• You are not logged in.

Making an Improved Colemak

  • Started by stevep99
  • 26 Replies:
  • Reputation: 117
  • From: UK
  • Registered: 14-Apr-2014
  • Posts: 975

As we know, Colemak is an excellent layout. One of the first optimized layouts, it has stood the test of time. Although some have proposed alternative layouts to address perceived flaws in Colemak, these tend to solve some problems at a cost of introducing new weaknesses.

In my personal view, Workman is a good example. It does a nice job of making better use of easier to reach keys, but at a cost of more same-finger bigrams and some awkward-looking placements, such as its D.

But instead of throwing the baby out with the bathwater, what if we asked the question what minor mods could be made to Colemak which address possible criticisms, but without adversely impacting on Colemak's strenghts? So the purpose of posting this topic is to look at some possible options.

Basically, following on from discussing the D>P>G mod, the question I am asking is: What changes can be made subject to this strict rule: Starting from base Colemak, no key can change fingers. Consequently, any such change would (i) leave same-finger bigrams unaffected and (ii) be much easier to learn than a change that required finger changes.

The main target here are the index fingers, given they cover 6 keys, and that's where potential improvements can be gained.

I have seen some keyboard diagrams where people rate each key position by effort to measure ease of typing. These are no doubt somewhat subjective - but perhaps it's possible to get some consensus at least on the preferred order of keys on a given finger. Starting from the index finger "home" position, I propose this order would be: (1)bottom-row, (2)top-row, (3)middle-row-centre-column, (4)bottom-row-centre-column, (5)top-row-centre-column, i.e.:

2 5   5 2
T 3   3 N
1 4   4 1

So now let's look at the actual frequency of letters.

LEFT HAND

t     9.056%
d     4.253%
g     2.015%
p     1.929%
b     1.492%
v     0.978%

As you can see, the correlation between key effort and frequency is far from optimal. This in my view accounts for the most common criticisms of Colemak.

Let's look at some options to improve things:

1. The D > P > G switch

D P
T G
V B

This is the mod that started me thinking about this topic. A clockwise rotation of those three keys, to leave D and G in improved positions. P is worse off but is least frequent of the 3.
+ more comforable D and G
+ G returns to Qwerty positon (one less changed key for new users)
+ Left centre-column usage reduced from 7.76% to 5.44%
- 3 keys change from standard Colemak

2. The D > V > B > G switch

P B
T G
D V

For this change, I think the Angle mod is a pre-requisite. The common D moves to the easiest available spot (C key on a standard board).
+ optimal D and easier G
+ G returns to Qwerty positon
+ Left centre-column usage reduced from 7.76% to 4.49%
- 4 keys change from standard Colemak
- Breaks the standard "Z X C V B" pattern

3. Move all 6 keys strictly according to effort/frequency stats

G V
T P
D B
+ optimal arrangement according to stats
+ Left centre-column usage reduced from 7.76% to 4.40%
- Breaks the standard "Z X C V B" pattern - position of V bad for paste shortcut
- only B retains Qwerty position

Personally I would rule this one out.

RIGHT HAND

n     6.749%
h     6.094%
l     4.025%
m     2.406%
k     0.772%
j     0.153%

The right hand index finger keys are already closer to what the effort-frequecy stats suggest as optimal. The only conceivable change to produce a benefit is:

The H-M switch

J L
M N
K H

A simple switch to move H to the easily reached M position.
+ Easier to type H
+ Elimate the much compained-of HE bigram. I make that HE/EH accounts for 3.49% of bigrams, ME/EM accounts for 1.15%
- Moves two keys which were unchanged from Qwerty.

SUMMARY

Of course, making any change at all technically makes it no longer Colemak, but by limiting ourselves to modest changes, all same-finger, I think the result could still be considered as in the Colemak "family".

For the left hand, the D>P>G switch is a quick win with very little disadvantage, especially for new Colemak users who would have to relearn one fewer change (G).

Q W F D P J L U Y
A R S T G H N E I O
Z X C V B K M , .

However, if we are talking about the best available Colemak-like layout, then I think the D>V>B>G switch used with the angle mod wins. The angle mod would be compulsory with this arrangement. By also combining with the H-M switch, it also gives a certain pleasing symmetry regarding the former centre-column H and D:

Q W F P B J L U Y
A R S T G M N E I O
Z X C D V K H , .

It does require 6 changed keys from base Colemak, which is more than ideal, but at least all are on the same finger. With its easier to type keys, but retaining Colemak's home keys and excellent same-finger bigram stats, perhaps this could be a very pleasant evolved Colemak? Anyone think it's worth a try?

