Advice on adjusting forecast intervals

[nerdcha]

Hello epiforecasts community! Thank you so much for sharing EpiNow2. I've been using it in a Twitter bot to generate daily forecasts during the latest covid outbreak here in NSW, Australia. It's a simple script, as in this gist, using basic epinow settings and 20k samples.

I've only got 17 sets of forecasts to evaluate so far, but it seems that while the median has been in the right spot, the forecast intervals have been a bit too conservative. The actual reported cases have been inside the model's 50% interval about 75% of the time, and likewise the 20% interval has turned out to be more like a 40% interval.

Could you suggest any ways to correct this? Would, say, tightening the priors on the Rt process make sense? And/or, are there other ways that I should be validating and debugging the model's output?