I only pay any attention to PG data, since those are the only ones my son attends/attended that record this stuff, but as you dig down there it's pretty obvious that a lot of the peak numbers are bad data.
As an example, http://www.perfectgame.org/Pla...vents.aspx?ID=345971.
All of this kids pitching numbers follow a pretty reasonable progression with two outliers, in June of '14 and again in March of '15. In his particular case, I don't know what's up with the 6/14 series, but the 3/15 game isn't actually him, it's another kid from the same HS where the data got recorded wrong.
If I'm "scouting" this kid from my sofa, I'm ignoring the outliers because it seems clear that the rest of the numbers paint the picture, but it's a potential problem for guys with only 1-2 data points (which is why I assume no one signs guys based on just 1-2 data points that haven't been personally verified).
I've got cases of my son's teammates where there's no way the above type of mistake (wrong player) was the issue, but where there are still spurious peak numbers reported. One lefty whose peak is listed as 83, but he always works 75-78, for example. In my experience, this sort of bad data exists for around half of the pitchers I personally know. Again, it's not really a problem if you're ignoring outliers for guys with lots of data points (which is the only sane thing to do, in my opinion). I've always assumed that this was the occasional batted ball exit velocity getting recorded with the pitching velocities, but that's just a guess.
PGs newly available Trackman data would appear to go a long way towards fixing that latter problem. It's detailed pitch-by-pitch, so it'd be easy to see one pitch weirdness if it exists. It appears to mis-classify pitches on occasion, so you'lee see a guy with an 85mph fastball follow that up with a 75mph fastball, but if you look at the totality of the pitches in a game it becomes clear that the 75mph pitch isn't actually a fastball. Unfortunately, at least at LakePoint, the Trackman data seems to suffer a lot from the first problem right now, getting pitches assigned to the wrong pitcher. My son's limited data set makes him an ambidextrous pitcher who goes upper 60s to low 80s from the right side and low to mid 70s from the left and with 4 distinctive arm slots. If only, he says. :-) That's probably an artifact of growing pains implementing the system, possibly combined with less attention to detail for players who aren't on stud-filled teams, if I'm guessing.
