 2-letter rolls for speed, and hand alternations for rhythm. Long rolls slow you down and get tiresome. Too much alternation can seem too hectic and cause errors. The best balance seems to be 60% hand alternation and 30% rolls.

You can try out my new BEAKL layout that applies this new theory: https://deskthority.net/wiki/Keyboard_layouts#BEAKL_.282016.29. It was created with the aid of AdNW optimizer by heavily modifying the configuration to limit home row usage on the inside and outside columns. Yet when tested on traditional tests that heavily favor the home row, it still scores better than almost all of them. Except MTGAP, which wins in most tests, but only by a tiny amount over BEAKL. In conclusion, BEAKL still achieves top notch scores by sacrificing the home row. In return, saves the pinkies and possibly grants higher speed ceiling with less effort (since the other 3 fingers are stronger and faster).

[jacobolus]

Of all the well-known layouts, evidence suggests that MTGAP takes the lead in most tests for both prose and code. Not surprising, since it was created from algorithms for finding the best layout. Colemak and BuTeck-(AdNW) are slightly behind.

Which evidence? I’ve never seen a serious scientific study about any of these layouts.

If you mean the “lead” when scored based on arbitrary choice of a heuristic scoring function.... IMO those should be taken more as the heuristic author’s opinion or preference, rather than anything rising to the level of “evidence”. There’s no logical rigor involved; the layouts are specifically optimized to match the heuristic with little if any real-world data gathered (e.g. about comfort, error rate, typing speed, or anything else). To say that a layout “wins” by such criteria is a form of question begging, or at best represents a kind of overfitting to an incomplete model.

Ideally pinky usage should max out at around 5% each and 8% total, but preferably even lower. Same with columns inside the index fingers. Altogether these 4 columns should do no more than 16% of the workload. Conversely, the 3 middle columns are much stronger and faster. Even the top and bottom rows in these 3 columns are typed much faster than the home pinky--sometimes 2x to 3x faster.

Do you have some evidence for these claims? Or is this just your personal preference? My personal opinion is that this analysis is substantially incorrect. I would expect real-world data to easily falsify it for a typical sample population of touch typists.

(If it’s just your personal preference, there’s no problem with that... use whatever layout you prefer on your own keyboard)

* * *

Also, sorry if I sound harsh; I’m not trying to stomp on anyone. Welcome to geekhack!

[LuX]

I find my pinky surprisingly strong on home and bottom row presses. Also as long as you float your hands when typing, it's barely an issue to press keys that are off from the home row, so what's most important is that the right keys are positioned for the right finger to avoid bigrams with the same finger. I also disagree that rolls slow you down or that they are tiring.

This just goes to show there is no one layout for everyone. If it works for you, great.

[PieterGen]

Welcome yellowfour.  Interesting layout, the BEAKL-layout. Looking at the figures it seems a mix of Dvorak and Colemak. And looking at the nice pictures, I still see the Dvorak/AdNW backgrounds in it, with all the vowels on one side. Now of course all this is subjective, so this is *my* *subjective* view of the layout:

- A 2.7% same finger use seems rather high !
- 64% of alternation is OK, though my personal preference is a bit higher.
- 14% adjacent keys is (to me) *way too high*. But, other people *love* rolls on adjacent keys.
- I don't see the figure on 3-letter combos on one hand (the AdNW optimizer shows this, when in trigram-mode. Often, a layout that has many "nice" rolls also has  many not-so-nice 3 letter combos on one hand. Or even more on one hand, such as (querty) secrets

I agree with jacobolus that there is no 'objectively' optimal % that the pinkies (or any finger) should do.

I agree with yellowfour that a "2 keys left,  2 keys right" rhythm may be ideal - although I do not have data to back this claim up (the mtgapcom site has some anecdotal data though, that seem to support this).

