'Best' is a difficult title to bestow on any layout, since they are all designed with different goals in mind and with different sacrifices to try and reach those goals.

QWERTY is the most common and everyone already knows it, and thus a vast majority of people consider it the 'best' because their main goal is familiarity.  Dvorak was designed independently of QWERTY, and so has trouble co-existing in a QWERTY world, but it is also the oldest and most popular alternate layout.  There are many Dvorak users that don't consider the ZXCV block to be a more valuable goal to aim for than a ground-up design with no compromises.

The most common opinion here is that Colemak is the 'best', because we think that staying as close as possible to QWERTY for the sake of keyboard shortcuts while still having a solid home row is the most important goal for a layout.  On the other hand, our layout is not installed by default on most machines and is different than the first and second most common layouts... those are the trade-offs that we make to reach our goal, and there are many people that find them unacceptable.  To them, our preferred layout is not the 'best'.

I will admit that I don't know much about Capewell, but that is really a moot point, since I've already decided what the 'best' layout for me is.

The only way to answer your question is for you to do a bit of research and see what is the 'best' for you personally.

that's great to hear =)

I haven't heard about anyone using a Capewell layout (I'm sure they exist but I don't hear from them) - you may be asking too much. I find it unlikely that many people have tried both more than briefly.

You have to realize that thorough research is required to answer your questions about ergonomy and speed, and such research would require funding.

Other than that, I'd think that Capewell/Dvorak/Colemak are all fast and ergonomic. It's possible (but hard to prove decisively) that Capewell is a bit faster - but by how much if so? (1%? 2%? More? Less?), and it's quite certain that Colemak is easier to learn coming from QWERTY and also easier to use if you have to use QWERTY as well at work or suchlike.

Most people here would agree that it's certainly worth ditching QWERTY for a more ergonomical layout for good. If you ask on this forum, guess which one we'll recommend?  ;)

Interesting blog there, SpeedMorph. Thanks for the read.

By default, I'll have to chime in here and tell that for me at least, the Colemak has been fully as fast and simple as the Dvorak - and I think I've used them both enough to judge by now. In fact, I'd say that the Colemak is faster for me, since I've managed to beat my old QWERTY and Dvorak test results by a fair bit with it (65 WPM over time vs. 62 for Dvorak and around 60 for QWERTY, iirc). And it's easier for me since it makes going back and forth between my layout and QWERTY (mainly at work and when I'm visiting people) much easier, it makes using Windows hotkeys much MUCH easier (I remember cursing whenever I'd hit Ctrl-W next to the Ctrl-V I wanted, thereby closing the window I tried pasting into) and... well... it just feels easier from what I can tell. My memory is probably far from infallible in that matter, of course.

I do realize that you never said anything else - you just pointed out your own initial reaction. Well, I hope you understand that I had to chime in nevertheless. That, and your starting lines were too good to not spark a bit of fun. :)

You've found out yourself that the most common letters (and digraphs too?) account for a lot of the action. A central tenet behind Colemak is that you can make lots of layouts that are good "enough", meaning that whatever speed or comfort there is to be gained beyond that point is fairly inconsequential and probably far less than the standard deviation in a population of trained typists anyway. We believe the Colemak to be one of these good layouts, meaning that if the Capewell or the Dvorak or whatever other layout you compare it to has a measurable speed/comfort benefit over the Colemak it'll be so small as to make no real difference (certainly so small as to require some solid research to prove!). So that's where the additional restrictions actually work to our benefit: Since the layout is already one of the good ones, the added perks of easy learning and easy interconversion with QWERTY and easy hotkey use in several systems/apps are almost pure benefits!

The problem with those statements is that we can't back them with research so we're left with just our own experience of it and that argument isn't waterproof by any standard. You'd just have to take the words of a small bunch of enthusiastic users on this forum, and some people naturally won't want to do that on face value. But so far, it's what we got. The theoretical tests that have been conducted have shown only small differences between the good layouts, and these are heavily influenced by how you weight the various quality measures - or for that matter, what your test syllabus consists of. Every person has an individual usage - I, for instance, write predominantly English of a fairly academic nature but I also use my own language Norwegian (mostly New Norwegian). As an example of individual usage, if you use Linux or Unix on a lot of different computers you may be annoyed with Dvorak's awkward placement of the LS digraph, and as a Norwegian I was very annoyed with how the Dvorak handled the second most common letter in Norwegian (R). I'm not saying that a layout should take my own small language into consideration, just pointing out that there are many different parameters playing in for many users. It's a tricky matter for sure.

It's a consolace for me as an individual user that even in Norwegian the Colemak seems to be one of the good layouts (we had a long thread about different languages a while ago). I doubt anyone could make one that's so much better as to make many users take the bother of learning a specialized and hard-to-learn layout to gain such a small edge that might never materialize at all - and that would be less efficient with a different usage pattern.

