Author Topic: Introducing the anishtro letter layout  (Read 4898 times)

0 Members and 1 Guest are viewing this topic.

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Introducing the anishtro letter layout
« on: Sun, 16 August 2015, 15:27:15 »
In the beginning of 2013 I learned to touch type in QWERTY on a standard keyboard.

Mid-2013 I had a Truly Ergonomic keyboard.

A few months later I started thinking about learning a new keyboard layout.

In the Spring of 2014 I had created my own. I’ve used it since and never looked back. It was a lot of fun to do, and I like it, but I don’t expect it to be popular or anything.

Today I finished a little write-up on it, and decided to share it :)

Code: [Select]
qlucj kpmw·
anisv bhtro
··ygx zfd··

   e   ·

You can read all about it here: https://github.com/lydell/anishtro

That link takes you to:

  • A keyboard layout (actually two!) you might find interesting :)
  • Links to tools to gather and analyze text to use as data when creating a layout, as well as the 13MB text I used.
  • My experience on choosing between standard and ergonomic keyboards and layouts.
  • My experience on how to create a layout, as well as generalized advice on how to create one.
  • A bit of my personal evolution from not interested at all to a keyboard geek.

Enjoy!

Offline stevep

  • Posts: 25
    • Colemak Mod-DH
Re: Introducing the anishtro letter layout
« Reply #1 on: Mon, 17 August 2015, 05:54:01 »
Interesting contribution, it looks quite nice given the assumptions it makes. I largely agree with the key ratings, especially the fact the centre-column keys are avoided. I don't think the top-row ring finger is so great though, but then, Workman also rates that highly for some reason.

The big sticking point for me would be any layout with the E on the thumb. Even with the most perfectly designed ergonomic keyboard with comfortable thumb keys, I would always want letters in the main section, and use thumb keys for other useful purposes, e.g. Shift, other modifiers, or a custom extend layer.

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #2 on: Mon, 17 August 2015, 14:59:27 »
Interesting contribution, it looks quite nice given the assumptions it makes. I largely agree with the key ratings, especially the fact the centre-column keys are avoided.

Thank you!

I don't think the top-row ring finger is so great though, but then, Workman also rates that highly for some reason.

That “some reason” is simple: People are different! What is comfortable is different to different people. Take beds for example. Good thing there are both many keyboard layouts and many beds to choose between!

The big sticking point for me would be any layout with the E on the thumb. Even with the most perfectly designed ergonomic keyboard with comfortable thumb keys, I would always want letters in the main section, and use thumb keys for other useful purposes, e.g. Shift, other modifiers, or a custom extend layer.

If you dislike having a letter on a thumb I respect that. But if you haven’t actually tried it I recommend you do. I used to think that I didn’t want a letter on a thumb either, but after having tried it I was amazed at how much I liked it! In fact, if I could do just one single improvement to QWERTY, it’d be moving the E to a thumb key (provided that you have a keyboard with more than one main key per thumb). You can read all about it here: https://github.com/lydell/anishtro/blob/master/how.md#the-one-single-change-to-qwerty

Offline Coreda

  • Posts: 700
Re: Introducing the anishtro letter layout
« Reply #3 on: Mon, 17 August 2015, 15:12:14 »
Really interesting writeup and data analysis. Reads like something I'd find on Hacker News. Enjoyed the programmatic steps arriving to the perfect layout for your needs.
« Last Edit: Mon, 17 August 2015, 15:13:51 by Coreda »

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #4 on: Mon, 17 August 2015, 15:52:18 »
Thanks for the kind words, Coreda! Given the time and energy (while still being lots of fun) it took to create the layout (and that writeup!) that feels really good to hear  :)

Offline jacobolus

  • Posts: 3642
  • Location: San Francisco, CA
Re: Introducing the anishtro letter layout
« Reply #5 on: Mon, 17 August 2015, 19:06:46 »
It’s nice that people keep trying to make new layouts. Definitely keep experimenting!

I’d love to hear your thoughts about what works and what doesn’t work about your layout. Are there any phrases that you find yourself making an unusual number of errors on? Any particular sequences that you find are now much faster or more comfortable compared to whatever layout you were using before?

Just at a brief glance, certain common combinations seem like they’d be pretty awkward to type. For example, THR, THO, ROT, HOR, etc. occur in many words, but this layout forces them to have direction-reversing sequences on the same hand, which are slow and error-prone. Having PH on the same finger is going to really be a pain. Even just swapping T and H seems like it would make a substantial improvement. Etc. Etc.

Side note: I really wish there were better data available showing precise keystroke timing of various skilled typists on various keyboard layouts (QWERTY, Dvorak, Malt, Colemak, whatever...), including a detailed account (perhaps synced with high-frame-rate top-view video) of precisely which finger they used to press each key. I’d love to see better analysis of which digraphs and trigraphs lead most often to errors, which sequences are fastest or slowest, and have some corpus of video evidence about hand motions to look more concretely at specific examples. Most of the heuristics people use to design keyboard layouts (to take a particularly egregious example, look at carpalx) are basically speculative fantasy, and don’t even have solid reasoning behind them, much less any empirical evidence. I honestly don’t understand what ergonomics researchers have been doing with their grants; all the existing academic research on this and other topics is so so lame.
« Last Edit: Mon, 17 August 2015, 19:09:19 by jacobolus »

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #6 on: Mon, 17 August 2015, 21:29:48 »
Whenever I see a thread like this, I have to fight the urge to post my own. :) I find them very interesting.

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #7 on: Tue, 18 August 2015, 14:50:28 »
I’d love to hear your thoughts about what works and what doesn’t work about your layout. Are there any phrases that you find yourself making an unusual number of errors on? Any particular sequences that you find are now much faster or more comfortable compared to whatever layout you were using before?

One of my favourite words are “this,” which is written by rolling the right ring finger to the index finger (an inwards roll) on the home row, and then doing the same thing with the left hand. A similarly typed word is “than.” “There” is also one of my favourites—really, any combination with E (being on the left thumb) is really nice.