I agree with Lux that there is no one layout for everyone.  :-)

[yellowfour]

I find my pinky surprisingly strong on home and bottom row presses. Also as long as you float your hands when typing, it's barely an issue to press keys that are off from the home row, so what's most important is that the right keys are positioned for the right finger to avoid bigrams with the same finger. I also disagree that rolls slow you down or that they are tiring.

The difference is in milliseconds, so your body can't detect the difference. You need special hardware or software to time how fast each finger types (like http://patorjk.com/typing-speed-test/). The difference adds up quickly in a matter of seconds and minutes.

From my own tests, the ring finger is consistently the fastest and the pinky is consistently slower. Surprisingly the fastest key (for me) is the bottom ring finger.

Also, the right hand is stronger and faster. From this I conclude that vowels should be put on the right side, due to more vowels have higher frequency--4 of 5 most frequent letters in English are vowels.

There are as many bad rolls as there are good rolls for the pinky, so that "advantage" is a wash. You don't like bigrams with the same finger, neither do I. So having the weakest finger (pinky) hit double letters should be avoided. Like Dvorak 'S' & 'L' and Colemak 'O' which are quite common double letters. Hence, you're agreeing with me that common letters, especially letters that are commoly doubled in English, should not be put on the pinky.

Welcome yellowfour.  Interesting layout, the BEAKL-layout. Looking at the figures it seems a mix of Dvorak and Colemak. And looking at the nice pictures, I still see the Dvorak/AdNW backgrounds in it, with all the vowels on one side.

The Dvorak/AdNW influence is coincidental since the final layout is created by the optimizer. Probably because the default AdNW itself has heavy Dvorak influence and my modifications didn't change those qualities much, if at all.

However, the previous BEAKL prototype layouts that I created manually did resemble Dvorak. Since it seemed easier to divide the keyboard by putting all vowels on one side, and then work on each half to find the right balance.

But yes it is somewhat surprising for BEAKL optimized layout to resemble Dvorak that much. If you look closely at other layouts not created from AdNW nor trying to improve Dvorak, some of them may not be that much different than Dvorak. MTGAP has 6 vowels (AEIOU+Y) on left half. Colemak has 5 on right half (only A is on left side.)

As for the stats:

- A 2.7% same finger use seems rather high !
- 64% of alternation is OK, though my personal preference is a bit higher.
- 14% adjacent keys is (to me) *way too high*. But, other people *love* rolls on adjacent keys.
- I don't see the figure on 3-letter combos on one hand (the AdNW optimizer shows this, when in trigram-mode. Often, a layout that has many "nice" rolls also has  many not-so-nice 3 letter combos on one hand. Or even more on one hand, such as (querty) secrets

- 2.7% same finger is actually low relative to other layouts. Other layouts: Colemak 3.0%, MTGAP 2.5%, Dvorak 3.5%, Capewell 3.2%, Qwerty 6.1%
- 64% alternation is already higher than the other layouts, which have between 50% to 62% hand alternation.
- 14% adjacent is too high? other layouts; Colemak 15.6%, MTGAP 16.9%, Dvorak 12.9%, Capewell 18.8%, Qwerty 18.0%
- the trigram stats:
10.866 no hand altern.
42.637 two hand altern.
5.648 seesaw
5.328 indir same finger

These seem a lot better than MTGAP and Colemak.

MTGAP:
14.489 no hand altern.
33.870 two hand altern.
8.414 seesaw
4.187 indir same finger

Colemak:
19.892 no hand altern.
34.560 two hand altern.
10.736 seesaw
4.487 indir same finger

I agree with yellowfour that a "2 keys left,  2 keys right" rhythm may be ideal - although I do not have data to back this claim up (the mtgapcom site has some anecdotal data though, that seem to support this).

It is MTGAP site. https://mathematicalmulticore.wordpress.com/2011/11/27/typing-data-preliminary-analysis/ He timed how fast trigrams are typed. The fastest trigrams in general are typed with 2-1 or 1-2 meaning 2 letters on one hand and the 3rd letter on the other hand. Of all the slowest trigrams, the bulk of them are typed all on the same hand.

