The same thing happens with CAT models (remember “near-term rates?)
From John Taylor’s post on the fed’s model error:
The solid black line plots changes in the estimated sacrifice ratio of unemployment to changes in inflation implied by the model over 64 vintages.
Could easily be an actuary’s pricing model. That’s overfitting, folks. Responding to specific (recent) data points without and real generalizability. We must be humble in what we can know about very complex things.