We can all infer that not all position groups impact the outcome of a season equally. This could be due to numbers (5 OL vs 1 or 2 TE) or nature of importance (QB vs long snapper). I’ve already looked at the correlation between talent level and outcome by whole roster and by offense vs defense. So this time, I just broke the same data down a little further by position groups.
Each scatter plot below has a box that shows the correlation and effect size. The correlation is how closely the two variables (2019 win % and average composite talent rating for that team’s position group) are related. The effect size (technically, a coefficient of determination, but that can be interpreted as effect size in regression) essentially shows how reliable that correlation is. Every relationship was statistically significant at 0.06 or lower.
QB was done differently due to sample sizes. When I used the average rating of all QBs on each roster, the outcome was random. In situations like Florida, where the highest rated QB was not the starter, the highest rated was used (So, Feleipe Franks for UF). Of course, this doesn’t make sense on the surface, but I surmised that using the highest ranking individual is a general indicator of how good the QB situation is for a team. If that QB gets beat out by a lower rated player, then the lower rated player may have been undervalued (or the higher rated player was a bust…). Either way though, using the highest rated player at the QB position had some value, so I used it. If you have a better explanation (and I hope you do), feel free to share.
Position Group Effect Table
Here is a table summarizing effect size by position
Surprising to me, is that the LB position group had the strongest correlation to win % for SEC teams in 2019. Keep in mind, this isn’t how important a position group is to winning – this is how well the composite rating of that position correlates to winning. Most of us would agree that effective QB play is key to winning. Here is that output visually in a radar graph:
Expectations scores around zero, whether positive or negative should be considered as meeting the expectations. Here is the table with each team’s total:
Looking at each team’s performance relative to expectations by position group is a neat way to peer into how well the on-field coaching is. It is important to note that teams with higher talent ratings will have harder times surpassing expectations and it will be easier to fail expectations. Not surprisingly, LSU had the best on-field coaching performance. Winning all your games and a Natty should cement that. Relative to their on-roster talent levels, Florida was second only to LSU. Imagine what Dan Mullen could do with elite talent… he’s already not far off with the talent he has.
Dan Mullen is one of the best on-field coach in the SEC, and could hit a championship mark if he improves his recruiting. I am skeptical LSU will have anywhere near the success they had last year when they caught lightning in a bottle. I expect Saban and Alabama to return to the top of the West. Georgia can recruit like crazy, but just doesn’t perform up to that talent level. They are a good team, that should be great.
I would focus hand-wringing over recruiting on QB, WR, and LB. The need for an elite QB is a given (though nothing I found here would indicate that, likely because of the small number of rated QBs on a given roster). But if 2020 holds form with 2019, LB roster talent and WR roster talent will more strongly predict win percentage. RB and TE position groups are the weakest predictors. We will see how this holds up in 2020.