• You are not logged in.

    Colemak's Cousin

    • Started by iandoug
    • 20 Replies:
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67

    Hi all

    I had another go at making a layout that kept ZXCV and right hand punctuation as per QWERTY, except for semicolon of course.

    The result ended up as a partial mirror of Colemak, though it was developed from scratch / first principles.

    Ended up with with two variants, the Fingers version does better on my version of KLA KLAnext, while the Words version does better at word metrics and overall.

    To get from Fingers (attached) to Words, rotate O A U one move counter-clockwise so that U is above E and O and A are both on the home row.

    Cheers, Ian

    Attachments:
    Offline
    • 0
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67

    Difference to Colemak.

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

    Looks like the biggest change is a swapping of index fingers e.g. T,P,G,D,B now right index instead of left, but otherwise pretty similar. Not sure what benefits it would have. Possibly the anti-pinky people might like it because the O moves. But then, the O-U swap could be done even in vanilla Colemak.

    Using Colemak-DH with Seniply.

    Offline
    • 0
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67
    stevep99 said:

    Looks like the biggest change is a swapping of index fingers e.g. T,P,G,D,B now right index instead of left, but otherwise pretty similar. Not sure what benefits it would have. Possibly the anti-pinky people might like it because the O moves. But then, the O-U swap could be done even in vanilla Colemak.

    T works better on right hand because distance to Shift key is less. And Th very common start to sentences.
    For same reason, having A and I on right hand. (in "words" version).

    Most frequent capitals in English: T A I S C E D M N R P O B

    Cheers, Ian

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

    T works better on right hand because distance to Shift key is less.

    ah ok, so I suppose you are assuming:

    - using ANSI not ISO board.
    - using the "official" opposite-pinky shift method.

    I guess that applies to some people, but I do wonder how many type using these supposedly "correct" methods in practice. It's not a factor for me, as there's no way I'd use the standard Shift keys, they are horrible and not fit for purpose in my book. So much so that I think having a better way to shift is much more important than optimizing the letter keys to the nth degree.

    Last edited by stevep99 (21-May-2021 16:50:27)

    Using Colemak-DH with Seniply.

    Offline
    • 0
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67
    stevep99 said:
    iandoug said:

    T works better on right hand because distance to Shift key is less.

    ah ok, so I suppose you are assuming:

    - using ANSI not ISO board.
    - using the "official" opposite-pinky shift method.

    I guess that applies to some people, but I do wonder how many type using these supposedly "correct" methods in practice.

    I didn't know there was another way on ANSI. :-)
    Apparently the previous speed typing champ only uses Caps Lock.

    We only get ANSI keyboards here, no ISO on sale is shops. And if there were. they would be UK version with symbols in the wrong place :-)

    Philipp Kiefer took my Uciea layout and applied some MiniMotion tweaks, which put the shifts on the index finger bottom row. I wasn't wild about that so I tweaked it further. This is current best layout on KLAnext, for ANSI, but it won't be everyone's cup of tea...

    Cheers, Ian

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

    I just meant that on ISO keyboards, both shifts are equally distant from the home positions.
    If you apply the Wide Mod, right shift is actually closer.

    That bottom row shift idea doesn't look much better to me. There are various of proposals around for shifts, including home row mod-tap, caps-lock/apostrophe (optionally sticky) and thumb keys. IMO thumb keys are the best option.

    If limited to standard boards, I'd say RightAlt makes for the best shift position. Alternatively, I've found that if using the Wide Mod, spacebar makes for a great shift key, with RightAlt being used for space. It might seem strange to reassign space, but that arrangement feels surprisingly comfortable.

    Last edited by stevep99 (22-May-2021 10:46:49)

    Using Colemak-DH with Seniply.

    Offline
    • 1
    • Reputation: 210
    • From: Viken, Norway
    • Registered: 13-Dec-2006
    • Posts: 5,343
    stevep99 said:

    - using the "official" opposite-pinky shift method.

    I guess that applies to some people, but I do wonder how many type using these supposedly "correct" methods in practice.

    Not many, it seems! It's a common topic in Colemak and Monkeytype Discord channels, and my impression is that few do the alternate-shift thing – at least, not consistently.

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

    Offline
    • 0
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67
    stevep99 said:

    I just meant that on ISO keyboards, both shifts are equally distant from the home positions.

    Sorry,  I don't think so ... at least not centre-to-centre as KLA measures.

    Cheers, Ian

    Offline
    • 0
    • Reputation: 210
    • From: Viken, Norway
    • Registered: 13-Dec-2006
    • Posts: 5,343
    iandoug said:
    stevep99 said:

    I just meant that on ISO keyboards, both shifts are equally distant from the home positions.

    Sorry,  I don't think so ... at least not centre-to-centre as KLA measures.Cheers, Ian

    That's hardly a matter of opinion? The left shift key on ISO is 1u down, 1.5u left from the centre of the QWERTY/Colemak A key. Correspondingly, the right shift key is exactly the same distance from the Colemak O (QWERTY semicolon) position. If you calculate distances to the center of the shift keys, I think you're doing it wrong. Unless you have some kind of proof that most typists hit the center of their Shift keys, which I patently do not. I hit the Shift keys like other 1u keys, at the minimal stretch, and it doesn't make sense to me to do otherwise. Of course, the keys being so wide there is some leeway at high speed I guess but in general I use the short stretch.