One thing I notice myself having a little bit of problem with sometimes are words with lots of A:s and N:s in a row, such as “bananas.” I do like rolling between the pinky and ring finger (in either direction), but that repetition is hard. Spamming “anananan” is a lot harder than “isisisisis” (the latter is also a roll on the same hand, but between the middle and index finger).

A bit ironically, I write a lot of JavaScript code. The file extension for JavaScript is “.js”, which is a same finger usage in my layout. Not a big deal, but quite funny. Also, in JavaScript (CommonJS to be precise) I type the word “require” a lot, which also contains a same finger usage: UI.

What I like best about my layout is that it feels like my fingers stay on the home row, and just take an occasional departure to a different row per word.

Just at a brief glance, certain common combinations seem like they’d be pretty awkward to type. For example, THR, THO, ROT, HOR, etc. occur in many words, but this layout forces them to have direction-reversing sequences on the same hand, which are slow and error-prone.

Actually, “through” and “thought” are two of my favourite words :) “prototype” and “short” don’t bother me either. I like rolling, even with more than two fingers, even when the same finger is part of two consequent rolls in two different directions (such as the middle finger when typing “through”)! I don’t think you can say that “direction-reversing sequences on the same hand” are “slow and error-prone” in general. I don’t think so. It’s probably different from person to person.

Having PH on the same finger is going to really be a pain.

Well spotted! PH is indeed one of the worst same finger usages of the layout (0.11%). Only CS (0.23%) and IU (0.14%) are worse. I actually noted that PH wasn’t a particularly good combination while designing the layout, and initially tried to avoid it. I was surprised that doing that sacrifice resulted in lower same finger usage overall!

Even just swapping T and H seems like it would make a substantial improvement.

Not according to me. My analyzer reports 0.1 percentage points more same finger usage and 0.08 percentage points less rolls in that case. I guess you did the change the get rid of PH? Well, instead you introduce FT (0.11%) and PT (0.10%). I tried to put the least common of the home keys on the index fingers, since that spot is exposed to the largest number of other keys—the whole center columns need to be taken into account as well. T is a lot more common than H.

Etc. Etc.

Huh?

Most of the heuristics people use to design keyboard layouts (to take a particularly egregious example, look at carpalx) are basically speculative fantasy, and don’t even have solid reasoning behind them, much less any empirical evidence

I agree with you regarding carpalx. However, at least in my case there’s no speculative fantasy and I have solid reasoning behind my choices! :)

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #8 on: Tue, 18 August 2015, 14:51:13 »
Whenever I see a thread like this, I have to fight the urge to post my own. :) I find them very interesting.

No, please do post your own! I’d (just like you) find it really interesting!

Offline jacobolus

  • Posts: 3642
  • Location: San Francisco, CA
Re: Introducing the anishtro letter layout
« Reply #9 on: Tue, 18 August 2015, 15:09:05 »
I like rolling, even with more than two fingers, even when the same finger is part of two consequent rolls in two different directions (such as the middle finger when typing “through”)! I don’t think you can say that “direction-reversing sequences on the same hand” are “slow and error-prone” in general. I don’t think so. It’s probably different from person to person.
“Rolling” is a combination like 4-3-2 or 5-4-2, etc. It’s not really “rolling” when you have a 4-2-5 or 5-3-4 or whatever combination. (Where fingers are labeled from thumb = 1 to pinky = 5) Instead, those direction reversals are much slower and more error prone than either alternating to the other hand or keeping the motion going in the same direction. Also bad are any combinations of length ≥3 which need to switch back and forth between two fingers among 3-4-5, e.g. 3-4-3-4 etc., since those fingers rely on shared muscles to flex, especially combinations where you need to move the fingers to new keys in between. The index finger is more independent, so something like 2-3-2-3 or 2-5-2-5 is fine, and isn’t even too bad when the index finger needs to handle multiple keys. To take examples from your layout, HOBO is pretty easy, but MOTOR and ROBOT really suck.

Quote
Even just swapping T and H seems like it would make a substantial improvement.

Not according to me. My analyzer reports 0.1 percentage points more same finger usage and 0.08 percentage points less rolls in that case. I guess you did the change the get rid of PH?
Swapping T and H mostly fixes some of the awkward home row combinations; getting rid of PH is just a nice side benefit. I suspect moving one of the right hand home row keys to the other hand would also help, but it would take a much more complete analysis.

Quote
I tried to put the least common of the home keys on the index fingers, since that spot is exposed to the largest number of other keys—the whole center columns need to be taken into account as well. T is a lot more common than H.
The index finger is the strongest and most independently flexible of the four fingers, by far.

Quote
Etc. Etc.
Huh?
My point was, if someone examined this layout for more than like 30 seconds, I’m sure there are many other similar improvements that can be made. I only briefly glanced at the home row of the right hand.
« Last Edit: Tue, 18 August 2015, 15:12:48 by jacobolus »

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #10 on: Tue, 18 August 2015, 15:20:45 »
Well jacobolus, you and I obviously have very different opinions when it comes to keyboard layouts :) Good thing we’re both free to choose for ourselves what to use. I had a lot of fun making it, and I enjoy typing on it (which I’ve done the last one and a half years), and those two where my only goals. Thanks for sharing your thoughts, it’s interesting!  :thumb:

Offline jacobolus

  • Posts: 3642
  • Location: San Francisco, CA
Re: Introducing the anishtro letter layout
« Reply #11 on: Tue, 18 August 2015, 16:11:59 »
Sure thing!

Sorry, I’m not trying to sound negative or discouraging, and I’m glad you came up with something that works for you. As I said before, it’s great if people experiment and report back findings.

I really wish there were some more solid data from ergonomics researchers, as it would be nice to have a more empirical basis for theoretical models of efficient combinations of finger motions. My analysis is based on first-hand experiments, some anatomical understanding of the muscles in the hand, construction of several prototype keyboards, and a whole lot of making weird hand motions while sitting in public places and getting funny looks for it. :) Hard data would be much better.

