Hi, author here! Lots of submissions have come in, and I'm starting to automate the analysis pipeline while I'm waiting for more submissions. Some highlights of

Hi, author here! Lots of submissions have come in, and I'm starting to automate the analysis pipeline while I'm waiting for more submissions. Some highlights of what I've found up to this point: - It appears that 10 people submitted test results from more than one keyboard. Awesome! That will be fun to dig into. That is what'll truly let me compare keyboards without being too worried about confounders. - The record holder submitted results from four (!) different keyboards. That beats myself as the previous record holder with only three keyboards. - Where the description made it possible to say how the keyboard was connected, wired is most popular (63 %), but it is also surprisingly common to wire the keyboard indirectly through some sort of USB hub, e.g. a dock or monitor. - The default USB polling rate of 125 Hz is by far the most common, in cases where the polling rate is possible to extract from the tap durations. - Split keyboards make up around 20 % of the keyboards where their design is known. The number of keyboards with programmable firmware is similar. - When I can judge the age of the computer used from the description, it is most often 3–4 years old. - The median reaction time on the reaction task across all trials is 260 ms. The tenth percentile is often considered the "practically optimal", and across all trials it was 210 ms. This confirms what I would expect, which is promising in terms of data quality. But it can depend a lot on what keyboard you have. - The mean reaction time within a submission is typically 20 % slower than the optimum. This probably means most submitters were relatively well-rested when they took the test. - There is actually a small learning effect within submissions. The mean reaction time at the 10th trial was 270 ms, and the mean at the 25th trial is 267 ms. That looks like noise, but the slope is remarkably consistent. (The first five trials have a mean reaction time of 277 ms, so any serious analysis would exclude those, I think.) - The optimum reaction time for users with wireless keyboards is 6 % slower than that for users with wired keyboards. (Wired through a hub is a few ms slower than wired directly to the computer.) - Users with Apple keyboards are 8 % slower than non-Apple keyboards. This is probably confounded by wired/wireless status. - External and built-in keyboards are equally fast, which is surprising. The mean reaction time of external keyboards ought to be confounded by wired/wireless status, and thus slower. But it could be the case that built-in keyboards are also slower enough to compensate. - Split keyboards are 4 % slower than non-split keyboards. This could be because the halves communicate by bluetooth (as it is in my case), or confounded by wired/wireless status. - Some keyboards have a weird tap anomaly where tap durations are at most a handful of milliseconds. Normal keyboards have tap durations of 50–150 ms. The tap anomalous keyboards are 24 % slower at the reaction task. I suspect they buffer internally to be able to produce those mechanically clean tap durations, but at the cost of response time? - The relationship between response time and polling rate is not clear. Probably confounded by a bunch of other things. Thanks again, and keep contributing data if you haven't already!