I'm not going to speak for
@AlexisLD, but there has been some reference to the fact that their processes, tooling, and software has all advanced over the years. I have also noticed some things that don't quite add up from my own experience and invariably it's when looking at older models when compared to newer. So hopefully he'll weigh in on data accuracy over time.
(I am happy to take these detailed discussions in some other thread if it is preferred. I don't want to divert too much from the original topic)
@Noodler, we just looked at this in details. There seem to be a problem for some of the data that were measured pre-2016 (we used very annoying force sensors back then, it must be it). We have no way to recover the data or correct it, so we removed every ski that was measured before then (about 160 skis). Sorry to have missed that and thanks for catching it. Note that some of the older skis in the database were measured post-2016 (e.g., in Blister's collection in 2017), so they are still in the database.
For the measurement done in 2017 (basically everything in the database that is older than 2017-18), we cross-checked with newer versions of these skis and the results seems to be more consistent. We looked at the
Brahma (in different lengths), the
Bonafide, the
Zero G 95 (seems like it got stiffer in torsion with years, which I would believe for a number of reasons but is not confirmed by the manufacturer), some
Ikonic, the
Wayback and the
Kendo. Some of these skis have changed through the years, so some variations would be expected and that even for skis that are claimed to be unchanged. We can either trust or not these measurements, but I think they are mostly useful even if less precise. Anyways, I would not bet my life on them but I think most of them are pretty good.
In 2018-2019, we introduced a "calibration ski" that we measure at regular intervals to rapidly detect calibration problems (i.e., multiple times during a day of measurements). It should all be good. However we don't measure the full length of the ski, so all measurement that depends on that will be wrong (e.g., setback, surface area, surface-to-weight ratio, rocker length, etc.). We are working on putting a N/A for all these values so that we don't confuse people.
In 2019-2020, we upgraded the forces sensors. They are now much more stable. Since then, we rarely need to recalibrate the machine. We still regularly check with the "calibration ski" to be sure. We also now save the full non-processed data from all the sensors, so we could correct any problems after the fact if needed or change the processing pipeline in the future. Most of the ski in the database have been measured post-2019.
In 2020-21, we added a device to measure the full length of the ski. We are not measuring the first and last 5 cm of the sidecut and camber. We interpolate to find the final camber height which should result in a fairly small error. However, now the all the measurement that depends on the full length of the ski will be fairly accurate (e.g., setback, surface area, surface-to-weight ratio, rocker length, etc.).
Next year, it is challenging to measure the width of ski with a convex base in our machine (but height/camber is OK). That is why so much of the sidecut is missing on these skis. They are not many of them, but we will modify the machine for next year...
Thanks for pointing this out. Let us know if you find anything else weird in the data. We want to provide the best information out there and it is helpful to have more than just our eyes to find ways to improve our process.