Simple analysis of layouts like “X% same finger use” or “X% alternation” or “X% rolling” aren’t in my opinion very valuable metrics per se, but only rough pointers for the most basic comparison. Typing isn’t done letter by letter or as digraphs, but rather is chunked up into common learned motions for words or parts of words, which get repeatedly practiced in context. A single motion sequence on one hand can be continued across a keypress by the other hand, or motions by the two hands can be interleaved. As such, it’s necessary to look behind by at least 3–4 keypresses to analyze any given keypress. It’s important to figure out where the hand and fingers are likely to be in context and what direction they’re moving, which arm/hand/finger muscles were recently active and how long before, etc. Additionally, different people have dramatically different typing technique, using different fingers for different keys, different amounts of arm movement, different muscles for flexing individual fingers, fingers in different resting positions, and so on. Some of the fastest typists change which finger they use for each key depending on the context. None of these subtleties are captured by any set of design heuristics or actualized keyboard layouts I’ve ever seen, and I think a big part of that is lack of good data.
« Last Edit: Tue, 18 August 2015, 16:14:09 by jacobolus »

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #12 on: Tue, 18 August 2015, 16:26:13 »
Just out of curiosity, which layout do you use?

Offline jacobolus

  • Posts: 3642
  • Location: San Francisco, CA
Re: Introducing the anishtro letter layout
« Reply #13 on: Tue, 18 August 2015, 20:30:58 »
I’ve spent considerable amounts of time trying Dvorak, Colemak, a Malt variant, an Adnw variant, and a couple of my own layouts to get a feel for them, but for routine typing I currently just use QWERTY, about 60% of the time on a scissor-switch laptop keyboard :).

My plan was to build my own keyboard(s) and then figure out the right logical layout after the physical layout was firmly established. But that project has been on hold for the last year or so, while I work on other stuff. (The last few months, mostly making maps and reading math books.)
« Last Edit: Tue, 18 August 2015, 20:33:11 by jacobolus »

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #14 on: Tue, 18 August 2015, 23:06:28 »
Whenever I see a thread like this, I have to fight the urge to post my own. :) I find them very interesting.

No, please do post your own! I’d (just like you) find it really interesting!

I want to try this one out...
108682-0
At least, once I have enough time to set up either the Maltron or the Kinesis I bought. :) Granted, the above is for the Ergodox, but most of the keys are going to be where I want them on the keyboards I have, just the ones on the edge need moved a bit.

In case anyone is wondering, I played around with MTGAP software for a couple of months last year (really heavily, too -- I ran several dozen quick tests, tweaking settings, plus quite a few overnight runs after the tweaks looked reasonable). I also tried little variations of the MTGAP results with the Keyboard Layout Analyzer on patorjk.com, and tried to get a feel with the keyboard itself.

I probably will hate some part of it and change it, but most of the unusual things I did on the left side (like moving the OR off the home row) I did for inward rolls, and because it resulted in relatively little use of the same finger twice in a row. Things like you and your bounce a bit back and forth, but all of variations I tried had quirks with common words. The right side is pretty clearly heavily influenced by MTGAP.

Not sure if I can program all the keys how I want them to be, especially splitting punctuation, but I'm going to try. I already have Soarer's converter for the Maltron.
« Last Edit: Tue, 18 August 2015, 23:08:49 by Snarfangel »

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #15 on: Wed, 19 August 2015, 13:28:28 »
Interesting layout, Snarfangel! It actually performs pretty well in my analyzer:

Code: [Select]
{ sameFinger:
   [ [ 'nr', 0.24390161897603993 ],
     [ 'pt', 0.09824084615667118 ],
     [ 'ao', 0.09609421189814711 ],
     [ 'lm', 0.061256665558003354 ],
     [ 'gs', 0.059433319591425664 ],
     [ 'kl', 0.040798982443635334 ],
     [ 'kl', 0.040798982443635334 ],
     [ 'lv', 0.03566516436752655 ],
     [ 'iy', 0.032070198561082605 ],
     [ 'fs', 0.025125966050073993 ],
     [ 'hl', 0.023975059730744822 ],
     [ 'hk', 0.020393025455978736 ],
     [ 'hk', 0.020393025455978736 ],
     [ 'hm', 0.01853088489436749 ],
     [ 'km', 0.01761274614523972 ],
     [ 'km', 0.01761274614523972 ],
     [ 'cx', 0.017354115511682606 ],
     [ 'bc', 0.01229788662564095 ],
     [ 'dt', 0.011405610939868895 ],
     [ 'dp', 0.007784782070069245 ],
     [ 'cz', 0.0061683406103372595 ],
     [ 'cq', 0.005780394660001583 ],
     [ 'bw', 0.005366585646310195 ],
     [ 'cw', 0.0038148018449674875 ],
     [ 'fg', 0.0035432396797325134 ],
     [ 'hv', 0.0022242234485912132 ],
     [ 'bz', 0.001875072093289104 ],
     [ 'mv', 0.001655236054765554 ],
     [ 'kv', 0.0015647153330205627 ],
     [ 'kv', 0.0015647153330205627 ],
     [ 'bx', 0.0009569333441613358 ],
     [ 'wz', 0.0008017549640270651 ],
     [ 'wx', 0.0007888234323492093 ],
     [ 'xz', 0.0003362198236242531 ],
     [ 'qx', 0.0003103567602685414 ],
     [ 'bq', 0.00028449369691282956 ],
     [ 'qz', 0.00018104144348998245 ],
     [ 'qw', 0.00016810991181212656 ] ],
  rolls:
   [ [ 'ht', 3.1143912946997787 ],
     [ 'an', 2.7104361081469164 ],
     [ 'or', 2.2012958429263745 ],
     [ 'st', 1.5176186946498116 ],
     [ 'ai', 0.7833404629177985 ],
     [ 'lp', 0.2623419831486624 ],
     [ 'cs', 0.23447453238288302 ],
     [ 'iu', 0.13753977092567524 ],
     [ 'oy', 0.10984043007170793 ],
     [ 'dv', 0.01916452994658243 ],
     [ 'df', 0.009621059568324783 ] ],
  sameFingerTotal: 0.9381308971017337,
  rollsTotal: 11.100064709384517 }

It’s interesting to see how you’ve chosen such prominent spots for many punctuation keys on the left, and that you’ve put way more letters on the right hand. Are you a programmer? If not your layout looks really good for programming when thinking about punctuation characters. Myself I’ve got most such characters in another layout around the home row.