Update: The results of these of these experiments resulted in a new Colemak mod called Mod-DH which is further discussed in this thread...

Last edited by stevep99 (21-Jan-2015 18:23:25)

Using Colemak-DH with Seniply.

Offline
  • 1
  • Reputation: 0
  • Registered: 28-Sep-2014
  • Posts: 7

There are a couple of issues I see.

1.  In order to successfully establish colemak as a viable alternative it requires stability.
2.  The differences between optimized layouts are very small percentage wise.
3.  Individual variations in taste will result in differing optimizations.

... That said
1.  Colemak will never be mainstream
2.  Learning one optimized layout is as good as another
3.  It will not improve typing speed over qwerty
4.  RSI from typing is resolved by improving wrist position.

  So have fun.

Offline
  • 0
  • Reputation: 0
  • Registered: 28-Sep-2014
  • Posts: 7
Offline
  • 0
  • Reputation: 117
  • From: UK
  • Registered: 14-Apr-2014
  • Posts: 975
tiramisuu said:

1.  In order to successfully establish colemak as a viable alternative it requires stability.
2.  The differences between optimized layouts are very small percentage wise.
3.  Individual variations in taste will result in differing optimizations.

1. Obviously I'm not suggesting official Colemak is changed. Just some possible mods which are hopefully improvements and address commonly cited flaws.
2. Of course, and it depends how you measure. Although I also suspect there are issues that arise from overuse of awkward keys / combinations that a simple metrics algorithms might not detect. This is shown by the that many alternatives have continued to be created in the last several years. It's partly a personal experience of how a layout feels overall.
3. That's why I posted my assumptions of prefered keys - it would be interesting to know if others don't agree with them.

Yeah I've read that page before. The trouble is those "improvements" end up looking nothing like Colemak, with many keys moved all over the place! Not the same approach I'm suggesting at all.

Last edited by stevep99 (25-Oct-2014 18:25:54)

Using Colemak-DH with Seniply.

Offline
  • 0
  • Reputation: 210
  • From: Viken, Norway
  • Registered: 13-Dec-2006
  • Posts: 5,343

(CarpalX is raving – that's not "improving Colemak" in my opinion, but making a new layout using the CarpalX model based on some of the ideas behind Colemak. To my eyes, that's a "broken improvement".)

I shudder and cringe at the thought of abandoning the Colemak main layer stability, especially considering all my technical solutions for xkb, PKL and graphics! Oh LORd no. The arguments would have to be weighty and so far they aren't. The thoughts are still interesting of course.

The only one of these mods I find really intrigueing is the D>P>G loop:
+ It moves one less key from its QWERTY position, leading to more layout efficiency per moved key (somewhat important to me but not to all).
(+ Furthermore, on my laptop there's a stick in the middle which makes it impossible to move the GHB keys for what that's worth)
+ The corresponding Tarmak progression would be more even: Not counting J there would be exactly 3 keys moved in every step!
(- But then, I'd have to make and support another set of Tarmak files which I really don't feel like doing at the moment – Tarmak just went through a revision too!)
+ The Colemak P position does feel more comfortable than the D one! The difference is minor though.
- A schisma in the "cult" of Colemak is something we need about as much as the schismas in Islam and Christianity...
- Mayhem. Pandemonium. Sleeplessness. Falling stars. Hell bursts open and the unholy walk the Earth.

For fun, I ran a quick-and-dirty test on the Patorjk Keyboard Layout Analyzer. It gives the same result for both variants, ±0.03 (<0.05% aka nothing at all) for the standard corpuses. But that analyzer has its own agendas as they all do. Might be interesting to see what the Workman analysis makes out of this mod (possibly including the H>M mod as well for good measure)! Obviously, the only real change modelwise will be the weightings of the three keys so the actual analysis run might not be all that interesting after all as it'll all be about your weighting choices anyway.

I'm having enough trouble trying to provide xkb and PKL layout variants in different flavors to US ANSI users and others who need their hand held. Making further options would be too much really, but for my own purposes I might enter some extra lines in the code here and there as an "easter egg" to enable this mod for the adventurous. That much is fairly simple to do as long as it doesn't have to be visible on the surface! On the surface, I really don't want to change the Colemak layout now for many reasons. But it might be nice to provide these little tools for the ones complaining about the middle column for instance.