[jacobolus]

The difference is in milliseconds, so your body can't detect the difference. You need special hardware or software to time how fast each finger types (like http://patorjk.com/typing-speed-test/). The difference adds up quickly in a matter of seconds and minutes.

How long a test did you do? You should probably do a test for a half hour or something, on whatever type of prose you commonly type (rather than some chunk of a classic novel) if you want a fair sample.

Then beyond that, you need to adjust the analysis to account for the way that brains chunk up words or other kinesthetic actions. A keystroke is going to be faster if it comes in the middle of a common "chunk" typed as a single choreographed motion.

So you need to adjust the analysis of timing based on (a) letter frequency, (b) commonness of the word or phrase being typed, (c) whether or not the letter is part of a transition between separate chunks (which depends on your own personal typing style and level of practice), (d) physical keyboard layout, which makes some keys easier to reach than others, apart from which finger they use.

Overall, I think this is an almost impossible type of test to get meaningful data from without putting in significant effort to account for the above. The simple numbers from your link are going to be severely misleading.

There are as many bad rolls as there are good rolls for the pinky, so that "advantage" is a wash. You don't like bigrams with the same finger, neither do I. So having the weakest finger (pinky) hit double letters should be avoided.

You should try to avoid one-finger, two-key bigrams for any finger. In my opinion, pressing the pinky twice is only marginally different than pressing the middle or ring finger twice. Pressing the index finger or thumb twice in a row should be somewhat easier though. It depends somewhat on how stiff your keyswitches are and how the keys are physically arranged, though. One of the problems with the pinky is that it’s physically shorter, which means that on many keyboards it is slower to e.g. reposition the pinky from the top row to the bottom than for other fingers.

It is MTGAP site. He timed how fast trigrams are typed. The fastest trigrams in general are typed with 2-1 or 1-2 meaning 2 letters on one hand and the 3rd letter on the other hand. Of all the slowest trigrams, the bulk of them are typed all on the same hand.

By far the fastest 8-gram is going to be something like the QWERTY ;lkjasdf.

Where letter sequences get slow is when you need to either reposition the hand in the middle of a sequence – e.g. typing QWERTY ,Y – or when you need to alternate back and forth between pinky, ring finger, and middle finger, which overload the same muscles for flexion/extension. The reason that “rolls” between these fingers work okay is that you can do them as a single motion. But if you need to type QWERTY sad or sas or something like weasel, it gets pretty slow. In the middle of a chunk, being on the top row or the bottom row isn’t really too big a deal (despite the weight given in most of these heuristics). It’s extremely fast to type a word like ointment in QWERTY, even though none of the letters are on the home row.

[PieterGen]

But if you need to type QWERTY sad or sas or something like weasel, it gets pretty slow.

This is what the adnw optimizer calls seesaw. It can happen on 3 (or more) letter combos on one hand. I found than seesaws really have a bad influence on (my) typing fun. Seesaw can also be called " bad rolls".  One other key parameter is "no hand alternation", which indicates the total percentage of 3-letter combos that are typed on one hand. 'Seesaw' (the bad rolls) is a subset these; the rest are so to speak "good rolls". 

BEAKL scores good in terms of 3-letter combos on one hand (only 10.9% of all 3-letter combos are on one hand) and
5.6% of all 3-letter combos are seesaw. Good.

For comparison, here are the stats of my personal layout (optimized for Dutch, but also OK for English):
Here's the layout:

buy,* qkvlfj
saeio gdtnrw
z**.ij pcmhx


Some stats - language Dutch: 7.1% of all 3-letter combos are on 1 hand; 3.6% are seesaw.  Same finger is 1.0%
for English language the stats are: 7.1% of all 3-letter c8ombos are on 1 hand; 3.8% are seesaw.  Same finger is 2.1%

These are only numbers though. I have some layouts that score better (numerically) but that I didn't like as much in real life. A sign that the model is not quite right. My layout has a high alternation (70% in Dutch, 68% in English), which is something I like, but which is very personal, other people prefer less alternation, that is OK  In the end it is about what someone likes in real life. 