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #16 on: Wed, 19 August 2015, 22:12:15 »
Interesting layout, Snarfangel! It actually performs pretty well in my analyzer:

Code: [Select]
{ sameFinger:
   [ [ 'nr', 0.24390161897603993 ],
     [ 'pt', 0.09824084615667118 ],
     [ 'ao', 0.09609421189814711 ],
     [ 'lm', 0.061256665558003354 ],
     [ 'gs', 0.059433319591425664 ],
     [ 'kl', 0.040798982443635334 ],
     [ 'kl', 0.040798982443635334 ],
     [ 'lv', 0.03566516436752655 ],
     [ 'iy', 0.032070198561082605 ],
     [ 'fs', 0.025125966050073993 ],
     [ 'hl', 0.023975059730744822 ],
     [ 'hk', 0.020393025455978736 ],
     [ 'hk', 0.020393025455978736 ],
     [ 'hm', 0.01853088489436749 ],
     [ 'km', 0.01761274614523972 ],
     [ 'km', 0.01761274614523972 ],
     [ 'cx', 0.017354115511682606 ],
     [ 'bc', 0.01229788662564095 ],
     [ 'dt', 0.011405610939868895 ],
     [ 'dp', 0.007784782070069245 ],
     [ 'cz', 0.0061683406103372595 ],
     [ 'cq', 0.005780394660001583 ],
     [ 'bw', 0.005366585646310195 ],
     [ 'cw', 0.0038148018449674875 ],
     [ 'fg', 0.0035432396797325134 ],
     [ 'hv', 0.0022242234485912132 ],
     [ 'bz', 0.001875072093289104 ],
     [ 'mv', 0.001655236054765554 ],
     [ 'kv', 0.0015647153330205627 ],
     [ 'kv', 0.0015647153330205627 ],
     [ 'bx', 0.0009569333441613358 ],
     [ 'wz', 0.0008017549640270651 ],
     [ 'wx', 0.0007888234323492093 ],
     [ 'xz', 0.0003362198236242531 ],
     [ 'qx', 0.0003103567602685414 ],
     [ 'bq', 0.00028449369691282956 ],
     [ 'qz', 0.00018104144348998245 ],
     [ 'qw', 0.00016810991181212656 ] ],
  rolls:
   [ [ 'ht', 3.1143912946997787 ],
     [ 'an', 2.7104361081469164 ],
     [ 'or', 2.2012958429263745 ],
     [ 'st', 1.5176186946498116 ],
     [ 'ai', 0.7833404629177985 ],
     [ 'lp', 0.2623419831486624 ],
     [ 'cs', 0.23447453238288302 ],
     [ 'iu', 0.13753977092567524 ],
     [ 'oy', 0.10984043007170793 ],
     [ 'dv', 0.01916452994658243 ],
     [ 'df', 0.009621059568324783 ] ],
  sameFingerTotal: 0.9381308971017337,
  rollsTotal: 11.100064709384517 }

It’s interesting to see how you’ve chosen such prominent spots for many punctuation keys on the left, and that you’ve put way more letters on the right hand. Are you a programmer? If not your layout looks really good for programming when thinking about punctuation characters. Myself I’ve got most such characters in another layout around the home row.


I'm not a programmer, though the guy who came up with MTGAP (Michael D -- Geekhack likes to censor his name) is. :)  So probably it carried over a bit from him.

The vowels tended to fall naturally on one side or another in the optimization program, which kind of makes me think Dvorak had a few good ideas (hehe). Since the program also tried to balance hand usage, and vowels tend to be more common than consonants (ETAOIN SRHLDCU and all that), that meant a lot of consonants tended to be put on the other side. Also, punctuation tends to be relatively rarer after vowels than consonants -- it's relatively uncommon to end words with vowels other than E (the order of last letters is E, S, T, D, N, R, Y, F, L, O, G, H, A, K, M, P, U, W or something like that) -- so putting them on the side with vowels seems a safe bet. The N and the R are something of a problem, but commas and periods tend to be the great majority (combined they total about 60% of all punctuation), with hyphens, double quote, and single quote/apostrophe another 30+%. Exclamation points, question marks, and the rest I stuck around NR are relatively few in number. I kind of wish comma and period could be next to each other because it would be easier to track them down that way when learning the keyboard, but hiding the period under the ring finger isn't as good IMO as putting it where Q normally is.

Anyway, that's a brief overview of what I did. :) Have you ran your keyboard layout through the patorjk.com site? I might try it if I have a chance to see how mine compares (I think MTGAP is a lot more accurate, but the ease of use of the patorjk keyboard layout analyzer is awesome, so I play with that a fair amount).

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #17 on: Sat, 22 August 2015, 03:20:45 »
The vowels tended to fall naturally on one side or another in the optimization program, which kind of makes me think Dvorak had a few good ideas (hehe). Since the program also tried to balance hand usage, and vowels tend to be more common than consonants (ETAOIN SRHLDCU and all that), that meant a lot of consonants tended to be put on the other side.

It’s pretty interesting that I didn’t optimze for hand alternation and didn’t care about “hand balancing,” but still ended up with all vowels except O on one side. (Though it is possible to move E to the other hand’s thumb.)

it's relatively uncommon to end words with vowels other than E (the order of last letters is E, S, T, D, N, R, Y, F, L, O, G, H, A, K, M, P, U, W or something like that)

When reading that statement, I just had to compare with my own data:

Code: [Select]
[ [ 'e', 17.91392048902396 ],
  [ 's', 13.349383147200076 ],
  [ 'n', 10.760706687174762 ],
  [ 'd', 9.91189940135185 ],
  [ 't', 7.21523345296389 ],
  [ 'r', 6.164119254079948 ],
  [ 'y', 5.569186338834357 ],
  [ 'a', 4.241262387211667 ],
  [ 'f', 3.9351270486423293 ],
  [ 'o', 3.8960923161516114 ],
  [ 'l', 3.6804105582016287 ],
  [ 'g', 2.9538592619562447 ],
  [ 'h', 2.9326576559841286 ],
  [ 'm', 1.9272160755728271 ],
  [ 'k', 1.1016413906866496 ],
  [ 'i', 0.890616060216524 ],
  [ 'c', 0.855791927042698 ],
  [ 'w', 0.7874811451839412 ],
  [ 'p', 0.73591368766765 ],
  [ 'u', 0.3041043436047189 ],
  [ 'b', 0.2586298709822833 ],
  [ 'x', 0.25308178717649593 ],
  [ 'v', 0.14043587133399218 ],
  [ 'z', 0.12849758385903903 ],
  [ 'j', 0.07207555301268381 ],
  [ 'q', 0.020656704884047528 ] ]

