Asset compared to Colemak:

• Asset (old and new) has 200% more same-finger ratio compared Colemak. This is IMO the biggest flaw with Asset. This slows down typing, interrupts the typing flow and increases strain on the fingers.
• In the new Asset, the semicolon there does seem a bit stuck, and it's a waste to put the semicolon on the strongest finger. Exchanging it with M isn't a good idea either, because when there are three similar punctuation keys stuck together it's too easy to get confused between them.
• Asset (old and new) overworks the right pinky by keeping the P there, and having to hit the PR and RP digraphs with the weakest finger.
• The old Asset has 150% more same hand home-row jumping than Colemak. The new Asset has 40% more same hand home-row jumping than Colemak.
• The old Asset doesn't have the ZXCV in place.
• Asset (old and new) keeps the Y in its QWERTY place. The somewhat common "my" digraph is very uncomfortable to type.
• The new Asset puts J, K and semicolon which are rare keys on strong positions, and puts L, P, Y and semicolon on weaker positions.
• Colemak moves 17 keys, new Asset moves 13 keys and old Asset moves 18 keys. Asset has a few things in common with Colemak, like having the same keys on the home row, and replacing the Caps Lock with Backspace.

I personally think that the new Asset is better than the old one.

Is there some open-source code available that one can use to calculate these statistics? Are you aware of a decent theoretical model that can predict touch-typing speed? I'm working on a model myself but it'll take awhile before it's ready, and making the model accurate would require measurements of real touch typists, and preferably non-Qwerty touch-typists actually. They're kind of hard to find…

The key idea of Asset with respect to other layouts is higher Qwerty similarity. I'm open to suggestions of how to increase theoretical efficiency without sacrificing that feature.

By the way, it's looking like the right side of the keyboard is the biggest trouble area in Asset. Rearranging four keys again (;UPK) yields this:

Q W J F G Y P U L ;
A S E T D H N I O R
Z X C V B K M , .

This arrangement has the same bottom row as Colemak, still has 12 letters that match Qwerty, and still keeps all letters on the same hand except R and J. It does not fix Qwerty's Y / MY flaw, but by moving U one space to the right, a few digraph contentions are removed, notably UN/NU (IU/UI is much less common). The strain on the pinky should be less with ";" in place of P, except for programmers, most of whom use ";" frequently (poor bastards.)

Regarding MY, I would suggest a trick which I use on Qwerty, namely, to hit Y with the left hand in this case. I would be interested to know if this trick is widely used by fast typists (and other tricks, for that matter. e.g. when I type ED, I hit E with the middle finger and D with the index finger. My whole bag of tricks gets me a speed of up to 70wpm, but it took years to aquire.)

That layout is better than the previous versions, but it now resembles Colemak more than it resembles the original (2004) Asset. The new version moves 15 keys while Colemak moves 17, so it is marginally easier to learn compared to Colemak, but it still has some of the issues of the last version.

• 100% more same-finger ratio compared Colemak.
• LO/OL (very common) same-finger digraph on slow finger.
• MP/PM digraph is uncomfortable to type.
• 125% more same hand home-row jumping than Colemak.
• Puts J is in a strong position, and Y is in a weak position. MY/YM digraph is uncomfortable to type. Hitting Y with the left hand is a bad idea, because it's a very long stretch and it can cause a lot of strain.

Shai, what is "same hand home row jumping"? My data shows that PM/MP are rare digraphs (0.09% of all digraphs in English), so I won't worry about them, but OL/LO are significant (0.47%). I looked at the OL/LO problem for a couple hours, but found that the only way to fix it (without creating new problems) required Y to move (along with U/J/P/K)—which is not bad, except that it would mean only 11 letters match Qwerty, which in turn makes it more similar to Colemak, which keeps 10 letters.

I'm glad you like the lessons. Feel free to use the code on your own website if you like.

So I guess what it comes down to is how much Qwerty similarity is desired. Asset keeps 12 letters while Colemak keeps 10; and in total, my Nov. 13 redesign also keeps 20/26 letters on the same finger, whereas Colemak only keeps 15/26. Unfortunately, how important it is to keep keys on the same finger is purely a matter of opinion, because it would probably take a quite expensive study to tell for sure.

Now, Colemak and Asset have many similarities, the only major differences I can see being that (1) Asset is more similar to Qwerty, (2) Asset's Y key could be better placed, and (3) Asset has higher same-finger typing (most significantly, LO, OL and MY).

I'd like to address the claim of "100% more same-finger ratio compared Colemak." That could be true, but I tried a corpus of several books at http://colemak.com/Compare and found that Colemak usually has under 2% same-finger typing, which makes under 4% for Asset. Now, if we assume digraphs typed with the same finger take twice as long to type as digraphs on different fingers, we would conclude that a Colemak typist would be able to type about 2% faster than an Asset typist.

So that's the trade-off. More similarity to Qwerty or 2% more speed. I guess we'll let the community decide. In the meantime I guess I should resubmit the new Asset…

It would be kind of nice if we could come to a compromise. For example: ARST. If you switched R and S then S would be back in its Qwerty position, and the Ctrl+S (Save) shortcut would be preserved. My digraph data indicates that if you switched R and S, the amount of same-finger typing would increase by 0.22% (i.e. 1.92% instead of 1.70%). I guess you wouldn't think it worthwhile, but tell you what… if you switch them, I'll drop out of this competition and replace my Asset page with a page that praises Colemak :)

Before I made Colemak, I developed a keyboard layout that was called Asetion:

12345 67890-=
qwfdg ymklp[]
asetr huion'
zxcvb j;,./

Asetion moved 15 keys (like the Nov. 13 Asset), and only moved 4 keys to a different finger (vs. 6 keys the Nov. 13 Asset), moved no keys from hand to hand (vs. 2 in the Nov. 13 Asset). But like the current Asset it suffered from some issues, mainly a higher same-finger ratio.