[LuX]

Interesting points. I find that a lot of n-grams on same hand on Colemak tend to feel quite nice. Take for example the word 'mountain' which I find a joy to type, despite having 4 letters on the same hand like a see-saw.
Another point is that physical letter placement is a big factor as well. 'Sad' (ring, pinky, middle) on QWERTY feels awkward, but 'you' (ring, pinky, middle) on Colemak, despite requiring the same finger order (except other had) feels much better. I find this is because the fingers on Y,O,U on Colemak are at a relatively rested position, while S,A,D on QWERTY requires to keep my fingers at clawed position. Also maybe because my right hand is a little better at fine motions.

And yes, all these numbers from different analyzers don't really say anything, because they fail to take the most important part into consideration. Which is...?

[jacobolus]

This is what the adnw optimizer calls seesaw.

Does it separate out combinations involving the index finger? Because the index finger is much more independent than the other three fingers, alternating back and forth between e.g. index finger and pinky is nowhere near as bad as alternating back and forth between middle finger and ring finger.

[Pro XKB]

This is what the adnw optimizer calls seesaw.

Does it separate out combinations involving the index finger?

It does not separate them out.  However, it is possible to define custom bigram and trigram categories, for gathering statistics, and to introduce custom optimization criteria.  The optimizer comes with a little AWK script that demonstrates this for “rolls”.  For seesaw without index finger, it can be modified to have a main body like so:

END {
    for(i = 0; i < n; ++i){
   # Skip thumbs and index fingers
   if(abs(finger) <= 2) continue;
        for(j = 0; j < n; ++j){
       # ditto, and skip hand alternation as well
       if(abs(finger[j]) <= 2 || finger*finger[j] <= 0) continue;
            for(k = 0; k < n; ++k){
      # ditto.
      if(abs(finger[k]) <= 2 || finger*finger[k] <= 0) continue;
      # Select combinations where the finger for the first and last
      # key are on the same side with respect to the finger operating
      # the second key.
      if((finger-finger[j])*(finger[k]-finger[j]) > 0){
          print "Trigramm", name, name[j], name[k], extra, "'seesaw-noindex'";
      }
       }
        }
    }
}

[PieterGen]

Thanks Pro XKB   

[bocahgundul]

soooo

what is the most optimized layout recently?

mod-dh?

[PieterGen]

soooo

what is the most optimized layout recently?

mod-dh?

LOL. No single answer because the jury is still out on what is "the best" and even on " how to judge that".  So, pick whatever you like. For most languages, both Colemak and Dvorak are fine choices. :-) 

[Phenix]

whats the best for German?

btw:
Is it possible to just make a vote? Would be interesting to see how many use which layout (using QWERTZ all day - this would be at least an indicator which I shall try out first)

[bocahgundul]

What's the best for English?

[PieterGen]

@Phenix and @bocahgundul - depending on your analysis, preference, hands, how deep in the rabbithole you want to go, etc.  These are some indications:

Good for English:
(1) Colemak.
(2) Other choice, more or less the same "rolly" feel, more different from Qwerty, but scoring a little better than Colemak: MTGAP.
ypouj kdlcw
inea, mhtsr
qz/.:  bfgvx

If you prefer more alternation,
(1) stock Dvorak is for English a fine choice.
(2) better is AdNW (beating Dvorak in both English & German)
(3) For fun I once calculated a pure-English version of AdNW, that makes nop compromises for German and therefore scores even better for typing English. I nicknamed this layour DvorMax.
kyu.* zlmdpv
rieao hnstcw
x?:,/ jqfgb

Good for German:
(1) Dvorak
(2) Even better: AdNW
Colemak is not so nice for German (too high same finger use). Although some people use it and like it, so as always this is personal.

