How good a layout is is very subjective. There are multiple factors to be taken into account, whose importance is different for everybody. So ultimately a layout, better than Colemak or Dvorak, would be a layout which performs better in each of these factors.

Now if you pick Colemak for example and try to improve any of the factors in play, you'll reduce another. So in other words I don't think there will ever be a layout that is ultimately better than Colemak (or Dvorak). Which also doesn't make either of them the best for everybody.

I personally agree with you about the finger load. There are quite a few layouts out there and if everyone makes a decent research I think most of people's requirements should be met by one of them. For the more enthusiastic people, there's always the harder way – of optimizing or designing their own layout. Though if that's first contact with alternative layouts I'd recommend try one (or even two) of the established ones for a few months first.

If I were to make a new layout for myself I'd probably try and make a design with the lowest same finger ratio possible while considering alternative fingering for some key combinations. Colemak is very good at lowering the same finger ratio and alternative fingering is not very often needed. But if it's included in the design factors maybe someone could come up with more advanced layout which requires specific finger movements for different digraphs to be optimal.

Ideally with every person of different finger lengths and different hand physiology, there will be an optimized layout for each individual.

I am sure that there will be another layout which is 0.1-0.2% better than Colemak in every aspects you guys mentioned, but it may take several months of computing powers to do the calculation and optimization, since Colemak is pretty optimized

As UltraZelda64 said we don't really know how optimized Colemak is. but it certainly beats all other existing layouts on many metrics.

I totally agree with that.

That would be true only if you want to optimize for the same factors as Colemak...

It's the same with me:

Edit: Don't use this layout, it has some bad problems, like typing words like 'you' is very uncomfortable.

h    l    d    w    b    k    y    u    m    
n    r    s    t    g    p    e    a    i    o
z    x    c    v    q    j    f    

True, it's a trade off. In my layout I sacrificed same finger use to place some less used letters on the center columns.

the truly ergonomic looks awesome!!

Predictive text sucks big time, unless you're predictable. I'm not, and spent more time correcting the correcting mistakes than I saved.

I don't think so. I type English and Norwegian with a mélange of foreign words in the mix. I've yet to find a text correction system that's up to my standards, and I've come across quite a few. If you're a language person, I just don't think the predictive implementations are up to the challenge!

I do realize that other people have such simple needs that they don't even use the AltGr key and only type in one language! For them I guess it's okay. For people like me who on occasion type words like 'Þröstur', 'Nhất' and 'κάκιστος' it's hopeless. ;)

Another data point: in my day job I work with hundreds of servers, all with "funny names", where prediction will not work either.  Some are existing English words and names, some just resemble existing words, and others have sequential numbers in them, etc.  I don't want to (and can't) pre-enter them all, as we have new ones every day.  Just entering those names character by character on a keyboard is reliable and predictable though.

So I agree with DreymaR that prediction is not a solution.  It may be an optimization for many use cases, but when designing a new input method it should always be optional, and not an essential part of the solution.

Back to the original question:

Yes, there is room for a layout better than Colemak or Dvorak. I think efforts on alternative layouts which ignore arrow keys are solving only part of the problem. Computers are not typewriters, we don't "enter" text, we edit text. Arrow keys are high-frequency, they should be in good spots.

For example (no claims to statistical significance), here is my usage data for, I think, a couple of hours of computing: some writing, some mail, some programming. (Letters are "what's printed on the keys", so the high-frequency "k" is really Colemak "e".)

The top-10 keys are Backspace, space, enter, command (Mac Os, like alt on Windows), shift, downarrow, e, tab, uparrow, i.

Of these, only 3 have decent spots; the rest are either conspiring to kill my wrists and pinkies out on the edges (backspace, enter, command, shift, tab), or focus on my wrists (arrows).

Similar story for Xah Lee's Emacs Command List Frequency

Some of these keys have little to do with the layout. If you want comfortable positions for modifier keys, arrows Backspace or anything but a letter, I'd consider something like DreymaR's extended mapping. Or if you want something simple to start with, just think of a decent shortcut and map it to an awkward key you use often. I for example have Alt+NEIU (on Colemak) to send Ctrl+← ↓ → ↑, which is really good for jumping through words in a text and not only. I also have shortcuts for ctrl+backspace, which I use all the time and many others.

As for the frequency of general use – it would be completely different from person to person, so it'd be hard to make a solution that suits everyone.