After I've created Asetion, I got some feedback from people who tried it out, I've made a lot of research on the subject and I've discovered some mistakes that I made in Asetion's design. During my research I also found out about Asset, and implemented some of the ideas in Colemak, such as the keys in the home row. Colemak is the result of this research.

The first question that everyone asks before learning a keyboard layout is: "is it better than Dvorak?". When people see that it loses on one criterion, such as same-finger ratio, they lose interest. So I think it's very important for a keyboard layout to be better than Dvorak on all the important criteria.

Read The Dvorak Simplified Keyboard: Forty Years of Frustration. The article compares Dvorak and QWERTY on several criteria:

• Hand overload/hand alternation. This is a controversial issue, but I found that there's a sweet spot for hand alternation. When hand alternation is very low (like QWERTY/Asset/Asetion) you end up typing very long word with one hand, which is obviously not a good thing. If you make hand alternation too high, you cause staccato typing like in Dvorak (while some like it, I and others don't). See http://colemak.com/Hand_alternation
• Unbalanced finger load. One of the biggest complaints about Dvorak is that it puts the L and S on the right pinky. As a simple exercise try typing repeatedly on QZQZQZ (QWERTY) with UMUMUM (QWERTY). You'll instantly notice the difference in the strength and dexterity of the fingers. In Asset/Asetion, this isn't taken fully into account, but it is a very important factor, especially on the weak fingers (pinky and ring fingers).
• Excess finger movement/finger distance. There's no doubt that cutting finger distance is a very good thing.
• Awkward strokes (same hand row-jumping). There's a good explanation of the issue in the article linked above. While this doesn't really affect speed, it is a very important issue for ergonomics. When you need to hit keys with the same hand, you place one finger on the bottom row and you prepare the next finger to type on the upper row. Frequently the posture of the hand in these situations can cause strain. You could avoid the situation by waiting that the previous finger will completely return to the home position, but then you lose on typing speed.
• Same finger ratio (not mentioned in the article, but is very important nevertheless). I've found that the best way to achieve high typing speed with a very low error rate, is to maintain a steady typing rhythm. On Asset/Asetion you'll lose the typing rhythm twice as often, which is very notable. Moreover all the other typing problems are multiplied when they happen on the same finger. The undisputed criteria that affect speed are finger distance and same-finger ratio. Again, if people see that the keyboard layout is worse than Dvorak on this aspect, they won't be interested. You must take into account that people might need to type in other languages where the same digraphs are much more common. e.g. in Spanish LO/OL digraph is about 200-300% more frequent than in English. Colemak is designed so that the sum of any single digraph and its reverse digraph exceed 0.10%.

Actually in the first version of Colemak, the R and the S were switched back as you suggest in order to preserve the Ctrl+S shortcut, but then I switched it back, because I found the same-finger issue annoying. This change alone causes about 30% more same-finger ratio. It might not seem like a big difference in absolute numbers, but I definitely notice the difference.

I found from my own experience that while keeping the keys on the same hand is important for easy learning, but keeping them on the same finger isn't. Colemak completely avoids cyclic replacements of keys, e.g. exchanging L and O that happens on Asset, or exchanging T and Y that happens on Dvorak. To type O in Asset, you type on the L QWERTY position. To type L on QWERTY you type on the O Asset position. This is very easy for your mind to mix up, especially if you're switching back and forth between layouts. I know that even after using Dvorak for over a year, I've made mistakes from time to time with the T and Y. The same class of errors were also common when I used Asetion.

A keyboard layout is something that you don't want to learn more than once in a lifetime. So while Colemak does try to be easy to learn ( http://colemak.com/Easy_to_learn ), it's important not compromise on the other issues. Remember that Colemak is easier to learn than the original Asset from 2004, which I wouldn't consider too difficult to learn either.

Colemak has already hundreds of happy users, and I promised that the layout would not change again. I don't see the suggested change as something positive. Not to mention that for every small change in the layout requires dozens of hours to update everything (the images, the typing lessons, dozens of different implementations, Vim, the Java comparison applet, etc.) in about 50 different places.

I would really appreciate your blessing, because if we are to defeat QWERTY we should unite forces.

All right, you've made a good case, except that the 30% figure seems too high. Where does your data come from? Mine (which is based on an analysis of 375MB of English books and short stories) suggests that if you were to switch R and S, the same-finger would increase approximately 10%, and you would still beat Dvorak on the same-finger statistic by at least 20%. Maybe it's different in Spanish?

I have a written a small program to calculate the statistics, I might release it in the future after I clean up the code. In the meanwhile I recommend using the Java comparison applet in http://colemak.com/Compare . I didn't modify any of the original applet code to calculate those statistics, but I've verified the finger distance and same-finger statistics there to be 100% accurate. One possible difference is that repeated characters are not counted as same finger. For any calculation you should take the figure there as authoritative.

There also some other minor reasons for switching the R and S.

• Better finger balance, doesn't overload the ring finger.
• Less same hand home row jumping.
• More comfortable hand rolling movements.
• A more logical and easier to remember order (QRST).
• Fits slightly better with the multilingual layout.

These are all small factors, but together I think they make for a better keyboard layout.

Correction: this doesn't affect same hand home row jumping.