Here's the forecasts ranked by MASE score; also threw in peak dates - had to leave out date ranges owing to lack of ability/time:

Thanks!

It's a nice way to show the results, I'm going to keep the idea.

can someone test if there is a statistically significant correlation between assumed/predicted peak date and accuracy -- it seems that the 10 best performing models all peak in the next 5 years, while 10 out of the worst 15 predict a peak 20 or more years from now.