I'm relying on the work of others here, but essentially, if you were to recursively walk through each team's ratings progression throughout the year, you would find that after any given game, you would be better off predicting that a team moves towards its YTD performance average rather than further away from it.

There are definite exceptions to this. There have been teams that have just progressively increased their ranking all season (Kyle Smith's new squad, for instance). But that's the exception, not the rule. Generally, a short term vacillation in performance tends to regress back toward the dominant YTD trend.

My hypothesis on this is that style-driven improvements or declines are more likely to sustain and not regress than improvements or declines driven by elements more influenced by luck or variance (jumper shooting for and against, free throw defense, non-steal TOs). Why I'm so suspicious of Penn's sudden surge is that its eFG% in the four-game surge has been 48%, 62%, 70% and 59%. It was 57% at La Salle. In the four and three game slides in Jan and Feb it was 49%, 41%, 49% and 49% and 42%, 55% and 47%.

Looking at it from a eFG% for vs. against, you see that during the slides they went:
+11%
-11%
-6%
+3%
-6%
-3%
-16%

And during this run:
+5%
+15%
+19%
+21%

I'm HIGHLY skeptical of runs driven by outsized eFG% differentials. FWIW, Penn's long run eFG% differential is +3% for the year.