Our lists are very similar, except that in my list, ending A is much more common than in yours. Here are the 100 most common words ending with A in my data:

Quote
a area la california australia canada america india data china sea media virginia pennsylvania via africa florida russia usa maria santa sega era drama opera idea indiana victoria carolina ma asia columbia georgia da philadelphia austria orchestra arizona oklahoma encyclopedia formula arena argentina korea mega minnesota saavedra louisiana extra na ka peninsula barbara iowa villa ca bulgaria macedonia nebraska anna fifa indonesia propaganda olympia fa manga bibra junta ha malaysia sa cuba silesia alaska nova uefa nevada gmina vienna camera atlanta ra ya chikara taxila viola ncaa romania rudra tea aka alabama faa listeria sierra wikipedia dakota pa wikimedia costa fda

Looks like there are loads of US states and countries in there. Not sure I type them so often. Might be affected from that all my text comes from Wikipedia articles.

so putting them [the punctuation characters] on the side with vowels seems a safe bet.

I thought a bit about it in the beginning if I should measure same finger usage involving punctuation characters, but ultimately decided that wasn’t necessary. Thinking about it, I’ve never been so bothered by a same finger usage involving a punctuation character that I’ve noticed it. But it would be a fun experiment to measure it some time, and compare our layouts :)

The N and the R are something of a problem

I think all layouts always will have at least one same finger usage that is “outstandingly” worse than the others. NR, at 0.24% according to my data, isn’t that bad. anishtro, Maltron, Colemak and Arensito all have CS, at 0.23%, as the worst same finger, while Dvorak has CT at 0.43%.

but commas and periods tend to be the great majority (combined they total about 60% of all punctuation), with hyphens, double quote, and single quote/apostrophe another 30+%. Exclamation points, question marks, and the rest I stuck around NR are relatively few in number.

Again, your data seems to be pretty correct compared to mine. Where did you find all these things?

Code: [Select]
[ [ ',', 31.321236766865002 ],
  [ '.', 28.48866973185998 ],
  [ '-', 8.086459701723205 ],
  [ ')', 7.359603199134533 ],
  [ '(', 7.340284753883007 ],
  [ '"', 5.737336759137625 ],
  [ '\'', 3.9979522448033387 ],
  [ ':', 3.2305270071864616 ],
  [ '/', 1.3252453442546945 ],
  [ ';', 1.1897747469283673 ],
  [ '%', 0.5131462019936636 ],
  [ '&', 0.37284599335445484 ],
  [ ']', 0.2368924348968395 ],
  [ '[', 0.23544355150297505 ],
  [ '!', 0.1352291167606831 ],
  [ '?', 0.11446178811529248 ],
  [ '#', 0.09562630399505448 ],
  [ '+', 0.06858048064291786 ],
  [ '*', 0.03549764314967931 ],
  [ '_', 0.027045823352136622 ],
  [ '|', 0.01835252298894985 ],
  [ '~', 0.01448883393864462 ],
  [ '>', 0.012556989413492 ],
  [ '<', 0.010866625453983463 ],
  [ '@', 0.009900703191407156 ],
  [ '^', 0.0067614558380341555 ],
  [ '}', 0.006037014141101924 ],
  [ '{', 0.0057955335754578475 ],
  [ '`', 0.0033807279190170778 ] ]

Have you ran your keyboard layout through the patorjk.com site? I might try it if I have a chance to see how mine compares (I think MTGAP is a lot more accurate, but the ease of use of the patorjk keyboard layout analyzer is awesome, so I play with that a fair amount).

Yes, tried to use that particular analyzer in the beginning, but as I write in How anishtro was made:

Quote
There are lots of different keyboard layout analyzers, for example one by Patrick Gillespie. This particular one lets you use your own text corpus, which meant that I could use my Wikipedia text. However, that analyzer (as well as many others) measures things I’m interested in. I want to focus on same finger usage and rolls. So ultimately, I created my own tools: ext-frequencies-analysis, as well as the scripts in this repository.

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #18 on: Sat, 22 August 2015, 11:11:17 »
(Lots snipped to keep the size down)
Again, your data seems to be pretty correct compared to mine. Where did you find all these things?


Well, a good chunk of it was picked and prodded out of the MTGAP page at Mathematical Multicore, https://mathematicalmulticore.wordpress.com/2012/02/27/introducing-the-thumb-keys/#comments. as well as http://mdic*kens.me/typing/letter_frequency.html . His code on Github also has lots of statistics: https://github.com/michaeldic*kens/Typing (check in the data folder). Take out the asterisk from the last two URLs -- Geekhack censors ****ens.

And Peter Norvig's page is amazing for the huge number of statistics he was able to get from the Google Corpus: http://norvig.com/mayzner.html . Seriously, this is a fun page.

Plus, I did lots of Google searches to find more information, like for punctuation: http://homepage.ntlworld.com/vivian.c/Punctuation/PunctFigs.htm

You've probably already seen most or all of these, but that's pretty much where I got my start. Then on the keyboard layout analyzer, I input things like the collected works of Mark Twain, collected works of Jane Austen, and other very large corpuses (corpii?) to see how the stats changed for my keyboard layout. :)

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #19 on: Sat, 22 August 2015, 11:49:55 »
Thanks! I’ll re-run my tools with the data from Peter Norvig’s page as well as Michael D.’s and see how it affects the results.

