I read it. More than once. Like I said, your numbers and statistics are meaningless and you simply can't draw any conclusions from them. (The problem with your penultimate comment above in post 23 and is that you don't know which of the three was more accurate. You recognize this in the article but then proceed and attempt to draw conclusions, or let others do so, even when you know the analysis is deeply flawed.) To do it better you need a known-good reference measurement - and implementing that sort of test is very hard to do.
The best-case scenario is to compare against a known-good reference -- I agree! However, it also useful (though not as rigorous) to compare each device's measurements to the mean of all devices used. (Bryan Litz says as much in his book, which I linked to in the article).
Unfortunately, I don't personally have access to the laboratory-like facility which would support such an endeavor, nor the budget for that matter. I knew that going in, which is why I set different goals for myself. As with all of my work, I'm much more interested in the truth than being right or wrong. If I make a mistake, or could do things better, I'm happy to accept constructive criticism so that I can learn from it and improve!
Regarding usability, your comments are fair except to say that
if you have a chronograph which picks up neighboring shots (i.e. accepting that for the moment) having the ability to download the data is rather pointless - your data dump includes a lot of samples that need to be deleted. For practical use and load development the ability to download to an excel spreadsheet simply isn't necessary. You have relatively small sample sizes and it is very easy to shoot and then note the velocity with pen and paper. Keying these into Adam's online calculator
https://www.autotrickler.com/stats-calculator.html avoids the need for constructing a spreadsheet and knowing the formulae for all the statistics. More importantly, confidence interval analysis and determining whether two sample sets are meaningfully different is much more powerful for load development than just knowing the simple average and SD/ES. Understanding the maths behind this and even inputting such formulae into a spreadsheet would stretch a lot of people. Magnetospeed and, I believe, LabRadar don't incorporate such analysis tools into their products.
There's a lot in here to unpack, so I'll try to be concise and enumerate each of your pieces of feedback.
1) As I wrote, everyone has different use-cases, so the "practicality" of it's usage is highly dependent on the type of range/environment folks have access too. It's subjective.
2) My point about being able to download the data was one of time-management and also goes along with the previous point, since it may not always be practical for the user to write down every reading the Two-Box displays. If a device records data and gives it to you without having to write it down or type it in manually, then it will save you time (and prevent possible typos).
3) One thing I did not write was that the screen glare was bad enough during this experiment that I may not have been able to do that anyway.
4) I agree that the simple statistics that folks commonly use are not the best indicators, and that spreadsheets kind of suck! I am working on a separate article which focuses purely on statistics, which will cover this much more thoroughly and will be sure to glaze over many eyes.
Of course, having a chrono that doesn't pick up neighboring shots is advantageous but practical use of the Two Box isn't difficult or annoying. With regard to positioning/setting it up, if you even care about this, I think you will find having the sensor boxes at the same height is much more important than having the boxes sitting level. So the bubble levels likely don't add much if anything at all.
I had written about a previous experience with the Two-Box where we used it at a busy 600-yard line. We weren't super careful with how we set it up (just plopped it down basically, like you had suggested earlier) -- the sensor boxes were clearly tilted, which we suspected had lead to them picking up neighboring shots. That's why I decided to be much more careful this time, even to use bubble levels to avoid any potential sources of error. It may not be necessary, but it also shouldn't hurt.
You raise an interesting point though -- were my boxes at the same height? I'm not sure I have a good way to measure that, honestly. Here's a photo snapped from the video where I was getting it lined up, so you can see for yourself. Obviously, the camera doesn't show the same angle as my eyes. Anyway, getting the same height actually seems like kind of a hard problem to solve -- you don't have a ton of wiggle room when shooting prone and no real way to measure their relative height.
It would be nice to see a solid-mount system for the Two-Box, like a rail, to make it a bit easier to align.
PS I have both the Magnetospeed and the Two Box. I don't use the Magnetospeed anymore. If I want to compare the SD/ES of strings of fire I shoot over the Two Box with a "just plonk it down" set up. Its measurement of variance (SD) isn't clouded by as much inherent measurement error and I don't care about the absolute figures per se. For absolute average velocity measurements, for computing (rough) expected trajectory for example, I just use my ShotMarker.
Hey, if the Two-Box works for you -- that's great! We're fortunate enough to get to use the ShotMarker every week during 600-yard F-Class practice. I really like that they provide velocity data, though I have not seen it's error rate published, so I'm not sure what variance to expect.
Anyway, it seems like you have some strong opinions about this and that's fine -- it's fine to disagree!