Capewell-Dvorak is yet an other layout that I found comfortable. Both for English and for Dutch, and I suppose it must be nice for German as well.
',.py qfgrk
oaeiu dhtns
zxcvj lmwb;

As the Deskthority wiki writes: a modified version of Dvorak designed by Michael Capewell. Attempts to solve many perceived faults with the Dvorak layout, such as:
- Placement of the L letter
- Swaps I/U because I is much more common than U and should be nearer
- ZXCV return to their original QWERTY positions

tl;dr
- Colemak is a good choice for English. Dvorak is also good. If you are willing to go a bit deeper in the rabbit hole, MTGAP, AdNW and Capewell-Dvorak are nice.
- Dvorak is good for German. AdNW and Capewell-Dvorak are also fine

As always: your milage may vary !!

[Phenix]

So what would you recommend me first (or your personal experience)?
AdNW and Capewell-Dvorak

Normal typing around 70% english, 30% German

If I need to do something for work/school its only German
----
doing something for work/school is not really often, maybe 3 times a month )

---
will weather be used with (depends on what arrives earlier)
1) ergodox
2) gh60
3) 8x16 Matrix keyboard (yet not ordered)

4) Model M (daily driver)

Should I get suited keycaps with matching legends?

[Pro XKB]

AdNW and Capewell-Dvorak

Normal typing around 70% english, 30% German

With that weighting, with the corpus that is included in the optimizer package, trigrams, and default configuration:

Aus der Neo-Welt 299.860 total effort   185.946 positional effort    left right
                   1.037 same finger rp   9.628 shift same finger top  5.2 11.9
  kuü.ä vgcljf    71.342 hand alternat.  24.577 shift hand alter. mid 37.2 31.9
  hieao dtrnsß     1.898 inward/outward  25.122 inward or outward bot  5.1  8.8
  xyö,q bpwmz      8.823 adjacent        20.578 shift adjacent    sum 47.5 52.5
                   4.587 no hand altern. 44.612 two hand altern.
                   3.395 seesaw           6.664 indir same finger
                  8.1 11.2 13.0 15.2 --.- --.- 18.0 10.8 13.8  9.9 Sh  2.5  1.1


Capewell-Dvorak  354.518 total effort   193.950 positional effort    left right
                   3.414 same finger rp   9.864 shift same finger top  5.3 11.3
  ü,.py qfgrkö    67.405 hand alternat.  33.439 shift hand alter. mid 35.5 30.3
  oaeiu dhtnsä     1.250 inward/outward  26.683 inward or outward bot  6.8 10.8
  zxcvj lmwbß     15.023 adjacent        18.946 shift adjacent    sum 47.5 52.5
                   7.746 no hand altern. 40.167 two hand altern.
                   4.124 seesaw           6.076 indir same finger
                  8.4  8.3 16.8 13.9 --.- --.- 16.8 11.5 15.4  8.8 Sh  2.2  1.3

Same finger repetitions are quite frequent with Capewell-Dvorak, so even if you disregard the disputable “total effort”, AdNW is probably preferable.

[Phenix]

ok, than I go with AdNW.
Can you provide some images/layout? (I see the letter cluster, but alligned onto a ergodox or an normal board would be better)

Maybe an layout with http://www.keyboard-layout-editor.com/#/

[PieterGen]

Read all about it on the Deskthority wiki and of course on www.adnw.de   Drivers & scripts are on the adnw website, for all major operating systems (windows, osx, linux, *bsd).

Pictures, among others of AdNW on an ergodox (patorjk.com keyboard analyzer picture): http://imgur.com/a/JIx9A

Example of minimalist keyboard (by GH/ DT user suka):

Here's a different minimalist version by GH user Oobly.

[bocahgundul]

I type 100% english and I use normal ANSI layout so AdNW is better or are there other layout that is better?