Online Findecanor

  • Posts: 4297
  • Location: Stockholm, Sweden
  • Dans öl tarvö... "Pöngö"
Re: Introducing the anishtro letter layout
« Reply #20 on: Sat, 22 August 2015, 13:15:49 »
Off-topic: I'm silly, I know, but... I get peeved that I always read this thread title as butthole-poop-hurl. (synonyms)

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #21 on: Sat, 22 August 2015, 16:10:07 »
Thanks! I’ll re-run my tools with the data from Peter Norvig’s page as well as Michael D.’s and see how it affects the results.

I would be interested in seeing that! Both have a ton of downloadable data -- Peter Norvig even has a 11 MB zipped file with all of his data in tabular form, if you like big spreadsheets. :)

Offline jacobolus

  • Posts: 3642
  • Location: San Francisco, CA
Re: Introducing the anishtro letter layout
« Reply #22 on: Sat, 22 August 2015, 19:42:50 »
I get peeved that I always read this thread title as butthole-poop-hurl. (synonyms)
Is this a swedish thing or what? I don’t get it.

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #23 on: Sun, 23 August 2015, 00:13:43 »
I get peeved that I always read this thread title as butthole-poop-hurl. (synonyms)
Is this a swedish thing or what? I don’t get it.

It is definitely not a Swedish thing, because I’m Swedish and I wouldn’t name my keyboard layout to such a thing. I think Findecanor is reading “anis” with a U instead of an I, and then “sht” with an I before the T, and finally “htro” with H and T swapped and a silent W at the end.

I admit “anishtro” isn’t the best name :)

There also happens to be an unfortunate hidden subliminal message in the right hand side home row, if you type all five of them from left to right, alternating with the left thumb between all keypresses except before the last one.
« Last Edit: Sun, 23 August 2015, 05:28:17 by lydell »

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #24 on: Sun, 23 August 2015, 06:39:13 »
I’ve compared with Norvig’s and Dıckens’ data now. They gave about the same results, but also resulted in anishtro3—yet a marginally different variation that may or may not be “better” and that I won’t bother learning. You may look at the diff if you want to read just about the newly added Norvig/Dıckens stuff.

Online Findecanor

  • Posts: 4297
  • Location: Stockholm, Sweden
  • Dans öl tarvö... "Pöngö"
Re: Introducing the anishtro letter layout
« Reply #25 on: Sun, 23 August 2015, 07:26:25 »
There also happens to be an unfortunate hidden subliminal message in the right hand side home row, if you type all five of them from left to right, alternating with the left thumb between all keypresses except before the last one.
Bah. That's is just a Greek prefix that can be used in lots of contexts.

Top right, right side, from right to left, a message about mechanical keyboards for games: Weapons of Mass Pwnage ...
Top left: you are obviously a fan of Star Trek The Next Generation... J Luc. piQuard...

;)

Offline spindle

  • Posts: 28
Re: Introducing the anishtro letter layout
« Reply #26 on: Tue, 01 September 2015, 03:51:56 »
Thanks for posting your reasoning/process behind your layout, it was a really interesting read.

I've been using a modified colemak with 'e' on a thumb, and I think I've got it to a point where I feel everything is comfortable on my physical layout:

Code: [Select]
qwfdk jluy;
arstg bnhio
zvcpx -m,./
   ␣   e 

Since you considered modifying Colemak yourself, I'd be interested to hear your thoughts on this layout. You mentioned that you didn't like 'v' taking the spot of 'x', but that hasn't bothered me yet (I'm not a js programmer :p). But I'd be interested to hear any of the other ideas that you tried out (you weren't at the stage of putting 'e' on a thumb though???).

Something that I also value in a layout is having similar lefthanded keys to Qwerty in the layout, as this makes using programs that have keyboard + mouse workflows (like CAD and graphic design software) nicer since they general give the best shortcuts to left handed Qwerty keys. Unlike Colemak, I don't really care about preserving the exact positions of just as long as they're on the same hand. Most of the ergonomic layouts that I looked at don't really seem to prioritise this.

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #27 on: Tue, 01 September 2015, 14:53:31 »
I think you layout is very good, and it scores pretty well in my analyzer (1.6% same finger, 6.14% rolls). If I had found this one before setting out to create my own I might have chosen it.

You have managed to avoid lots of same finger usage in ways I recognize from my own layout, from Arensito and from Maltron. All of your letter columns are very sensible. The only unfortunate pairs are BL and BM.

Personally I prefer a bit more rolls, though. The left hand side home row is pretty good, but on the right side you lose some because H is only really a good roll together with T. But that’s just me.

think I've got it to a point where I feel everything is comfortable on my physical layout

Perfect! I’m glad that you’ve found something that really works for you. I believe that keyboard preferences are very personal, so finding something that works for _you_ is very important.

Since you considered modifying Colemak yourself, I'd be interested to hear your thoughts on this layout. You mentioned that you didn't like 'v' taking the spot of 'x', but that hasn't bothered me yet (I'm not a js programmer :p). But I'd be interested to hear any of the other ideas that you tried out (you weren't at the stage of putting 'e' on a thumb though???).

I even started to modify Colemak. That’s when I ended up with V on the same spot as you once. I think this is a good example of how we’re all different. I can’t explain it, but any time I type VA using your layout I feel something I don’t like along my entire wrist. If you don’t notice anything in particular with that combination, then by all means go for it!

You’re also right that I wasn’t at the stage of putting E on a thumb while trying to modify Colemak, so I never tried that. I don’t really remember anymore what I tried. What I do remember is that I never came up with anything I liked and concluded that modifying an existing layout was not the way to go for me.

One single thing I remember from Colemak was that I disliked typing “you” with just a few fingers on the same hand. I really like using both hands in my layout. What do you think?

Something that I also value in a layout is having similar lefthanded keys to Qwerty in the layout, as this makes using programs that have keyboard + mouse workflows (like CAD and graphic design software) nicer since they general give the best shortcuts to left handed Qwerty keys. Unlike Colemak, I don't really care about preserving the exact positions of just as long as they're on the same hand. Most of the ergonomic layouts that I looked at don't really seem to prioritise this.

