I also don't find the

max_horizon

argument particularly clear. If I'm working with weekly data and my horizon is 13 weeks and I want 13 models, I can set

max_horizon=horizon

and I get 13 models. Is it then possible to still have 13 models if I am working with the same data at daily granularity (91 day horizon) and where each model predicts a 7 day sequence of 1 week at daily granularity? I receive errors about

horizon can't exceed max_horizon

which implies not.

This isn't currently possible. If it's something valuable to you, can you please open a feature request here?