A name for this beastie. I thought of 'ColemaD' since a) it moves the D and b) it's utter madness. Given that its focus is to clear the middle columns, it might be called ColeMose? :D

Last edited by DreymaR (26-Oct-2014 19:37:21)

*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

Offline
  • 0
  • Reputation: 117
  • From: UK
  • Registered: 14-Apr-2014
  • Posts: 975

Thanks for the comments. I undersand the desire to not create confusion with multiple potential versions of Colemak out there. I should stress that I don't see this proposal as being in any way anti-Colemak, quite the opposite in fact: There are obviously differences of opinion about the centre column issue, and for those who like it the way it is there is no need to change anything. I see this mod as being aimed at those who try Colemak but experience problems with the lateral movement, and out of frustration look to other, sometimes inferior, layouts, or abandon the idea entirely. I think this mod effectively solves the issue for those people while still retaining Colemak's other benefits. Having a broad ecosystem of minor mods and customizations might actually be a strength for Colemak rather than a weakness.

In my case, when I was learning Colemak, I found reaching for the D H (and G) keys awkward for a while. Even though I got to the point of accepting them, a niggling doubt remained, but I preferred to avoid changing anything and stick with the official version. But I've also started to think, what with wide mods, angle mods, custom layers on Caps and AltGr, my keyboard has effectively become very personalized already. I think the trouble is, the kind of people who seek out alternative keyboard layouts are by their nature not the types to settle if they see a potential improvement. This may be both a blessing and a curse for Colemak!

So, I figured having suggested it, I should at least try it out. I have now partially implemented it (the left hand part, DVBG). This means I have also implemented the angle mod at last, as it the new D position only makes sense with it. My previous wavering over the angle mod was partly about the loss of the easy-to-reach C position, to be replaced by V. This mod solves this problem for me nicely by promoting D to that prime real estate. I'll let you know how it goes!

Oh yes, I suppose it needs a name. ColemaD is kind of funny - I almost quite like it!  While it's technically a new layout, I think of it more as a Colemak mod. A more prosaic name could be just the "D4-mod" for the left-hand part, based on the main key changed and the number of keys. The full mod would when be the "DH6-mod". (Could be D4H2, but that sounds too much like a kind of bird flu!). Probably a better name could be thought of.

Last edited by stevep99 (26-Oct-2014 15:32:35)

Using Colemak-DH with Seniply.

Offline
  • 0
  • Reputation: 210
  • From: Viken, Norway
  • Registered: 13-Dec-2006
  • Posts: 5,343

I agree that a handful of mods can be useful additions to the layout! As long as they're modular, I'm a big fan of adding a heap of mods for the adventurous. As you say, this makes it possible to keep tweaking and improving your layout while staying with the Colemak concept. I was just worrying about my own capacity to sort and cater for the ever growing arsenal of good ideas... ;)

The problem may be that while learning you'll draw a few hasty conclusions. Most people do. Unless you have a little patience you may end up overdoing it with so much to choose from. And then you're stuck with something that isn't as practical as it could be – but that may not be a big problem after all.

I think your numbering scheme is unintuitive and a poor basis for a naming convention. I've come up with a few such schemes in my time (including the infamous Tarmak-ETROI and its ilk) and trust me, it ends in tears. You'll want something people can actually relate to like 'D-up', 'raiseD', or maybe 'swapD'/'swapM' for the left/right-hand side tricks. I'd then refer to 'swapD' as the minimal mod (D>P>G) or name it 'swapDPG' for clarity. 'Colemak-swapDPG/HM' is clear albeit brutal.

Last edited by DreymaR (26-Oct-2014 19:33:46)

*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

Offline
  • 0
  • Reputation: 0
  • Registered: 28-Sep-2014
  • Posts: 7

Like qwerty the hard part is learning to type words rather than letters.   Beyond the initial optimization to reduce the flailing that qwerty requires effort is far better spent on typing hand position and posture.   A quality 60% mechanical keyboard or a split angled or maltron like affair could be helpful.
If you really want to obsess then finding and remapping a datahand could be fun.

There are plenty of theoretical improvements to be had but none of them have been shown to overcome an individuals inherent speed of thinking and typing.   

The only thing I have seen that raises this bar is court based stenography where words and phrases are chunked together not unlike how competitive memorizers chunk digits using the major system.

So throwing out standard alpha keyboards entirely and focusing on phrases, words, trigrams, digrams could create fundamental improvements in keyboarding that can produce 200+ wpm with practice.

