I enjoy it - the fickleness adds spice.
But actually I just thought about the ENS spread again, a bit more scientifically. What’s actually happened is the mean goes up and down more. So this implies the opposite of what might seem intuitive: the wavier the nean, the more the model has nailed the day on day fluctuations consistently between perturbations. A flat mean line means lots of different fluctuations on different runs averaging out.
So rather than changing its mind, the GFS ENS has actually made up its mind.
Originally Posted by: TimS