Statistically Speaking - Michigan v. Utah

Measuring the ‘Harbaugh Effect’

While all this hype for a season opener in Salt Lake City is a nice break from trying to get excited for an FCS opponent, Michigan at Utah has brought a tiresome storyline along with it. Harbaugh is the white knight, riding in to save Michigan and slay the dragon, who this week wears a Utah Utes uniform. It hardly seems as though they even notice the Utes are playing, and many expect the Wolverines to roll right over the 'cupcake' Utes.

We can use data to put an objective face on the idea of a savior coach. The data includes any coach of a major college program with previous NFL or major conference head coaching experience who was brought in after the previous coach was either fired or resigned under pressure. I came up with a solid list, 16 hires that can be fairly described as ‘saviors’ for flagging or failing programs.

In-conference scoring differential gives a fair sense of how these coaches performed in their first year, compared to the final year before the previous coach was let go. Some coaches' teams get much worse while others engineer major positive change.

Turns out, it’s easier to improve a team that was really bad rather than one that was only sort of bad. That doesn't bode particularly well for Michigan; this data projects them to improve by about 2.2 points. In 2014, UM trailed the Utes by 11.51 points in overall efficiency, so this projected improvement doesn’t promise to get them within striking distance of a win.

Another way to look at this data is F/+ rank. Rather than being record- or points-dependent, this metric is looking at efficiency and explosiveness on offense and defense. There is less data here, but it has a stronger correlation and is well suited to a predictive model. I am not convinced that there is a causal relationship, but the correlation is undeniable.*

Teams tend to improve the most in years 1 and 2 and then start to level off as the system is installed and the team reaches maturity. It’s reasonable for the Wolverines to expect to improve by twenty or so in these rankings, which would put them pretty close to where the Utes were last year.

One question this data can't answer is how long this first season improvement takes. One would expect some improvement to happen right away, as the new coaches' better scheme and preparation would have an immediate effect. Additional improvements will probably take time, as players learn the system and the coaches' superior development skills come into play.

The ‘Harbaugh Effect’ is real. Teams which hire proven coaches after down years tend to improve significantly from the year before. The issue in this game is not whether or not Michigan will improve, but how much and how quickly. They were thoroughly overmatched in 2014, and getting a win in Salt Lake City looks to be a tough row to hoe.

Advanced Metrics

The Utes beat the Wolverines handily in 2014, allowing no offensive touchdowns and scoring efficiently, despite being stymied in the running game (2.2 yards per carry). The Utes also were an average of 11.51 points better than the Wolverines last season, according to ESPN’s efficiency metrics. That’s a lot of ground to make up without personnel improvements.

This early in the season, we don’t have a lot of advanced metrics to work with. They tend to get more accurate as the season goes on, and the truly rigorous approach would be to not release any numbers until at least week 7. ESPN has released complete FPI projections, and FootballOutsiders has released half of their projected stats, S&P+ (FEI, the other component of F/+, tends to favor the Utes compared to S&P+). Both measures project the Utes and Wolverines to be nearly evenly matched this year, with a slight advantage for the Wolverines.

While preseason projections are usually pretty reliable, I don’t place a lot of value on them here. In Utah’s case, the addition of FEI numbers always helps bring the Utes up, and the use of weighted four and five year performance averages gives too little credit to the steady upward trend in talent and performance since graduating to the PAC 12. Michigan’s roster is in a state of disarray following graduations and transfers of several key players.

Michigan was a below-average team last year, and they took heavy hits in just about every area. On offense, the Wolverines lose their starting center, starting QB, and best wide receiver. On defense, they let both starting defensive ends go, the starting MLB (and leading tackler), three contributing DBs, and their starting NT is missing the season with an injury.  UM is also completely replacing its special teams, having lost its kicker and punter to graduation, as well as the starting return man to a transfer (one of at least eight UM transfers).

Neither Iowa transfer Jake Rudock nor Shane Morris have demonstrated the ability to stretch a defense with their arm.  The running game should be a bright spot; the offensive line returns most of its talent, although UM is replacing its center, which may interfere with more complex blocking schemes. There are a lot of running backs to choose from and none has emerged as a true feature back, but all have talent. This set up suggests a Michigan team that will make its hay running into the teeth of opposing defenses, with the vast majority of plays developing within a few yards of the line of scrimmage.

Michigan’s defense was very good last year, and they should stay that way with a new coaching staff. The front seven will take a step back, but likely not a large one. The secondary may improve somewhat with the addition of a healthy Jabrill Peppers. This defense will be salty, and points will be hard to come by for any offense they face.

A major advantage for the Utes is on special teams (get used to reading that). You never know what you’ll get with a specialist, but if either the place kicker or the punter for the Wolverines struggles in what promises to be a field position grind of a game, expect the Utes to capitalize.

Vegas & Massey

The line opened at -3 for the Utes, which seemed quite low at the time. The majority of bettors agree, as the line has continually shifted and is closing in on a full touchdown advantage, currently resting at -4.5. I find this line conservative, given last year’s result and the fact that game will be played in Salt Lake City.

Massey, a ratings aggregator that measures forty two different ratings and uses a mathematical formula to come up with a consensus, places the Utes at 28th and the Wolverines at 50th.

Stat-Head Pick

At its core, the UM offense is one that is installing a new system, training a new quarterback, converting a guard to replace a center, and trying to figure out what options it has at the perimeter. The defense should be good again, but will feel the loss of experience, at least at first. Harbaugh can transform this team, but it’s going to take more than one offseason.

This isn’t a game that will be out of reach for the Wolverines like it was last year; they will be a big play or two away from coming away with the W. The Utes are going to be more effective than the Wolverines on defense and special teams, and the resulting field position advantage promises better results on offense. I’m giving roughly a touchdown to Michigan to compensate for the Harbaugh Effect, but it’s not enough to make up the existing difference between these two teams, especially considering the Utes’ projected improvement and the flip in home field advantage. In this environment, it’s hard to see how Michigan’s offense, replacing important pieces and still in its infancy, will get into the end zone.

Utes 20, Wolverines 9.

* You would expect top tier teams with strong recruiting to have down years only infrequently, and bounce back quickly, regardless of the coaching staff. If teams don’t improve, their coaches are fired or replaced, leaving only the best coaches in the data (Nick Saban and Brian Kelly are the only coaches with more than 3 years). The data set is pretty small and I didn’t use a control sample to contrast with. This data serves gives a reasonably accurate profile of what to expect, but not necessarily why you should expect it.