It’s not a property of the data by any means. I guess I am used to seeing that “smoother” trend when using a median vs a mean, and it just didn’t seem to be the case here.
It makes more sense that my assumption/expectation was wrong, rather than two separate packages being incorrect and producing the same result.
But, here is the data, I’m using plotted behind the mean and medians, just so you can see what the distribution looks like:
I’m thinking I should instead remove the extreme outliers (maybe a density based algorithm), and then try the rolling mean and medians again. 