Feng: How many games will UM win in Year 1 of Harbaugh?

Ed Feng
Special to The Detroit News
Jim Harbaugh

Michigan has a rich tradition of football excellence. This visual shows the program's performance by both wins and computer ratings over the past 30 years.

This shows Michigan's wins, and computer ratings by The Power Rank, over the past 30 years.

The computer ratings in the bottom panel result from The Power Rank algorithm. These calculations take margin of victory in games and adjusts for strength of schedule. The rating gives an expected margin of victory against a bowl subdivision team for each season.

In 1984, the year before the 30 seasons shown in the visual, Michigan QB Jim Harbaugh broke his arm. Bo Schembechler's team finished the season 36th by my computer calculations.

For the next 22 seasons, Michigan finished in the top 25, a remarkable streak of consistency. This era included the tenure of three coaches and the national championship season in 1997 under Lloyd Carr.

This streak ended in 2007, Carr's final season. Since then, Michigan football has failed to consistently rank among college football's best. Can the return of Harbaugh as coach bring the program back to its past greatness?

This article projects a win total for Michigan in 2015 based on preseason analytics. Turnovers in 2014 play a big role in this prediction. Let me explain.

The randomness of turnovers

Turnovers fueled Michigan's failures last season. Senior QB Devin Garnder struggled, and the offense gave away the ball 26 times. The defense played well but could only generate 10 takeaways.

It's tough for even elite teams to survive a turnover margin of minus-16. Turnovers can have an enormous impact on the score, and Michigan's poor performance in this area led to its rank of 57th in my computer calculations.

Fortunately for Michigan, turnover margin from last season has no ability to predict turnover margin this season. Randomness plays a large role in turnovers. Sometimes a tipped pass falls harmlessly to the ground. Other times, it land in the hands of a defensive back who returns it for a touchdown.

Michigan most likely had a flukish year with turnovers in 2014, which suggests they were better than the 57th team in the nation. Their turnover margin should regress to the mean in 2015, which implies a turnover margin closer to zero.

Regression to the mean makes no guarantees, though. During each of the three seasons under coach Rich Rodriguez, Michigan had a turnover margin worse than minus-10.

Projecting a win total

At The Power Rank, my preseason college football rankings come from a model that considers a team's performance the past four years, turnovers and returning starters. Despite its simplicity, the model has predicted the game winner in over 70 percent of games since the 2005 season.

Michigan is 34th in these preseason rankings. The Wolverines get a big bump from their 57th ranking of last season because the model recognizes how much the randomness of turnovers hurt them.

The preseason model also gives a win probability for Michigan in each of its games, which gives this breakdown of the schedule:

Almost certain wins: UNLV (97.3 percent).

Very likely wins: Oregon State (82.2), BYU (67.2), at Maryland (75.2), Northwestern (77.7), Rutgers (83.3), at Indiana (74.0).

Toss-up games: at Utah (40.4), Michigan State (45.5), at Minnesota (53.9), at Penn State (46.1).

Unlikely wins: Ohio State (23.5).

Adding up the win probability in these games gives a projected win total of 7.7.

Jabrill Peppers

Teams that exceed the expectations of this preseason model get extraordinary play from a superstar. The most likely candidate for Michigan is Jabrill Peppers, the highly-recruited defensive back that played sparingly due to injury last season. His speed and size could take the defense from very good to elite.

Another possible route to exceeding this projected win total is special teams. Harbaugh hired specials teams coach John Baxter, who had an exceptional track record at Fresno State and USC. A late-game blocked kick could earn an extra win for Michigan.

How does Michigan go under their expected win total? Graduate transfer QB Jake Rudock gets hurt and the quarterback position degenerates into a canon whipping footballs at the ankles of receivers. Another bad year with turnovers would also hurt Michigan.

The Harbaugh effect

The preseason model is simple but accurate. However, it makes no adjustments for Harbaugh, a coach who engineered remarkable turnarounds with Stanford and the San Francisco 49ers.

This visual for Stanford football has the same format as the Michigan team history above and shows the quick turnaround in four short years.

The Jim Harbaugh years at Stanford are highlighted in dark red.

It wasn't always Roses in Palo Alto. In Year 3 of the Harbaugh regime, Stanford suffered a gut-wrenching loss to rival Cal when Andrew Luck threw a late-game interception.

In the same way, Harbaugh's tenure at Michigan will not always be smooth. The honeymoon period with the Michigan fan base ends when they travel to Utah for a tough opening game.

However, Harbaugh most likely brings a culture of toughness to the Michigan program. You might not see it against Utah, but you may by the time Ohio State visits the Big House in November.

While my preseason model does not consider the Harbaugh factor, you can make your own subjective adjustments to the predicted win total of 7.7.

Ed Feng has a Ph.D. in chemical engineering from Stanford and runs the sports analytics site The Power Rank. Email Ed Feng here.