I am of a similar opinion to NottNott. Mod-DH changes enough to be useful and doesn't mess with what isn't broken, on the whole.
One thing you should take into account is that metrics, no matter how robust, are usually accounting for factors common to users as well as the personal views of what the individual behind the metric considers important. It's a good starting point, but not ideal because there is no effort to address a deeper issue - individual users adapting to the layout and their comfort levels with long-term usage, which may not always follow the data obtained. It's possible to study the human factor in depth, but not by designing keyboard metrics and having a computer give you the results without extensive human testing.
Another issue is the flawed nature of keyboard entry generally. It's actually pretty sub-optimal by nature when you think about it, with fingers being responsible for individual keys that are struck in a particular order at speed. As a result, there will always be at least a handful of combinations which "break" the efficiency or are not as comfortable as they could be, which is a necessary compromise under such a system. This particular issue of method is probably one of the reasons that, no matter how fluent the typist, you never seem to see typing scores that go very far beyond 212 WPM or so. Stenography and other chorded methods tackle this problem with a completely different paradigm and do achieve comfortable and fast results at the cost of approachablility. So what we really want in a keyboard layout is going to be maximal efficiency and comfort within the limits of the input method itself.
If you do attempt Mod-DH, which is worth it for the comfort benefits in my view, make sure that you do so on a layout which affects the smallest number of common keys if possible. On an ISO layout in a Wide configuration, the only real hurdle I have encountered for non-letters in day-to-day typing is the relocation of slash/question mark. The ANSI layout does not have the benefit of an additional key off to the right on the home row that can be used for the apostrophe, so moving things around further is needed. The ISO layout is the more comfortable of the two in this regard, in the experience of several users.
Finally, both rolls and alternation are addressing similar issues, and Colemak does employ both to achieve what it does. In fact, with alternation specifically, a case could be made that it does this a little better than Dvorak in giving the right hand only a little more to do, rather than favoring it more significantly. I appreciate alternation, but find that certain common sequences in a roll are quite essential for comfort once you have mastered them. Opinions vary widely on this point and I don't know that the advantage of rolls is as decisive as some people think, but they are ergonomic. Another point about Dvorak is, for all the flaws in its research, they did have the right idea in taking it straight into the "real world" of real people using it. This is the difference between efficacy and effectiveness as psychologists use the terms. A particular intervention or treatment may work well in a clinical setting (efficacy), but the results outside the lab are what everyone wants to see (effectiveness).
Long story short, I think the typist ultimately determines what will be most efficient and comfortable for them, though there are common factors for all of us that must be considered first. This may serve to highlight why ergonomics fall as much in the field of psychology as they do the hard sciences, and in some cases are purely psychology's domain (the study of ergonomics actually is a lesser-known specialization in industrial/organizational psychology, for those wondering).
Hopefully you found something useful in that lengthy exposition. Years of exposure and consideration have led to these conclusions, and I'm still learning.
Last edited by azuvix (16-Jun-2018 15:02:40)