Offline
  • 0
  • Reputation: 117
  • From: UK
  • Registered: 14-Apr-2014
  • Posts: 975
tiramisuu said:

So throwing out standard alpha keyboards entirely and focusing on phrases, words, trigrams, digrams could create fundamental improvements in keyboarding that can produce 200+ wpm with practice.

You may be right, but in practice I'm only going to work with the tool that's readily available on my desk right now. Anyway, I don't dare primarily about speed. I type at best 60 wpm and am satisfied with that. For me the main driver is wanting a comfortable typing experience.

I like to think of it this way: Considering the left hand side mod only, if I spend 4% of my time tying D, and 2% typing G, then applying this switch results in a improved experience 6% of the time. OK so you can point to a marginally worse B position, but the overall net effect is still a significant gain in comfort.

Using Colemak-DH with Seniply.

Offline
  • 0
  • Reputation: 117
  • From: UK
  • Registered: 14-Apr-2014
  • Posts: 975

Thought I'd report back, having now adopted the left-hand part of the mod (the one referred to as "DVBG" above). I can say it's a definate keeper. The changed D takes a bit of getting used to, but having it in that new location feels really good. The G is nicer too in its Qwerty position. Kind of odd having a tiny bit of qwertyness back!

Some thoughts on the two possible mods to the D position:

Although I called it the "D>V>B>G" switch, actually this is perhaps unfair, as on a standard keyboard the V key really only "moves" to its original Colemak/Qwerty location. Assuming you are starting from vanilla Colemak, which seems the fair starting point for comparison, the mod is actually this:

1. apply angle mod to Z X and C keys.
2. apply the new locations to D, G and B.

So, not counting any angle mod adjustments, it's also only 3 keys changed. I should call it the "DBG" switch instead!

I feel that DBG+angle is far superior to DGP+angle for the following reasons:

1. I have come to realize that the bottom row index-finger key is actually a crucially important key, almost as important as the home positions! Having V there seems like a missed opportunity. It's much better to have a common key like D there. (This is actually the killer reason)
2. It keeps P unchanged in a comfortable position, and moves the rarer B to the harder-to-reach diagonal.
3. Given the choice of moving P or B, I'd rather move the B as it's rarer and awkward to reach on an unmodified Colemak/Qwerty keyboard anyway.

The only downside I can see is it looks a little less Qwerty-like with it's unusual B position, but that's a small price to pay for the significant gains.

Where I do have some doubts on the worthwhileness, is the right-hand (HM switch) part. I think it is a clear improvement, but that has to set against moving two previously unchanged keys. Since M is a mid-level frequency key, the gain is not as great as for the left-hand. Probably the right-hand switch is for people who struggle with the HE bigram or don't care about Qwerty similarity.

However, I think the case for the left-hand part of the mod is extremely strong. Certainly I shall be sticking with it.

Last edited by stevep99 (29-Oct-2014 15:21:15)

Using Colemak-DH with Seniply.

Offline
  • 0
  • Reputation: 210
  • From: Viken, Norway
  • Registered: 13-Dec-2006
  • Posts: 5,343

The awkwardness of reaching B has been one of the selling points for the angle mod.

To me, less QWERTY-like is bad. I liked the thought of keeping G in place, but if you move the B and V (yes, if it changes fingering I'll consider it moved!) you lose more than you gain.

But if you're happy with it, best of luck and keep trying it out. Maybe you'll find it a keeper in the long run or maybe you'll find it not quite worth the while as I've done with several tweak attempts. Hooray for clever tricks!

*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

Offline
  • 0
  • Reputation: 0
  • Registered: 28-Sep-2014
  • Posts: 7

I've started playing with Plover (stenotyping on qwerty keyboard) as a distraction from the frustration of learning Colemak.

It is quite similar to getting someone to punch you in the face in order to relieve the pain of a hang nail.   It makes me wonder if it was this painful 30 years ago in typing class on the old manuals.   The memory has completely left me.   

Colemak now seems like a trivially easy transition.   I highly recommend downloading plover and giving steno a try if you have a nkro keyboard.   It will make your hands shake and your head hurt :)

Offline
  • 0
  • Reputation: 210
  • From: Viken, Norway
  • Registered: 13-Dec-2006
  • Posts: 5,343

I've tried to feel what positions are the best and not-so-best and I came to a slightly different conclusion: The B/K positions aren't much worse than P/L, and feel better than D/H! Also, the upper-row positions aren't equal between hands because of the 1/4 stagger of that row, so J is quite a bit worse than G.

2 5   6 2
T 4   4 N
1 3   3 1

Heh.

