Ah sorry you need to score them independently and then combine…
Not any more!
Lots of work from @nikosbosse here whilst being heckled by @sbfnk and I.
This just supports the energy score for now (wraps scoringRules) and probably needs an update method or similar to make it easier to redefine multivariate objects as well as map from sample to multivariate objects (both likely implemented as methods for as_forecast_multivartiate_sample).
Interested to see what people do with this. For myself, I don’t currently have a clear view of what is good practice in terms of treating as a related forecast (I would assume horizon).