Thank you, Shai! It is ever so nice of you to take the time to explain these things over and over again. As I've said before, I'm the kind of guy who needs to know and understand (scientist that I am, hehe). By explaining this, you make me gradually see all the strengths of the Colemak as well as the amount of thought and work that has gone into it. It's so intriguing to learn by doing.

On a side note: I suppose you'd like to write up a compendium of all these considerations at some point. Each time a stubborn n00b like myself comes along, we have to be shown and convinced all over again, and a thorough FAQ is the way to alleviate your workload. You've already got much here on your pages, but take note of the wrong turns each newcomer makes and you'll get a tip as to what could've been made even more clear and/or put more in our faces as we enter. The truth deserves to be out there... but not too far out!  ;)

Here's a layout I made that's meant to keep QWERTY positions intact:

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

- 15 keys are in their original positions: Q W P A S G H Z X C V M , . /  This is more than ASSET or Colemak.
- Only 4 keys change hands: E L F ;
- Same finger is better than ASSET
- Shortcuts Z, X, C, V, A, S, Q, W, P preserved (ctrl-P is seldom used, but it is still preserved)

I agree.

If home row goes ASRT, it is at cost of top row being QFWP. According to your arguments, it is better to keep W in place than to keep S and move W.

The old colemak was like this (Shai tried to keep Ctrl+S Save shortcut):

Merged into this thread.

By "popular" request ;-)   I analysed the layouts that were mentioned in this thread. As always, I used the AdnW-algorithm. Remarks:
- Totalscore (lower is better) is debatable, for it is based on AdnW's assumptions:
     - low distance is desirable (same as Colemak wants)
     - balance over fingers and hands (50/50) is desirable (same as Colemak wants) 
     - sameFinger % should be as low as possible (same as Colemak wants)
     - adjacent key use should be as low as possible (= no rolls), (Colemak wants the opposite, Colemak wants rolls)
     - alternation is good (no idea what Colemak wants, I'd say not as much as AdnW wants)

However, the next metrics are more or less objective:
     - Distance (lower is better), based on location of keys, Qwerty Y and B are the "farthest away" whereas Qwerty asdf jkl; are the closest. 
     - SameFinger %
     - Adjacent Keys %
     - Alternation
     - Finger use
     - Hand use

OK, here we go. Based on English text. Please ignore the symbols äöüß, they played to role in the analysis.  THE BLACK BLOCK IS SCROLLABLE. COLEMAK AND OLD COLEMAK ARE IN THERE
 

Asset            336.804 TotalScore  187.904 Distance              left right
                   3.098 SameFinger    9.189 Shift-SameFinger   ob  5.9 10.1
  qwjfg ypulüä    55.754 Alternation  41.114 Shift-Alternation  mi 38.4 31.6
  asetd hniorß     1.072 In/Outward    8.072 IndirKollision     un  6.8  7.2
  zxcvb km,.ö     22.632 adjacent     12.644 Shift-adjacent    sum 51.2 48.8
                  9.2  8.4 14.8 18.8 --.- --.- 18.4 10.8 12.1  7.6 Sh  1.1  1.7


AssetDreymaR     327.095 TotalScore  185.128 Distance              left right
                   2.509 SameFinger    9.734 Shift-SameFinger   ob  7.8  8.2
  qwfpg jluyüä    56.275 Alternation  40.442 Shift-Alternation  mi 38.4 31.6
  asetd hniorß     1.079 In/Outward    8.642 IndirKollision     un  6.8  7.3
  zxcvb km,.ö     22.752 adjacent     10.977 Shift-adjacent    sum 52.9 47.1
                  9.1  8.4 16.8 18.7 --.- --.- 18.8 10.8  9.9  7.7 Sh  1.1  1.7


SpeedMorph       310.240 TotalScore  186.765 Distance             left right
                   1.835 SameFinger    9.977 Shift-SameFinger   ob 10.9  7.5
  qwldb jfukpß    57.116 Alternation  41.537 Shift-Alternation  mi 30.8 37.3
  asrtg hneoiö     1.212 In/Outward    8.590 IndirKollision     un  5.4  8.1
  zxcvü ym,.ä     16.808 adjacent     10.539 Shift-adjacent    sum 47.2 52.8
                  9.2  8.4 12.9 16.7 --.- --.- 17.9 15.4  8.9 10.6 Sh  1.2  1.6


OldColemak       306.230 TotalScore  187.107 Distance             left right
                   1.372 SameFinger   14.424 Shift-SameFinger   ob 7.8  9.0
  qfwpg jluykü    58.143 Alternation  39.607 Shift-Alternation  mi32.7 37.3
  asrtd hneioß     1.058 In/Outward    8.725 IndirKollision     un 6.8  6.6
  zxcvb öm,.ä     16.502 adjacent      7.470 Shift-adjacent   sum 47.2 52.8
                  9.1  8.8 10.7 18.7 --.- --.- 18.1 15.4  9.8  9.5 Sh  1.1  1.7


Colemak          307.722 TotalScore  186.227 Distance              left right
                   1.348 SameFinger   14.374 Shift-SameFinger   ob  7.8  8.2
  qwfpg jluyäö    58.143 Alternation  39.607 Shift-Alternation  mi 32.7 37.3
  arstd hneioü     1.041 In/Outward    8.772 IndirKollision     un  6.8  7.3
  zxcvb km,.ß     17.492 adjacent      9.009 Shift-adjacent    sum 47.2 52.8
                  9.1  7.8 11.6 18.7 --.- --.- 18.8 15.4  9.8  8.7 Sh  1.1  1.7

Interpretation:
- finger & hand use is about the same in all layouts
- distance is about the same in all layouts
- Colemak and OldColemak do best on SameFinger
- The Asset layouts have a higher AdjacentKeys, meaning that words like in are typed with middle-index. in Colemak in is typed with ring-index. as is typed in Asset with pinky-ring; in Colemak it is typed with pinky-middle. Here you see the lower AdjacentKeys of the Colemak layouts.  (For the record: Dvorak and Dvorak-like layouts such as AdnW optimize for even much lower adjacent keys, in the order of around 9%, in contrast to Colemak's 17%. This goes at the expense of rolls of course) 
- Both Colemak-versions are virtually the same, I do not see any significant improvement over (new) Colemak over OldColemak.