"model sufficiency" a lot of actual researchers spend entire lifetimes to model complex stuff at the lower level possible when a good model tuned on real data

"model sufficiency" a lot of actual researchers spend entire lifetimes to model complex stuff at the lower level possible when a good model tuned on real data will generally give better predictions. relevant quote : "all models are wrong, some are useful"