Discussion about this post

User's avatar
Negitrage's avatar

Regressing Prometheus "disagreement with consensus" vs "realized surprise" might be more interesting? AFAIK nearly everyone who tries is able to to get a decent correlation of forecasts to realized. Not to detract from the model, just saying probably can find more compelling datapoints to prove reliability?

Expand full comment
MFrankel's avatar

Pretty big miss from Prometheus but actually in line with scatter plot spread -- shows 'actual' can go to 0.1% when estimate at 0.39%

Expand full comment
5 more comments...

No posts