I like that you have a really good reason for keeping lefthanded QWERTY keys on the same side. I prefer to hold modifiers with one hand and press the letter of a keyboard shortcut with the other, but I don’t use such mouse intensive programs much, though. But when I do, I agree that it is annoying to move between the keyboard and the mouse all the time. I try to avoid the mouse as much as possible so that’s why I didn’t prioritize that. Most ergonomic layouts seem to be optimized for _typing_ – lots of typing – which can easily  be done without using a mouse at all. I like that you’ve made a layout that is optimized for a bit different practical use cases.

Offline spindle

  • Posts: 28
Re: Introducing the anishtro letter layout
« Reply #28 on: Tue, 01 September 2015, 18:32:17 »
You have managed to avoid lots of same finger usage in ways I recognize from my own layout, from Arensito and from Maltron. All of your letter columns are very sensible. The only unfortunate pairs are BL and BM.

Yea I agree, 'bl' is probably the worst same finger bigram on my layout. In Colemak 'k' takes that column, and I found that it was worse because 'k' appears in 'kn', 'nk' and 'lk' which together I find appear signifcantly more often than 'bl' and 'bm'. I don't think I can do much better without significantly modifying Colemak to point where I might as well start from scratch.

One single thing I remember from Colemak was that I disliked typing “you” with just a few fingers on the same hand. I really like using both hands in my layout. What do you think?

Typing "you" in Colemak doesn't bother me. I think I remember it feeling a little weird to type at first, but I think I quickly developed a motion where I quickly tap 'y' and then quickly roll from 'o' to 'u'. After that it started feeling fun to type. If I try type out the letters slower and uniformly it doesn't feel as nice.

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #29 on: Wed, 02 September 2015, 10:00:05 »
I just reprogrammed a Kinesis 110 with my YOR UIAN layout a couple of days ago (as close as I could get, anyway -- don't know of a graceful way to split punctuation, like / ? and ' ", let alone number punctuation combinations like 1! and 2@ and the rest -- especially parenthesis!), and have gotten to the point that I'm mostly able to type slowly without using my cheat sheet. I like it quite a bit, and the keys are where I expect them to be.

I am waiting for the new Kinesis Advantage, to see if it's even more programmable, so I can do things like have punctuation unshifted and numbers shifted on the top row (since I have a separate number pad anyway). It will be fun!

Offline vvp

  • Posts: 790
Re: Introducing the anishtro letter layout
« Reply #30 on: Wed, 02 September 2015, 13:00:51 »
I am waiting for the new Kinesis Advantage, to see if it's even more programmable, so I can do things like have punctuation unshifted and numbers shifted on the top row (since I have a separate number pad anyway). It will be fun!
You can do that even with the current Kinesis. Lets say you want to swap 5 and % (Shift-5). Define macro with trigger 5 and content % (Shift-5). And define macro with trigger Shift-5 and content 5.

Really, aside from occasional bugs, Kinesis firmware is pretty good. It's on-the-fly remap and macro features are nifty. Of course, the current kinesis firmware has also some pretty bad features, like e.g. switching keyboard to MAC mode when you press chord Equal-M. You will hate that if you are a programmer since it can be pressed by accident when you are writing stuff like "p=malloc(4)" - if you press m before releasing = then you are screwed  :)
Any configuration stuff should have been done through a chord starting with otherwise unused key which is hard to press by accident like e.g. the Program key.
AFAIK, the only open firmware which has both on-the-fly remap and macro is chrisandreae's firmware.

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #31 on: Wed, 02 September 2015, 17:09:31 »
I am waiting for the new Kinesis Advantage, to see if it's even more programmable, so I can do things like have punctuation unshifted and numbers shifted on the top row (since I have a separate number pad anyway). It will be fun!
You can do that even with the current Kinesis. Lets say you want to swap 5 and % (Shift-5). Define macro with trigger 5 and content % (Shift-5). And define macro with trigger Shift-5 and content 5.

Really, aside from occasional bugs, Kinesis firmware is pretty good. It's on-the-fly remap and macro features are nifty. Of course, the current kinesis firmware has also some pretty bad features, like e.g. switching keyboard to MAC mode when you press chord Equal-M. You will hate that if you are a programmer since it can be pressed by accident when you are writing stuff like "p=malloc(4)" - if you press m before releasing = then you are screwed  :)
Any configuration stuff should have been done through a chord starting with otherwise unused key which is hard to press by accident like e.g. the Program key.
AFAIK, the only open firmware which has both on-the-fly remap and macro is chrisandreae's firmware.

Awesome! I will have to try that when I get home. Thanks!

Offline jeffgran

  • Posts: 126
  • Location: Denver
Re: Introducing the anishtro letter layout
« Reply #32 on: Sat, 12 September 2015, 12:04:51 »
@lydell This is really cool.

I was looking at your code to see if I could modify it for what I'm thinking about doing. It looks like you somehow have the e key "hard coded" to the thumb area? The different layouts in `layouts.js` all only have the finger areas defined.

What I was thinking about doing is something similar to what you did but using my own keylogger data (which I still have to start recording). I'm a programmer and I use emacs (and mutt for email) most of the day, so I end up using a LOT of keyboard shortcuts mixed in with actual typing, so I wanted to take that into account when I develop my layout.

Anyway thanks for sharing. I do agree with you on the rest of your priorities (rolling feels so nice, same finger twice in a row feels so annoying) besides the e on the thumb. I have too many modifiers to do that.

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #33 on: Sat, 12 September 2015, 12:34:00 »
I was looking at your code to see if I could modify it for what I'm thinking about doing. It looks like you somehow have the e key "hard coded" to the thumb area? The different layouts in `layouts.js` all only have the finger areas defined.

The letter E is not hard coded to the thumb area. Nothing prevents you from adding E on any other key, which can also be seen in some of the layouts in layouts.js. What’s true, though, is that my analyzer does not care about thumb keys at all. My analyzer is not a tool that tells you which of two or more layouts is the best. It only measures same finger usage and rolls (rolls the way I define them), because that’s tedious to do manually, as opposed to the rest of  the characteristics I was looking for in a layout.

What I was thinking about doing is something similar to what you did but using my own keylogger data (which I still have to start recording). I'm a programmer and I use emacs (and mutt for email) most of the day, so I end up using a LOT of keyboard shortcuts mixed in with actual typing, so I wanted to take that into account when I develop my layout.