    Calculating to the center of the Shift keys doesn't even work, in fact. On my laptop, the left Shift key is 1.25u and the right Shift key 3u. On other boards, the widths vary greatly, between 2.75u and 1.25u wide RShifts. So that kind of measure would be very very wrong to use. I don't think any wise typist would hit the 2.75u right Shift key exactly on the middle...

    https://deskthority.net/wiki/Shift_key

    Last edited by DreymaR (23-May-2021 16:30:31)

    *** 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:

    If you calculate distances to the center of the shift keys

    I think that might be the problem, as the patorjk analyzer does this, and both iandoug's and my forks are based on it. It's a good example of where analyzers shouldn't be taken too literally, as they typically make simplifications and approximations for the sake of making calculations tractable, but unfortunately the real world is a lot messier. I certainly wouldn't rely on any of those analyzers to accurately model use of shift keys, space bar, modifiers, etc - at least, not without some careful thinking about and accounting for how they are used in practice in some way. I did add Fitts' Law to my fork a few weeks back, including the idea that larger keys are easier to hit, so this effect of large keys being further away should be mitigated.

    Last edited by stevep99 (23-May-2021 17:01:18)

    Using Colemak-DH with Seniply.

    Offline
    • 0
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67
    stevep99 said:
    DreymaR said:

    If you calculate distances to the center of the shift keys

    I did add Fitts' Law to my fork a few weeks back, including the idea that larger keys are easier to hit, so this effect of large keys being further away should be mitigated.

    :-) ... and I took it out of mine (along with Exponential Distance) ... it benefitted "wide" / right-shift layouts without achieving the intended aim of trying to punish sub-optimal placement of high-use keys. I feel the various "wide" layouts are fixing the wrong problem ... ANSI / ISO sucks, get ortholinear style... :-).

    It would be nice if the Chinese started churning out affordable boards, even if they are membrane switches ... just to get people onto a better form factor. </rant>

    Since then I've categorised ANSI layouts into "conventional" and "unconventional" where "conventional" implies that these (at least) are as per US ANSI:

    1. home keys
    2. key-finger assignments (as per touch-type accepted theory)
    3. space key
    4. enter key
    5. both shift keys
    6. AltGr

    I've also been having discussions with Phillipp about how to include some use of the backspace key, at say 5% or 10% (or x% where I have no idea what x is...), in which case I would need to add 7. Backspace to that list, at which point all the Colemak variants suddenly become "unconventional", which creates political rather than technical problems ... :-)

    Maybe my implementation of Fitt's was wrong / too heavy, because it also had Den's other distance penalties as well.
    So I will probably relook at it again in the future.

    Cheers, Ian

    Last edited by iandoug (24-May-2021 09:02:53)
    Offline
    • 0
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67

    Input: Alice Ch 1.

    Steve's scoring vs mine. Wide variant wins.

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

    Please never ever use Alice Ch. 1 as corpus, for anything...?

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

    Offline
    • 0
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67
    DreymaR said:

    Please never ever use Alice Ch. 1 as corpus, for anything...?

    Yes. you are right :-)

    I was just trying to make a point and used Alice because it's the go-to for such things... other inputs would be similar, the other layouts would probably be in a different order with better input.

    Wide layouts score higher because RShift and Enter are closer to pinky home, not because their letter arrangement is better, which is what the game is all about.

    Cheers, Ian

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

    Alice may have used to be the go-to, but it really shouldn't be. But for this particular purpose I suppose it sufficed.   ̄(=⌒ᆺ⌒=) ̄

    When Wide layouts score better because of a smaller Shift and Enter travel, that indicates that their arrangement is indeed better, which coincides with my opinion of them as a Wide (CAWS) user. So that's definitely part of the game! However, it's unfair and quite uncertain to compare Wide and non-Wide layouts on these analyzers. Punctuation placement, Wide/Angle and similar mods are things I much prefer to keep modular so you can add them as you please without changing the main layout's properties much.

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

    Offline
    • 0
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67
    DreymaR said:

    Alice may have used to be the go-to, but it really shouldn't be. But for this particular purpose I suppose it sufficed.   ̄(=⌒ᆺ⌒=) ̄

    When Wide layouts score better because of a smaller Shift and Enter travel, that indicates that their arrangement is indeed better, which coincides with my opinion of them as a Wide (CAWS) user. So that's definitely part of the game! However, it's unfair and quite uncertain to compare Wide and non-Wide layouts on these analyzers. Punctuation placement, Wide/Angle and similar mods are things I much prefer to keep modular so you can add them as you please without changing the main layout's properties much.

    In the interests of The Greater Good I knocked together a new "test" corpus made up of 40kB chained English bigrams, 10kB chained Code bigrams, and 10kB chained Proglish bigrams, which is itself an 80:20 English:Code mix. The is the same ratio I'm using for current round of keyboard evaluations (thankfully neartly done).

