This message was deleted.

Hey. Can you try something like this? You're probably missing a date or something similar. The important thing isn't the shape but the actual values, since an inner join is performed between the expected future and what you provide. With that snippet you're able to replicate the structure that is being used to validate, you can then join it with that you're providing and see if you're missing something (example)

Thank you, Jose. Yes the issue is that my data (neither train nor valid) does not have a value for every day, but only business days. Would the recommended approach be to add rows for the missing days, and if so would it be best to leave the data for missing days as NaN or impute values (e.g. propagating values from the most recent non-missing day, i.e. most recent actual data)?

Can you try setting

freq='B'

?

yes 👍