For the reported risk factors in the new risk models, is there any report-out
of the standard error of these risk coefficients (or equivalently t-scores or p-values per coefficient)
and overall R2 or adjusted R2 of the models?
The general rule in regression models is to pay less attention to coefficients/factors with high standard errors
(because the high standard error means you don't know it very well).
If you try to optimize based on coefficients that have high standard errors, your optimizer may
be chasing randomness.