I realized the reason my data seemed subjective is because it used the first data point as the initial speed for the model. But since the GPS exhibits noise, there is some error in the first data point. And since I can slide my starting point around, that adds a subjective element. Error in starting speed will translate into an error in Cd.
I made the model initial velocity a fit parameter instead. The subjectivity in my data analysis disappeared.
This data is garbage. Probably my bullnose has no effect, at least in this incarnation.
I have always felt like the drag in my car goes down after it's run 5-10 miles, especially in cold weather. One day I jacked all the wheels up to figure which breaks are dragging, and I found that the front wheels are at fault. But I don't think it's the breaks. I can't hear any rubbing or sliding at all. It must be a differential or the transmission. (Yes, it was in neutral.)
I think what we're looking at in this chart is a change in temperature and viscosity of an oil that needs to be changed. The manual transmission is the most likely candidate. This car has 222k on it, there's a decent chance is has the originial manual transmission, and there's a decent chance the transmission oil has never been changed.
The only thing this chart suggests to me is that my impressions were correct and until I fix it there is no point in doing any drag testing on this car. (Or
maybe it would work out okay if I did a warm-up drive first, but it would still be a wild-card.)