Last edited by DreymaR (10-Nov-2014 16:20:56)

*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

Offline
  • 0
  • Reputation: 2
  • Registered: 25-Oct-2013
  • Posts: 136

Nice layouts Tiramisuu! But, perhaps it can be improved? I'd say the best way to make a really improved Colemak is the same as Aus der Neo Welt (adnw.de) improved on Dvorak, that is:

- Use the goals & constraints of, in this case, Colemak
- Choose which text corpus you want to optimize for. English, French, Slovak, Urdu? Or some mix? Adnw chose 50% German, 50% English, but if I would type Slovak and nothing else, i would prefer a layout optimized for that language.
- Calculate layouts, using an algorithm
- Test the best scoring layouts to see how they feel

I'd say that the goals & constraints of Colemak are:

- minimize distance, same finger use, row jumps, home jump.
- maximize rolls
- make sure there are more inward rolls than outward rolls
- aim for a 50%/50% in hand use
- some goal for finger load (pinky/ring/middle/index, maybe 8%, 12%, 16%, 14% ? ) 

I am not aware of formal rules on these goals. Keyboard optimization is all about compromises. What is better: a slightly lower distance or lower same finger? Are more rolls always good, even if it leads to heavy imbalance between hands? Etc. So you must set some arbitrary rules here.

HARD CONSTRAINTS:
- swap backspace & capslock
-  keep ,./ in place. Keep ZXCV in place. Use the following mask. The remaining 22 English alpha keys must be put in positions 01 through 22, other characters only on keys marked F 

01 02 03 04 05   06 07 08 09 F
10 11 12 13 14   15 16 17 18 19 F
 Z  X  V  C 20   20 21 22  ,  . / 

- of the remaining 22 leters, 6 must be put on their qwerty location
- only 2 letters may change hands, compared to the qwerty locations.


As said, ADNW did this for Dvorak. The resulting layout (made for 50%  English, 50% German) is as follows. For fun, I used the ADNW algorithm to calculate a version optimized for 100% English as well, I called that one DvorMax.  (PS, don't mind the ,'?/ etc, I only optimized the letters)
 

Dvorak  (optimized for English ??)
  ,.py fgcrl'
  aoeiu dhtns;
  ?qjkx bmwvz

Aus der Neo-Welt  (optimized for English/German)
  ku"./ vgcljf
  hieao dtrnsx
  y?,qb pwmzß

DvorMax (optimized for English)
  kyu." zlmdpv
  rieao hnstcw
  xß?,/ jqfgb

Is all this worth the trouble? The answer is Yes. Look at the scores (done with the same ADNW algorithm - I think it is OK to use a 'Dvorak inspired algorithm' to evaluate Dvorak-like layouts).


Lower is better: *   Higher is better: #

Dvorak           348.893 Total score(*)       186.358 Base score(*)        lefthand righthand
                   2.662 SameFinger(*)         13.058 Shift-SameFinger(*)    top     6.0 16.8
  ß,.py fgcrlö    70.517 Alternation(*)        33.563 Shift-Alternation(*)   mid    36.1 30.5
  aoeiu dhtnsü     1.617 In--/outwards(#)       6.624 IndirSameFinger        bottom  3.0  7.6
  äqjkx bmwvz     11.139 Neighbouring(*)       20.861 Shift-Neighbouring(*)  sum    45.1 54.9
                  9.7  8.3 13.0 14.1 --.- --.- 16.5 13.3 13.7 11.4            Sh     1.8  0.9

Aus der Neo-Welt 337.581 Total score(*)       189.071 Base score(*)        lefthand righthand
                   2.052 SameFinger(*)         17.995 Shift-SameFinger(*)    top  4.5 12.0
  kuü.ä vgcljf    70.173 Alternation(*)        25.534 Shift-Alternation(*)   mid 38.3 31.8
  hieao dtrnsx     2.043 In--/outwards(#)       7.458 IndirSameFinger     bottom  6.1  7.3
  yö,qb pwmzß      7.827 Neighbouring(*)       16.064 Shift-Neighbouring(*)  sum 48.9 51.1
                  8.9  9.6 12.9 17.6 --.- --.- 18.9 11.4 10.8 10.0            Sh  1.7  1.0