[PieterGen]

I type 100% english and I use normal ANSI layout so AdNW is better or are there other layout that is better?

Hi Bocahgundul, it depends on your preferences, so try them out (at least a few days / evenings). All are much better than Qwerty.

- Colemak is a fine choice. Pro: closer to Qwerty. Neutral: more rolls, less alternation then Dvorak. Depends on how this feels to you.
- If you want a better Colemak, MTGAP is very nice. The cons are that you have to make your own scripts, so that may be hard for some.

- Dvorak is another fine choice. Pro: is standard on all operating systems, even Windows. Meaning you can use Dvorak withour any USB sticks etc. on virtually every computer, it's "just a setting away". Neutral:  more alternation, less rolls.
- If you want "a better Dvorak" there are 2 possibilities for English:
  - AdNW (it beats Dvorak when typing English). AdNW is good for German AND good for English. If you only type English, AdNW is still better than Dvorak.
- Capewell-Dvorak is another "better Dvorak". It keeps some qwerty shortcuts, but all in all it's not as good as AdNW. So in the Dvorak family, AdNW is the best guess for typing English.

All in all this leads to three layouts that are all good for English and also well supported (meaning either built in in Operating Systems or available as software for easy install or use off USB-stick) :
- Colemak (and improved qwerty)
- Dvorak
- AdNW (an improved Dvorak).

My advice would be to try both AdNW and Colemak and see which one you prefer. Give both some time. I myself prefer AdNW by far for English, but again this is very personal, lots of people prefer Colemak. Diversity is good   

[tp4tissue]

There is no MOST anything layout..

Avoid dvorak.. That guy doesn't know anything about typing..

The HOME ROW is not straight..


People think a standard home row is asdfjkl'   that's wrong.. it's actually   awefjio'

No long time typer who's made it to 120wpm ever moves their hands back to asdfjkl'

[davkol]

[tp4tissue]

They're trying to make the point that speed is limited by the layout..

This is simply not true.

Speed is limited by parsing technique and muscle-memorization of longer word chains


The layout has no conclusive impact on ultimate speed..  because alternating hand use,  keystroke balance are not the final barriers.

[davkol]

EDUCATIONAL TEST BUREAU. 1950. International Commercial Schools Contest Awards Won by Simplified Keyboard Typists in Competition with Stan-
dard Keyboard Typists. International Typewriting Tests.

[bocahgundul]

I type 100% english and I use normal ANSI layout so AdNW is better or are there other layout that is better?

Hi Bocahgundul, it depends on your preferences, so try them out (at least a few days / evenings). All are much better than Qwerty.

- Colemak is a fine choice. Pro: closer to Qwerty. Neutral: more rolls, less alternation then Dvorak. Depends on how this feels to you.
- If you want a better Colemak, MTGAP is very nice. The cons are that you have to make your own scripts, so that may be hard for some.

- Dvorak is another fine choice. Pro: is standard on all operating systems, even Windows. Meaning you can use Dvorak withour any USB sticks etc. on virtually every computer, it's "just a setting away". Neutral:  more alternation, less rolls.
- If you want "a better Dvorak" there are 2 possibilities for English:
  - AdNW (it beats Dvorak when typing English). AdNW is good for German AND good for English. If you only type English, AdNW is still better than Dvorak.
- Capewell-Dvorak is another "better Dvorak". It keeps some qwerty shortcuts, but all in all it's not as good as AdNW. So in the Dvorak family, AdNW is the best guess for typing English.

All in all this leads to three layouts that are all good for English and also well supported (meaning either built in in Operating Systems or available as software for easy install or use off USB-stick) :
- Colemak (and improved qwerty)
- Dvorak
- AdNW (an improved Dvorak).

