Thanks for this, Jesse! This is very insightful. Following-up on this, does it make any sense my comment on computing the one-step difference between consecutive elements of the GRW? In my head, I see these differences as a way to capture the overall trend and trend changes within historical data. So if I extract them and sample from them to mimic a similar behavior into the future, I should be able to keep pushing the same trend observed in the past. Not sure if this sounds reasonable and it it has any theoretical / practical support.