DvorMax          291.491 Total score(*)       179.388 Base score(*)        lefthand righthand
                   0.882 SameFinger(*)          7.423 Shift-SameFinger(*)    top  6.1 13.1
  kyu.ö zlmdpv    65.718 Alternation(*)        28.070 Shift-Alternation(*)   mid 39.5 31.4
  rieao hnstcw     1.016 In--/outwards(#)       6.177 IndirSameFinger     bottom  3.3  6.5
  xßü,ä jqfgb      8.665 Neighbouring(*)        9.226 Shift-Neighbouring(*)  sum 49.0 51.0
                  8.8  8.7 14.2 17.3 --.- --.- 15.8 11.0 14.3  9.9            Sh  2.0  0.8

FOR FUN  .... QWERTY
Qwerty           640.636 Total score(*)       346.732 Base score(*)        lefthand righthand
                   6.804 SameFinger(*)          6.299 Shift-SameFinger(*)    top 28.0 20.2
  qwert yuiopü    52.755 Alternation(*)        41.474 Shift-Alternation(*)   mid 22.1  9.5
  asdfg hjkläö     1.080 In--/outwards(#)      11.226 IndirSameFinger     bottom  6.8 13.3
  zxcvb nm,.ß     21.628 Neighbouring(*)       12.586 Shift-Neighbouring(*)  sum 56.9 43.1
                  9.1  8.4 18.5 20.9 --.- --.- 18.4  8.9 12.1  3.6            Sh  1.1  1.7

DvorMax scores better than Dvorak but also then Aus der Neo Welt (ADNW), which is logical. I made DvorMax for English only, so when tested with English only text, it will of course score better than ADNW. Used with German or mixed text, ADNW will score better.

Anyway, DvorMax and ADNW compared to the original Dvorak show:
- a clearly improved overall improvement (total score)
- lower same finger use
- the same alternation
- better balance of hands (51/49 for DvorMax and ADNW against 55/45 for Dvorak)
- lower neighbouring keys - in contrast to Colemak, Dvorak wants to avoid direct rolls, like (colemak) AR or (EI). So, by Dvorak rules, lower neighbouring keys is good.

(for illustation only: Qwerty scores bad with a high same finger, bad balance (57/43), bad total score, low alternation, uneven finger use. Scroll down in the black block to see Qwerty)

conclusion
I think that, using this same philosophy, Colemak may be improved as well. Not as drastically as Dvorak though, because Colemak is younger and better optimized (to the Colemak ideal) already. Plus there are a lot of constraints, so less room for improvements. But, there may be some good improvements, all within the Colemak framework. Or am I missing something here? Thoughts?

edited for formatting

Last edited by pieter (23-Nov-2014 17:20:12)
Offline
  • 0
  • Reputation: 117
  • From: UK
  • Registered: 14-Apr-2014
  • Posts: 975
pieter said:

I think that, using this same philosophy, Colemak may be improved as well. Not as drastically as Dvorak though, because Colemak is younger and better optimized (to the Colemak ideal) already. Plus there are a lot of constraints, so less room for improvements. But, there may be some good improvements, all within the Colemak framework. Or am I missing something here? Thoughts?
edited for formatting

I presume you are allowing keys to swap fingers. When you allow keys to switch fingers, you tend to run into problems with increased consecutive same finger, as Colemak is already well optimized in that regard. I also think that your hard requirements of keeping ZXCV in place plus at least 6 other keys in Qwerty locations mean improvements won't come easily. You might be able to come up with an alternative layout using this approach, but I doubt it would bear much resemblance to Colemak any more. By all means try it out though, it would be interesting to see what you came with.

Using Colemak-DH with Seniply.

Offline
  • 0
  • Reputation: 2
  • Registered: 25-Oct-2013
  • Posts: 136

@stevep99 If Colemak is only Colemak if it has ZXCV and lots of letters on the same locations as Qwerty, then improvements are indeed hard to get. And of course the algorithm will aim for a low same finger rate!

If I can get the source code to do what I want (must dive into that), I'd like to calculate of few versions:
- English only, Colemak strict (ZXCV + 6 other keys on Qwerty locations) 
- English only, Colemak free (XCV + 6 other keys on Qwerty locations)
- English only, Colemak free (XCV + 5 other keys on Qwerty locations)

- Language mix, Colemak strict (ZXCV + 6 other keys on Qwerty locations) 
- Language mix, Colemak free (XCV + 6 other keys on Qwerty locations)
- Language mix, Colemak free (XCV + 5 other keys on Qwerty locations)

For language mix, I may use several mixes: English/Spanish; or English/French; or German/Spanish; or EN/SP/FR/ ; etc.  Or optimized for one language, such as only Slovak, or Swedish or Spanish. This is actually very easy to achieve, the first part is harder.