My advice would be to try both AdNW and Colemak and see which one you prefer. Give both some time. I myself prefer AdNW by far for English, but again this is very personal, lots of people prefer Colemak. Diversity is good   

I've tried dvorak but the layout is not for me cause its been hurting my pinky finger because of the heavy use of right pinky finger and I'm searching for another layout


I can't seem to find a simple straightforward information of the MTGAP layout and do you have any comment on the Colemak-DH mod?

[iLLucionist]

Is there good software that helps you switch from layout, to train yourself for example for colemak. Like goodtyping.com with pictures and all?

[algernon]

Is there good software that helps you switch from layout, to train yourself for example for colemak. Like goodtyping.com with pictures and all?

I've been using http://thetypingcat.com/ with great satisfaction. Has Colemak tutorials too.

[iLLucionist]

Is there good software that helps you switch from layout, to train yourself for example for colemak. Like goodtyping.com with pictures and all?

I've been using http://thetypingcat.com/ with great satisfaction. Has Colemak tutorials too.

Thanks! The sounds are annoying though, but that can be turned off easily.

[PieterGen]

I use amphetype. You have to build it yourself, but it's nice and simple and doesn't require an internet connection.

For lessons, I analized a bunch of text material I wrote myself (scraped off my own disk drives  ). I used a simple python script to find the most used
- two letter combinations (also known as digrams or digraphs
- three letter combinations (yep, trigrams or trigraphs)
- 4 and 5 letter words - my greek is rusty, so I just call them 4-grams and 5-grams 

I type mainly in Dutch and English. Therefore, I complemented these x-grams with the 1,000 most common Dutch words and the 1,000 most common English words. As found on the internet.

I think (correct me if I'm wrong) that typing lessons philosophy is/was "keyboard centric" in the sense that exercises are based on the keyboard: first you type words on the homerow, then words on the top row, then words on home+top row, and so on. I thought that "word centric" would be better. That is: first train the most common combinations and words. Then less common combinations.

Anecdotal evidence from the adnw-mailing list is by the way that:
- going cold turkey is as fast as a step by step, deliberate training. This resonates with me when I think of learning languages. No better way to learn spanish than surrounded by people who *only* speak spanish.
- accuracy > speed. So first focus on not making mistakes, then on speed. The reason is that you need to form muscle memory. The muscles are dumb, they only repeat patterns. If you constantly type wrong patters, you learn the wrong patters. BTW, accuracy > speed is also who musicians learn. Look at the super fast metal guitar players, or classical piano players - they both train on accuracy, not on speed. Speed is a byproduct of playing accurate.
- training (on amphetype or other programs, or online) is usefull to identity your weakest words / combos, and then train specifically on those words.

Now excuse me, I must start my Amphetype and do some excercises    I'm a mediocre typist myself 

[iLLucionist]

I use amphetype. You have to build it yourself, but it's nice and simple and doesn't require an internet connection.

For lessons, I analized a bunch of text material I wrote myself (scraped off my own disk drives  ). I used a simple python script to find the most used
- two letter combinations (also known as digrams or digraphs
- three letter combinations (yep, trigrams or trigraphs)
- 4 and 5 letter words - my greek is rusty, so I just call them 4-grams and 5-grams 

I type mainly in Dutch and English. Therefore, I complemented these x-grams with the 1,000 most common Dutch words and the 1,000 most common English words. As found on the internet.

I think (correct me if I'm wrong) that typing lessons philosophy is/was "keyboard centric" in the sense that exercises are based on the keyboard: first you type words on the homerow, then words on the top row, then words on home+top row, and so on. I thought that "word centric" would be better. That is: first train the most common combinations and words. Then less common combinations.

Anecdotal evidence from the adnw-mailing list is by the way that:
- going cold turkey is as fast as a step by step, deliberate training. This resonates with me when I think of learning languages. No better way to learn spanish than surrounded by people who *only* speak spanish.
- accuracy > speed. So first focus on not making mistakes, then on speed. The reason is that you need to form muscle memory. The muscles are dumb, they only repeat patterns. If you constantly type wrong patters, you learn the wrong patters. BTW, accuracy > speed is also who musicians learn. Look at the super fast metal guitar players, or classical piano players - they both train on accuracy, not on speed. Speed is a byproduct of playing accurate.
- training (on amphetype or other programs, or online) is usefull to identity your weakest words / combos, and then train specifically on those words.

