I've started doing just this, thanks -- I'd forgotten about it.

Also, I've resolved to not check my speed stats daily, as I'd got into the habit of doing, because that tempts me to optimize for speed instead of deliberate practice and learning.  After two weeks I'll give myself a few typing tests as close as I can make them to how I've measured my speed in the past, and see how I'm doing.

I just discovered how much just the fact that I'm typing into Amphetype messes up my QWERTY.  So it's not only the keyboard I'm typing on that's a big cue for me!

Right now I'm typing this in QWERTY perfectly fluently with no errors.  Thirty seconds ago I was trying to type QWERTY in Amphetype (as a baseline against which to judge my Colemak) and totally failing to get much above 40 wpm -- sometimes more like 25 -- because of continual mixing up with Colemak key positions.

It amazes me how dramatic the difference is.

It doesn't worry me, I think it just shows up how, though the "switch" is in there somewhere, I don't yet have as good conscious control of it as I thought I did -- another thing that needs a bit of practice I suppose.  But as with typing QWERTY on my ergodox that I usually use just for Colemak, I'm not sure whether the effort is worthwhile: do I want to create layout interference for myself just to measure my QWERTY baseline speed in Amphetype?  Not sure.

I decided despite Dreymar's words of encouragement not to push even the QWERTY-in-Amphetype thing as a baseline measurement and instead just keep going with the practice.

I seem to have broken my recent stasis after concentrating on practice for a few weeks without any speed tests -- only a few day's measurements but I reckon I'm at about 55 wpm now.  This time the I did my tests strictly on first attempts on unseen text only -- so that's likely a bigger improvement than the numbers suggest: previously, I'd been using the average of all my runs on each text -- and it takes me a bunch of runs to get to 98% accuracy typically.

So I think the suggestion to use the Amphetype feature to generate practice texts using words that you stumbled on previously was pretty effective.

Right now I have a thing where I type D all the time when I want a G.  I feel like that's actually a sign of getting a bit more comfortable with the layout -- I think relaxing a bit may have made me stop putting in the effort to reach for that key :-)

Onwards towards 65 wpm!  I'll get there eventually!

60 wpm!

As usual by now, that's based on doing a few days of sessions in Amphetype where I go for speed rather than deliberate practice, then looking at the Performance tab -- the last few times I've I just been looking at the raw data for Texts and sort of doing a 'robust fit by eye' (i.e. ignoring the outliers).  When looking at this number I don't really try to look for improvements smaller than maybe 2.5 wpm -- you can still get some idea if you're getting better slowly by looking at the moving average, but that lags your current best wpm of course.  Total hours practice right now is roughly 57 (172 practice sessions of about 20 mins each).

Glad my feeling that I'm still making progress has finally shown up in the speed tests!  "Just" 5 wpm to go to hit my original target of 65.  I'm sure I'm going to hit that target now -- for a while at 50 and 55 I wondered if I was making enough real progress to get there, but of course progress does get slower as you get faster.

I also feel more fluent and have to concentrate less, though recently I've been making a lot of mistakes with R and S again.  I'm sure that actually starting to use it "for real" will inevitably throw me back a bit again -- for example I'll have a lot of new "words" to learn if I start doing coding -- lots of key shortcuts that I know very well with QWERTY and not at all with Colemak yet.  Also, my own day-to-day vocab I'm quite sure doesn't exactly match that of Marcel Proust or Joseph Conrad.

And, damn it Dreymar, I am tempted by mod DH.  Good reasons not to:

1. I don't feel enthusiastic to be patching spacemacs to support "KNEI" motion keys -- straightforward enough to start off with, but then I'm in the 0.001% of users who use that and maybe fighting a time consuming battle to keep it working with all the third-party code I end up using.  Spacemacs key bindings are complicated enough to deal with already!  On the other hand I'm probably almost in that position already with plain Colemak, and "K" and "H" are both already motion keys so maybe I won't hit too many extra problems that I won't already with "HNEI".
2. More vertical motion: avoiding that was the whole reason I started using Colemak in the first place.  That was mostly for middle and ring fingers though.
3. Re-learning.  But I think I might enjoy it.

