I believe smoothing is a regression where all of the data is fitted to a line. If there were a major trend it would show up in any amount of smoothing. We’re talking about thousands of data points (or more) over a short amount of time here so tens here or there that are abnormal aren’t a) going to be felt and b) not accounted for. If the car were breaking up on the dyno I think it would show up since that would likely consist of hundreds of data points. If the car were surging, I think that too would show up in the data.
Sent from my iPhone using Tapatalk
Not always the case, it's been very apparent to me that even though the car I drive doesn't change (and nor the fuel I use) that the first WOT tunes weren't able to handle part throttle conditions and would surge while the car hunted for the correct AFR output (among others).
After picking out specific rev ranges that had issues I was able to get my tuner to adjust the tune in those slots and allow for smoother part throttle driving.
Yes, major trends would show up, but the point of smoothing is, as you stated, to average out the numbers and essentially create fewer sampling points to connect. If you have fewer sample points to join, the smoother the line becomes when comparing data in between the start and end of the dyno run.
I would imagine there aren't many oddities at part throttle, they would be harder to detect on a WOT pull anyways, I just always saw a lower smoothing to be a more accurate representation of the true performance.