I don't have much in the way of old speed records.
Today: 415 CPM (that is 83 WPM), 419 CPM, but made 2 mistakes (out of 77 words).
I didn't really take the test seriously, or sit properly, and it's my 1st test on that site. Plus AutoCorrect appeared twice and ruined my flow.
In the past I did a few tests in a sitting and kept the best score. 94 WPM on an Apple Aluminium Keyboard with Numeric Pad, and 94 WPM on TextBlade a few months later.
Objective measurements are important, but I wanted to mention, I no longer have forearm pain on a workday, something that bothered me every workday prior to TextBlade. Also, in terms of productivity, it's hard to measure differences, but I certainly do a lot more arrow keys, selection, cut-and-paste, screenshots, and use strange shortcuts to start/terminate a debug session and step through code. All of those tasks are now done without my fingers leaving home row. I also no longer stretch my hands to reach backspace, forward delete, or even braces and brackets.
Also, I don't "practice" typing since I do it every workday (and as here, non-workdays too). My speed at certain patterns would improve with practice, but there isn't really a set of practice that is suited for coders.
For example, with Visual Studio Intellisense, if I wanted to type:
By the time I got to
it would suggest, and by
the suggestion would certainly be correct.
Prior to TextBlade I would type a lot more rather than use the arrow keys to choose between suggestions, because moving my hands would incur a greater penalty than just typing more characters. Now, I'm free to use arrow keys without leaving the home row, so I use Intellisense suggestions far more.
And though I use keyboard shortcuts almost exclusively, to this day, I will struggle to remember whether it is Opt-Shift-Arrow, or Ctrl-Shift-Arrow, to select the previous word, etc. But with TextBlade, I just use the selection layer, and I don't even think about it.
It's almost "magical" how I just select exactly what I want, without interrupting the thought process.
These productivity gains are hard to design a quantitative test for, and, for those who care that much about productivity, hard to engage volunteers to run through batteries of tests.