Offline
  • 0
  • Reputation: 210
  • From: Viken, Norway
  • Registered: 13-Dec-2006
  • Posts: 5,343

To follow up on this topic (although the main discussion moved to a new one):

This is what I personally ended up with after some thinking and trying:

3 5   6 3
T 4   4 N
1 2   2 1

...with only a minor difference between 1 and 2. This led me to the D>B>G and H>K steps (or 'Dbg/Hk'), which I now use and like a lot. They're less intrusive than moving the V and M keys, and feel about as effective to me. It really depends on how you personally judge the '1' vs '2' positions in the figure above. Also, for a non-staggered keyboard they won't work but there Stevep99's Dvbg/Hm mod should still be good.

Last edited by DreymaR (15-Jan-2015 11:24:58)

*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

Offline
  • 0
  • Reputation: 117
  • From: UK
  • Registered: 14-Apr-2014
  • Posts: 975

Seems you've grown really like those (default colemak) BK positions!  I am not quite so convinced though.  Definitely they are better than I first thought, but I would always say I prefer up/down more than lateral motion, which would lead me to agree with your earlier scheme:

2 5   6 2
T 4   4 N
1 3   3 1

For some reason I seem to like the PL positions more than most other people.

Last edited by stevep99 (15-Jan-2015 13:00:34)

Using Colemak-DH with Seniply.

Offline
  • 0
  • Reputation: 210
  • From: Viken, Norway
  • Registered: 13-Dec-2006
  • Posts: 5,343

Maybe you have long index fingers! Their lengths do vary a lot between individuals, in part because of testosterone:
https://en.wikipedia.org/wiki/Digit_ratio
My index fingers are slightly longer than my index fingers though, so I'm in that camp too.

Or maybe it's about hand angle. After starting with the WideAngle mod, I noticed a dislike of the Colemak G position which was attributed to the improved hand angle. I don't know how much impact the hand angle has upon the PL positions though.

Another thought: My piano playing may have trained me for slight lateral movements.

Obviously, this is about personal preferences too. I'll try to support both our variants for the most part, as I see the need for both.

Maybe I am ready to make the Curl mods GUI selectable soon. That raises a new question: Whether to make them options so they can apply to any locale (but confuse the hell out of non-Cmk[eD] users!), or individual layouts but only available for the US locale? Because it'd be simply too much to have six eD variants for each locale: KS, US, KS-DH1/2 and US-DH1/2! I'm leaning toward the latter alternative – this'll give the mods some exposure and make the default layout easily selectable, while still giving locale users the option to edit the colemak file to get any locale variant.

*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

Offline
  • 0
  • Reputation: 117
  • From: UK
  • Registered: 14-Apr-2014
  • Posts: 975
DreymaR said:

Maybe you have long index fingers! Their lengths do vary a lot between individuals, in part because of testosterone:
https://en.wikipedia.org/wiki/Digit_ratio
My index fingers are slightly longer than my index fingers though, so I'm in that camp too.

Yeah, mine look to be about the same length. I'd say on my left hand, index finger is slightly longer than ring. Yikes, looking at the list of traits doesn't make for happy reading, does it?  Mind you, it doesn't say how strong the correlation is. I suspect in reality the correlation is very weak.

GUI selectable sounds really cool. Some casual Linux users (like me) would probably look for that first. I'm not sure what the significance is of the locale question in terms of what it means to configure it - but wouldn't US locale probably mean an ANSI keyboard - which would make these mods less attractive to begin with? Probably most converts would be non-US (or at least ISO keyboard) users...

Last edited by stevep99 (15-Jan-2015 18:59:11)

Using Colemak-DH with Seniply.

Offline
  • 0
  • Reputation: 210
  • From: Viken, Norway
  • Registered: 13-Dec-2006
  • Posts: 5,343

From what I gather, the correlations are quite strong actually. But not to worry, it mainly means that we don't suffer overmuch from testosterone poisoning. We may have some side effects such as a longer projected life span, better relationships and a few other problems along those lines...

The locale has nothing to do with the keyboard model really, that's one beautiful thing about xkb. And the pc105 ISO keyboard model is actually usable for the pc104 ANSI keyboards too; it just has another graphic on the "Keyboard Layout Chart" and otherwise even the extra key gets defined although you wouldn't be able to actually press it on an ANSI board. So US locale doesn't mean much and for me it's the "neutral ground" of sorts, where Tarmak and mirrored Colemak and everything that hasn't got special locale symbols in it resides. Therefore, it's the place for this mod as well.