They all improve significantly over QWERTY according to models based solely on my own typing (using them for others is a big leap but we have nothing else to really go on that is of much more stature than that besides Dvorak's own work).  With that caveat, here's what applies to me:

Besides layouts I've made myself, which ignore things that are probably crucial to you such as ergonomics (in favor of speed-only), the layouts that consistently win out in my models are carpalx's.  So if you want a layout that has some similarity to QWERTY, I'm sure he has a very good one for you.  After that, and again I'm talking only about speed, so compare other features such as ergonomics for yourself (I assume everyone I'm mentioning in this post does a decent job), it's really a wash between Asset, Colemak, and Capewell-Dvorak (NOT Capewell-evolved).  Here's my "numbers" so you can apply your own taste:

QWERTY:  100% of QWERTY speed, 100% of QWERTY distance.
Dvorak:  109% of QWERTY speed, 45% of QWERTY distance.
Colemak:  113% of QWERTY speed, 45% of QWERTY distance.
Capewell-Dvorak:  112% of QWERTY speed, 45% of QWERTY distance.
Asset:  113% of QWERTY speed, 46% of QWERTY distance.
CarpalX Full:  117% of QWERTY speed, 50% of QWERTY distance.
CarpalX 10 Swaps:  115% of QWERTY speed, 57% of QWERTY distance.
CarpalX 5 Swaps:  111% of QWERTY speed, 64% of QWERTY distance.

This is how I would estimate it, and it's based on data from my own typing _only_, so take it with a grain of salt.  That being said, others seem to say it sounds right, and I think so too.  I think I'm barking up the right tree here.  As you can see, they pretty much are a wash.  Just get away from QWERTY if you want to drop that fat distance, and make sure you gain some ergonomics, too (to your taste, I suppose, else Dvorak is the only true authority).

I could list a few others, but my bet, if I had to make one on someone besides myself, is with CarpalX.  He seems to have taken conservative ergonomic assumptions and ended up coming up with faster keyboards that don't overly emphasize lowest possible distance (I'm assuming that the slightly higher distance is actually something you're paying in order to get superior ergonomics, which seems to be confirmed somewhat by the "side effect" of more speed, something he doesn't really shoot for at all).  Again, making wild assumptions, my taste is that keyboards seem to focus only on lower distance and same finger digraphs, with minimal ergonomics.  I feel like Capewell surpasses this, but CarpalX takes the cake.  My own layouts only optimize speed and probably destroy your wrists, but if you want a number to shoot for, imagine a layout with 135% the speed of QWERTY and 50% the distance--thats what I'm trying to learn, but I may end up making a new layout with more ergonomic standards and probably less speed as a result.

So pick one you trust the "other features" of--they all are better than QWERTY _for sure_, unless they have such huge errors in their ergonomics that they actually manage to hurt you more than QWERTY (which you probably know we have no reason to think has _any quality_ in terms of injury-reduction other than those that come accidentally with old technology jam-reduction and typing TYPERWRITER on the top row starting with an alphabetical layout etc.).

I also think Asset has been overlooked; it seems on par with Colemak, but perhaps Colemak is also on par with it and has more features or whatever.  Just get off of QWERTY if you are looking for something more.

I took my amphetype trigraphs and took categorized them according to what they had in common, i.e., trigraph type A consists of XXY where X can be A B or C and Y is P or Q, etc.  It's complicated but basically I focused on patterns that were excessively slow and patterns that were excessively fast.  These are not letters based but motion based.  In QWERTY I found that the "slow class" I defined corresponds to 60.1% of all typing, whereas in Dvorak it's 37.8%, Colemak 41.2%.  The "fast class" I found to be the least optimized, but there are still significant differences between popular layouts in this category, perhaps by accident (my guess is that some of Dvorak's ergonomic optimizations actually _minimized_ the natural amount of "fast trigraphs" that QWERTY already has--Dvorak's excessive hand switching prevented most of the kinds of fast trigraphs in my model which of course only applies to my own typing).  Dvorak has fast trigraphs only 11.9% of the time, compared to QWERTY's 14.3% of the time, and Colemak's 17.4% of the time.  When I take the estimated speed of the slow trigraphs and the fast trigraphs and apply what its overall effect should be, I end up with speed estimates which I can compare across layouts.  So while Dvorak has the "least bad typing" you might say, Colemak makes up for it with "more good typing", which is why it has better speed according to my prior post.  I have made a layout that gets "bad trigraphs" as low as 28.3% (while still being decent overall--not a suicide pact layout), and another layout (my preferred) that gets fast trigraphs to take up 27.% of all trigraphs typed--a big gain because IMO it's largely untapped.  CarpalX's layouts also do well in this regard which gives me more confidence that my model of my own typing is not completely worthless.  In this respect I also rank Asset and Capewell about Colemak, but not really enough to outweigh the overall package of Colemak, so Colemak really remains the single fine choice for an overall package, unless you don't care for its features at all and only want fast typing or a different layout's apparent ergonomics.