This should be talked about more with computer models. Statistical significance is never reported and by and large it isn't very good. This is particularly true for W/L models. For example, I ran a W/L based model and Texas was #1 and BYU was #19. But I cannot reject the hypothesis that those two teams are of equal quality. The p-value was 0.223 so its not even that close.

For a MOV model, precision is much higher. So, for example, in a pure MOV poll USC is #1. I get statistical significance in terms of the difference in rating between USC and the 9th rated team (Alabama) but not USC and the 8th rated team (OK State). Statistical precision is better for MOV polls because the data are richer. However, its not like it is fantastic even in this case (and this is in sample significance and doesn't take into account that I tested more than one hypothesis to come up with significance threshold) .