Unfortunately, more persuasive: I think I *like* Colemak mod DH.  :-|

It's not "mod DH" anymore, nor is it Curl(DH) although I may keep using that name. Its name is Colemak-DH, or just "the DH mod".

The DH mod won't lead to more vertical motion I think? My index finger resting points are at the lower edge of the home keys, instead of their middle. So while the vertical stretch to B and J do become a bit longer, those are rare enough keys that I don't notice any real difference. Shai didn't like it from what I've heard, but I guess he didn't try it for long and went in with a bit of a bias. Nearly all the feedback I've heard has been praise saying how comfy it is, and I agree.

And it's quite easily learnt.

I haven't used Spacemacs and I don't know what it is, but that sounds like a fairly minor problem. Also, Extend usually fixes any such worries. Unless you're on Windows and Spacemacs is a program with a keyboard hook, in which case it'll conflict with EPKL.

Well, then... Spacemacs can run together with Extend and you can just use Extend-arrows instead of KNEI navigation and Bob's your uncle.

If your index fingers don't have a vertical motion problem, then the DH mod should be fine I hope. Let us know if you experience peculiarities!

62.5 wpm.

Same way of measuring as before more or less: average over 3 days, judged by eye in Amphetype performance tab looking at "<ALL TEXTS>", no damping, not trying to judge it any finer than in 2.5 wpm steps.

So close!  I thought I was going to hit 65 this weekend.  Tune in next month or three for the next exciting installment...

65 wpm!  Hit my average speed target at last.

10 months 12 days to get here.

Top speed in Amphetype in QWERTY before I started on Colemak was about 110 wpm, top speed in Amphetype so far with Colemak is about 87.5 wpm.  Not really a fair comparison: that underestimates my QWERTY skill because I didn't try that long compared with constant (albeit brief) practice with Colemak.

Ironically today's practice session was terrible, lots and lots of mistakes -- but not bad enough to mess up the average over the three days I measured, and even today not far below 65.

I do still feel quite clumsy with Colemak, but now I feel like I have enough speed to use it daily so I know it will improve from here.  I'm going to start with spare time projects because I rely so much on typing speed at work.  First thing is to get used to working in emacs and coding -- I've started that a little bit already in the past few weeks.

Thanks for the encouragement, especially DreymaR!

Also, for the record: I think my average over all practice sessions was probably about 20 minutes (it was more like 30 mins at the beginning), and I did 217 practice sessions up till today.  So that's about 72 hours practice in total.

Am I right that you put Q on your middle finger and F on the left index? I wouldn't do that... what about the FA/AF bigrams for instance? Putting C on a thumb would make sense, but what did you put on the old C position then?

I've been using only Colemak, more or less, since I last posted.  My QWERTY has rusted away -- for a while I told myself some time I'd spend a few days going back to QWERTY to prove to myself I can still do it, but I don't want to lose the painful days it would take to get back into it.  I can tell it would come back from experiments in that direction, but I'm sure it'd take at least a few days for it to stop being a disaster -- maybe a few weeks by now, I'm not sure.

I just tried typeracer for the first time in a long time, in a few tries won a race at 79 wpm.  I'm still definitely quite a bit slower than I was with QWERTY -- but that's never been my main goal and I've not really done any dedicated deliberate practice in a long while (I'll still hit words that annoy me and practice them a few times just in the course of normal typing -- but I did that with QWERTY too).

It's hard to tell how much if at all it's helped with my original goal of helping my fingers, but I think it has a little, and the proxy goal of reducing finger travel seems a sure thing.  I do still like the layout, though in a way it's less satisfying that QWERTY just because I'm still less accurate, fast and certain with Colemak now than I was with QWERTY.  Still happy that I made the switch though, and I still continue to slowly improve I think despite the lack of dedicated deliberate practice.