Now excuse me, I must start my Amphetype and do some excercises    I'm a mediocre typist myself 

Thanks! That looks very promising.

[Kinami]

I made the switch to the Workman layout a year ago, and since then have been using only that. I chose it based on feel, after typing a few sentences each in Workman, Dvorak, and Colemak.

The good points are that I can really feel the decreased finger travel distance as I type--much less hand movement is needed. The balance between both hands is also nice if you spend all day typing. And I agree with the comment above that you need to decide whether you prefer alternation or same-hand rolling. I think alternation is better in my case--that way your second hand can already be getting into position as you're typing the present letter.

As far as guiding philosophies go, Colemak always struck me as a half-baked attempt that was more interested in not scaring Qwerty users than in true optimization.

The main thing I've learned from this switch is that there are costs to picking something too original. What will you do if you get a job where you need to work all day on a computer you aren't allowed to customize heavily? Workman is the most exotic you can get and still be compatible with normal machines, since it's among the default options for most Linux systems, can be added to Windows with a portable application, and has 1 iPhone/iPad app that supports it. Anything more rare, and you're tied to your workstation. At this rate I'll probably switch to Dvorak sometime in the future, simply for the sake of being able to transfer more easily between devices.

[iLLucionist]

The main thing I've learned from this switch is that there are costs to picking something too original. What will you do if you get a job where you need to work all day on a computer you aren't allowed to customize heavily? Workman is the most exotic you can get and still be compatible with normal machines, since it's among the default options for most Linux systems, can be added to Windows with a portable application, and has 1 iPhone/iPad app that supports it. Anything more rare, and you're tied to your workstation. At this rate I'll probably switch to Dvorak sometime in the future, simply for the sake of being able to transfer more easily between devices.

1. If you are not allowed to change your PC or get your own hardware, your boss is a m*therf**ker and you should seek employment elsewhere. I had to work with eclipse on a i3 with 512MB memory in 2012 with a 17" screen OVER NETWORK which was 100mbps. I quit.

2. If seeking a proper boss is not an option, bring your own hardware. That's what I do now: I take my mbp everywhere I go. Have my own screen and mech board at the office.

[algernon]

1. If you are not allowed to change your PC or get your own hardware, your boss is a m*therf**ker and you should seek employment elsewhere. I had to work with eclipse on a i3 with 512MB memory in 2012 with a 17" screen OVER NETWORK which was 100mbps. I quit.

2. If seeking a proper boss is not an option, bring your own hardware. That's what I do now: I take my mbp everywhere I go. Have my own screen and mech board at the office.

Good luck getting custom HW into a nuclear plant =)

Also, when you have to switch between computers all day, and use stuff you do not own or control (think troubleshooting, for example), you are stuck with what is available. Carrying a keyboard everywhere, all the time, is not always an option.

[iLLucionist]

1. If you are not allowed to change your PC or get your own hardware, your boss is a m*therf**ker and you should seek employment elsewhere. I had to work with eclipse on a i3 with 512MB memory in 2012 with a 17" screen OVER NETWORK which was 100mbps. I quit.

2. If seeking a proper boss is not an option, bring your own hardware. That's what I do now: I take my mbp everywhere I go. Have my own screen and mech board at the office.

Good luck getting custom HW into a nuclear plant =)

Also, when you have to switch between computers all day, and use stuff you do not own or control (think troubleshooting, for example), you are stuck with what is available. Carrying a keyboard everywhere, all the time, is not always an option.

Ahh yess... that would be difficult indeed. I would be scared to work in a nuclear plant anyway.