When you get around doing that, please share! It’d be very interesting to read about.


Also, if you’d like to use any of my code don’t hesitate to ask for help if it is difficult to understand. I made it primarily for myself, but tried to keep the code clear as well as include comments and write documentation, but it might be lacking.

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #34 on: Sun, 13 September 2015, 22:48:59 »
I tweaked my layout a bit to improve that stats--well, okay, more than a bit of tweaking, though there is a lot of overlap. I did this for my Kinesis, so I left off the ErgoDox keys I don't have. I'll add a simplified version without AltGr characters, but here is the complete version in all its glory:
 110940-0

I'm hoping "Inline text link" means I can attach the patorjk.com import data so other people can play with it.
* UP_OIANY_4.txt (14.16 kB - downloaded 66 times.)

Edited to read: Actually, I think an OU HNAIY format for the left hand might be even better. The problem is that it's so much fun to play around with layouts, it's hard to settle on just one.
« Last Edit: Mon, 14 September 2015, 12:01:44 by Snarfangel »

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #35 on: Thu, 17 September 2015, 12:52:35 »
Interesting layout, Snarfangel! It actually performs pretty well in my analyzer:

Code: [Select]
{ sameFinger:
   [ [ 'nr', 0.24390161897603993 ],
     [ 'pt', 0.09824084615667118 ],
     [ 'ao', 0.09609421189814711 ],
     [ 'lm', 0.061256665558003354 ],
     [ 'gs', 0.059433319591425664 ],
     [ 'kl', 0.040798982443635334 ],
     [ 'kl', 0.040798982443635334 ],
     [ 'lv', 0.03566516436752655 ],
     [ 'iy', 0.032070198561082605 ],
     [ 'fs', 0.025125966050073993 ],
     [ 'hl', 0.023975059730744822 ],
     [ 'hk', 0.020393025455978736 ],
     [ 'hk', 0.020393025455978736 ],
     [ 'hm', 0.01853088489436749 ],
     [ 'km', 0.01761274614523972 ],
     [ 'km', 0.01761274614523972 ],
     [ 'cx', 0.017354115511682606 ],
     [ 'bc', 0.01229788662564095 ],
     [ 'dt', 0.011405610939868895 ],
     [ 'dp', 0.007784782070069245 ],
     [ 'cz', 0.0061683406103372595 ],
     [ 'cq', 0.005780394660001583 ],
     [ 'bw', 0.005366585646310195 ],
     [ 'cw', 0.0038148018449674875 ],
     [ 'fg', 0.0035432396797325134 ],
     [ 'hv', 0.0022242234485912132 ],
     [ 'bz', 0.001875072093289104 ],
     [ 'mv', 0.001655236054765554 ],
     [ 'kv', 0.0015647153330205627 ],
     [ 'kv', 0.0015647153330205627 ],
     [ 'bx', 0.0009569333441613358 ],
     [ 'wz', 0.0008017549640270651 ],
     [ 'wx', 0.0007888234323492093 ],
     [ 'xz', 0.0003362198236242531 ],
     [ 'qx', 0.0003103567602685414 ],
     [ 'bq', 0.00028449369691282956 ],
     [ 'qz', 0.00018104144348998245 ],
     [ 'qw', 0.00016810991181212656 ] ],
  rolls:
   [ [ 'ht', 3.1143912946997787 ],
     [ 'an', 2.7104361081469164 ],
     [ 'or', 2.2012958429263745 ],
     [ 'st', 1.5176186946498116 ],
     [ 'ai', 0.7833404629177985 ],
     [ 'lp', 0.2623419831486624 ],
     [ 'cs', 0.23447453238288302 ],
     [ 'iu', 0.13753977092567524 ],
     [ 'oy', 0.10984043007170793 ],
     [ 'dv', 0.01916452994658243 ],
     [ 'df', 0.009621059568324783 ] ],
  sameFingerTotal: 0.9381308971017337,
  rollsTotal: 11.100064709384517 }

It’s interesting to see how you’ve chosen such prominent spots for many punctuation keys on the left, and that you’ve put way more letters on the right hand. Are you a programmer? If not your layout looks really good for programming when thinking about punctuation characters. Myself I’ve got most such characters in another layout around the home row.

Lydell, I was looking at your information again (and on the GitHub side). On your layout, did you add the cs and sc probabilities? I see

> layout(layouts.anishtro)
{ sameFinger:
   [ [ 'cs', 0.23447453238288302 ],


but from Norvig, cs is 0.023% (643,530,723) and sc is 0.155% (4,363,410,770). I could be misreading your numbers though.

Offline lydell

  • Thread Starter
  • Posts: 42
  • Location: Sweden
Re: Introducing the anishtro letter layout
« Reply #36 on: Fri, 18 September 2015, 11:26:08 »
Lydell, I was looking at your information again (and on the GitHub side). On your layout, did you add the cs and sc probabilities? I see

> layout(layouts.anishtro)
{ sameFinger:
   [ [ 'cs', 0.23447453238288302 ],


but from Norvig, cs is 0.023% (643,530,723) and sc is 0.155% (4,363,410,770). I could be misreading your numbers though.

Yes, I’ve added all bigrams that are the same when reading one of them backwards together. I call them pairs. To me it is not interesting how common a pair of letter is one way or the other, only the combined frequency.

Offline Snarfangel

  • Posts: 276
Re: Introducing the anishtro letter layout
« Reply #37 on: Fri, 18 September 2015, 16:42:01 »
Lydell, I was looking at your information again (and on the GitHub side). On your layout, did you add the cs and sc probabilities? I see

> layout(layouts.anishtro)
{ sameFinger:
   [ [ 'cs', 0.23447453238288302 ],


but from Norvig, cs is 0.023% (643,530,723) and sc is 0.155% (4,363,410,770). I could be misreading your numbers though.

Yes, I’ve added all bigrams that are the same when reading one of them backwards together. I call them pairs. To me it is not interesting how common a pair of letter is one way or the other, only the combined frequency.


Ah, okay, thanks!