    I've included my ansi version of Mod-DH as well as Steve's iso version, and was surprised by the result ... was expecting the opposite.

    Wide still wins.

    Re your other comments, that's why I've split ANSI. Current categories for the evaluation are:

    ANSI Top 30                                 Ergo Top 30
    
    All                                         All
    Conventional                                No AltGr layer
    Unconventional                              Uses AltGr layer
    Conventional + QWERTY shift pairs           No letters on thumbs
    Conventional + QWERTY shift pairs + ZXCV    No letters on thumbs or AltGr

    Cheers, Ian

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

    I lost track of what we're trying prove here, but the conclusion of iandoug's experiments seems to be that the Wide Mod is an excellent mod from the analyzers' point of view. But you can apply the Wide Mod to most layouts, so if we are trying to access how good a mod it is, wouldn't we compare Colemak Std vs Colemak Wide, Colemak-DH Std vs Colemak-DH Wide, etc? For layout comparisons, I also think it's better to compare using all ANSI or all ISO, not mix and match, otherwise the change form factor affects the results. My analyzer does have both ISO and ANSI versions of Colemak-DH BTW.

    And of course, I can't help but mention an extra thing that makes a huge difference in practice but is unaccounted for in the analyzers: the Extend Layer!

    Using Colemak-DH with Seniply.

    Offline
    • 1
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67

    In my view, our aim is to find the optimal letter layout for an English keyboard. Obviously this immediately discounts ANSI/ISO because it is bad in so many ways, but since the installed base is so huge (not to mention being Official Standards (still)), we try to make the best of it.

    Wide mods (or refingering) do not improve the letter layout. What they improve is some intersection of the ANSI form factor (which in itself is a kludge to mimic physical typewriters because anything else would imply retraining costs...)   and how QWERTY was translated to ANSI.

    Originally M would have been on right pinky, which is a lot better than ;.
    (hotlink to Wikipedia failed, so attached)

    I know this layout changed early on.

    So yes, wide mods score better, but they're exploiting flaws in ANSI/ISO, not improving same-finger-bigrams, for example.

    Other hacks which move the shift keys or enter key are even more effective.

    [The original point was Colemak's Cousin but we've drifted a bit into problems with using Fitt's Law in KLA... :-) ]

    Cheers, Ian

    Last edited by iandoug (25-May-2021 09:58:02)
    Attachments:
    Offline
    • 0
    • Reputation: 210
    • From: Viken, Norway
    • Registered: 13-Dec-2006
    • Posts: 5,343

    The big question with that is whether to keep the symbol keys away or not. The semicolon is already messed up in the letter block, so for simple analysis I just use that as the "extra symbol" even with layouts that put something more useful in that spot (like apostrophe). Just to keep analysis to the letter block only, as it were.

    However, more optimized layouts tend to move symbol keys anyway, and then the bets are off aren't they? For instance, the E-comma SFB on Colemak makes some modders swap comma and slash. Analyzers have mixed opinions on the end result of that, but it does wonders for the SFB stat at least.

    So if we can move symbol keys, then the Wide mod will be a part of that. You could optimize a layout differently for a Wide configuration than for a non-Wide one.

    My basic philosophy is, like it seems you want to do, to look at the main letter block separately from symbol mods and geometric mods like Wide. So in a simple world, I think I agree with your approach. But it will fail to pick up on some things like the effect of symbol-optimized layouts. MTGAP is such a layout. People say it's better because it scores highly in analysis; I don't know much about its merits but it does change not only symbol key positions but layer mappings. What an analyzer makes of that is a good question!

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

    Offline
    • 0
    • Reputation: 7
    • Registered: 18-Nov-2017
    • Posts: 67
    DreymaR said:

    The big question with that is whether to keep the symbol keys away or not. The semicolon is already messed up in the letter block, so for simple analysis I just use that as the "extra symbol" even with layouts that put something more useful in that spot (like apostrophe). Just to keep analysis to the letter block only, as it were.

    My basic philosophy is, like it seems you want to do, to look at the main letter block separately from symbol mods and geometric mods like Wide. So in a simple world, I think I agree with your approach. But it will fail to pick up on some things like the effect of symbol-optimized layouts. MTGAP is such a layout. People say it's better because it scores highly in analysis; I don't know much about its merits but it does change not only symbol key positions but layer mappings. What an analyzer makes of that is a good question!

    This is basically what I asked a few weeks back in the thread "What makes a fair fight?".

    It's also why I put layouts like MTGap in different category to Colemak, because MTGap changes the QWERTY shift pairs, and also does not have ZXCV, which is a major constraint (forcing A and O to the outer fingers).

    MTGap beats Colemak on Patrick's KLA, but it's the opposite on mine. Patrick is ignoring distance on home keys (and vertical distance in general).
    When I started, and was using Patrick's version, MTGap was always "the one to beat".

    I have finished all my testing, busy trying to finalise the front end to the results (and update the Letter Layout DB), but the comparison between well-known layouts is a bit different now to the version I posted a few weeks back. Attached.

    Note: uciea-minimo3 is Very Unconventional. Optimal-trigram changes shift pairs.

    Offline
    • 0