As mentioned elsewhere, I actually have an ISO keyboard made in USA with US key labels. Gotta love that. The ISO keyboard was developed by IBM as an improvement, but the conservative US market didn't want it because of the distance to the left Shift (which is no more than the distance to the right Shift dammit) and the shape of Enter (which I don't have a problem with... using the Wide mod; otherwise I see the point).

*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

Offline
  • 0
  • Reputation: 117
  • From: UK
  • Registered: 14-Apr-2014
  • Posts: 975
DreymaR said:

From what I gather, the correlations are quite strong actually. But not to worry, it mainly means that we don't suffer overmuch from testosterone poisoning. We may have some side effects such as a longer projected life span, better relationships and a few other problems along those lines...

Nice! Plus optimism as well, perhaps..  :)

DreymaR said:

As mentioned elsewhere, I actually have an ISO keyboard made in USA with US key labels. Gotta love that. The ISO keyboard was developed by IBM as an improvement, but the conservative US market didn't want it because of the distance to the left Shift (which is no more than the distance to the right Shift dammit) and the shape of Enter (which I don't have a problem with... using the Wide mod; otherwise I see the point).

For a long time I also looked enviously at ANSI keyboards with their easier-to-access left-shift key. I think the Return key is better on ANSI boards too. I had even considered switching to an ANSI board in the past - but that was before I discovered the power of the angle mod of course.  Were it not for the possibility of the angle mod, I would say ISO is not an improvement at all.

A nice idea for ANSI users (even with Qwerty) would be to apply the wide mod - then both shifts would be easy to access and in symmetrical positions.

Last edited by stevep99 (18-Jan-2015 17:23:49)

Using Colemak-DH with Seniply.

Offline
  • 0
  • Reputation: -5
  • From: Banned
  • Registered: 02-May-2015
  • Posts: 116
DreymaR said:

I've tried to feel what positions are the best and not-so-best and I came to a slightly different conclusion: The B/K positions aren't much worse than P/L, and feel better than D/H! Also, the upper-row positions aren't equal between hands because of the 1/4 stagger of that row, so J is quite a bit worse than G.

2 5   6 2
T 4   4 N
1 3   3 1

Heh.

You forgot, that there is ½ stagger in bottom row too.

R   T   Y   U

 F   G   H   J

   V   B   N   M


P   G   J   L

 T   D   H   N

   V   B   K   M


3   5   6   4

 0   4   4   0

   2   7   2   1

Banned from Colemak

Offline
  • 0
  • Reputation: 210
  • From: Viken, Norway
  • Registered: 13-Dec-2006
  • Posts: 5,343

The standard is 1/2 stagger from home to bottom row. And no, I haven't forgotten that? But maybe Piotr forgot that I'm calculating with the Angle mod? Indeed, the old QWERTY B position is very bad (1.5 keys lateral stagger) – almost as bad as the QWERTY Y position (1.75 keys). I'm referring to the QWERTY C vs V positions on the bottom left row, both having a lateral stagger of 1/2 keys.

The assessments Piotr quotes are not current though. I almost agree with him:

 3   5   7   4

  0   4   4   0

1   2   6   2   1
Last edited by DreymaR (31-May-2015 13:55:11)

*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

Offline
  • 0
  • Reputation: 210
  • From: Viken, Norway
  • Registered: 13-Dec-2006
  • Posts: 5,343

Hola, stevep99 and whomever it may concern! :-)

I took a look at the key effort model over at the mod comparison page, and realized that I'm almost in full agreement with the model now (a few decimals notwithstanding).

    keyboard_scores.png
    Stevep99's assessment of key effort

Just a few points, but to me they're fairly interesting ones:
• Why not make the color scheme go 1.0–1.5/1.6–2.0/2.1–2.5 etc? That'd make the upper and lower rows more nuanced (which I think would be right).
• (Sure, to be perfectly consistent the 1.0 keys should then have a separate color but I don't think anyone will push that point.)
• My pet peeve would be to petition for a weight of 2.5 instead of 2.6 on QWERTY V and N! With the color scheme fix above, that'd make them dark blue. ;-)
• I do believe that both the outward index finger curls and the ring finger curls deserve a nicer color than the nastier "purple" keys.

Should you disagree with the above, I would contend that you should then at least change the little finger home keys to cyan. That'd be an advancement to the cause too!

What do you say? Anyway, thanks for your efforts in making typing even more fun!

Last edited by DreymaR (14-Aug-2017 14:19:44)

*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

Offline
  • 0