Let’s start with the question.

A simple yet complicated one. Still, binary.

Who’s going to win the Rose Bowl semifinal of the College Football Playoff? Alabama or Michigan?

This takes a turn when you consider the wild discrepancies and variations of opinion in the matter. It becomes man versus machine in an anomaly that fascinates a veteran of college football analytics and outcome prediction.

Known simply as Josh from The Football Nerds, this lawyer-by-day said he’s never seen anything quite like the disparity of opinion for this Alabama-Michigan showdown.

Put simply, the computer models love — borderline adore — the Wolverines.

The oddsmakers and expert human predictions range between skepticism and straight-up disagreement.

Take The Football Nerds forecast for example. Their model predicts a final score of Michigan 28, Alabama 17 (when rounding to the nearest full number).

A few other computer score picks:

• Michigan 27, Alabama 16 on BCFToys.com. It comes with a confidence percentage of 76.2%
• Michigan 18, Alabama 11 on the Bassett Football model with a 70-80% chance of Michigan winning.
• Michigan 30, Alabama 24 on TeamRankings.com, a site that’s 74.8% accurate picking winners according to thepredictiontracker.com.

Not exactly comforting if you’re an Alabama fan booking flights for Houston and the College Football Playoff championship game.

There’s a but.

A big but.

If all of these models are calling for as much as an 11-point Michigan win, why is Alabama only a 1.5-point underdog in most sports books? The Crimson Tide opened as a 2.5-point underdog but was bet down to 1.5 points relatively quickly.

Also of note, ESPN pointed out this is the first time in 77 non-conference games Alabama is an underdog. That’s a streak that dates all the way back to the 2008 season opener when Clemson was favored over Nick Saban’s second Crimson Tide team, one that ultimately crushed the Tigers 34-10 in what’s viewed as the first statement of the Saban dynasty.

Anyway, a heavy majority of national writers who’ve made public predictions favor Alabama over Clemson. It’s Alabama by an 8-2 margin when compiling viewpoints from Sports Illustrated, Fox Sports, The USA Today Network, Athlon, Bleacher Report and The Sporting News.

Alabama even got two of the five nods from The Detroit Free Press.

So … what gives?

Josh from The Football Nerds said several factors come into play when explaining why the computer models swing so heavily toward Michigan.

It begins with the utterly dominant nature of the Wolverines’ early-season performance against a schedule considerably weaker than Alabama’s. It feasted on non-conference teams East Carolina, UNLV and Bowling Green, then the weak underbelly of the Big Ten before facing its first ranked team on Nov. 11 at Penn State. The computers love consistency, even in a vacuum.

The models viewed Michigan as almost a perfect team through a 31-24 win at Maryland with The Football Nerds model projecting it would beat 130 of the FBS teams before the relatively close game at Maryland.

Alabama, meanwhile, had a less linear path to Pasadena. Obviously, it had the Week 2 loss to fellow semifinalist Texas but there were other bruises to their algorithmic resumé.

“With Alabama’s case, you have a team with some real clunkers,” Josh said. “On the field, Alabama had an absolute dud against South Florida, an absolute dud against Arkansas, which I think a lot of people outside of Alabama overlook how had that was statistically.”

He said those two games had the performance of a team on the caliber of Nebraska — a team ranked between 60 and 70 nationally. The Texas loss and Auburn escape were closer to a top 30 to 40 team.

Then on the high end of the spectrum, games against LSU and Kentucky — had they been their average performance, “they’d be the highest-rated team in the nation,” Josh said.

But when you average it out, Alabama has a few computer-detected warts that Michigan avoided. The Wolverines have close to a clean sheet, albeit against a schedule that doesn’t compare with the difficulty of Alabama’s.

“When you look at scoring predictions like in our model,” Josh said, “It creates a unique model for each team based on how they did on the field. Well, Alabama has a few games where they struggled to score. So when they project them against the top defense and it sees that yardage projection and production projection really low, it starts to think this will look more like the team they were against South Florida.”

Of course, the models don’t factor in the fact quarterback Jalen Milroe was benched for that 17-3 win at South Florida — one that featured an offense that looked nothing like the one that beat then-No. 1 Georgia in the SEC championship.

It’s also worth noting most of these pro-Michigan models, The Football Nerds included factor only action on the field. Others, like ESPN’s Football Power Index (FPI), account for talent level by factoring recruiting rankings. There Alabama is No. 5 whereas the Crimson Tide falls to 11th in both the FEI and College Football Nerds’ power rating that don’t include recruiting.

Ultimately, Josh from The Football Nerds said everyone should take the computer models with a large grain of salt. He said it’s telling the point spread is considerably closer than any computer model would project. Playing weak schedules can trick models into assuming blowouts against low-level opponents will carry over to all opponents.

“There’s something to be said that there are chinks in the armor when you look at Michigan that lend you to think that Michigan probably isn’t the team on paper,” Football Nerd Josh said.

Case in point, Michigan has one of the lowest consistency ratings for hitting the score projections The Football Nerds’ model produces. Alabama’s low too given the wild fluctuations from early season to late.

“Without a doubt, there’s an argument that all computer models are either overvaluing Michigan or undervaluing Alabama,” Josh said. “Almost certainly, both are true. Michigan is not as dominant as they appear on the computer and Alabama is a much better team than they appear on the computers. And you end up with this weird result where the Vegas line is tremendously closer than any power rating or metric is wanting to suggest.”

These models also don’t account for the fact the Big Ten was more top-heavy than the SEC or the fact it’s considerably harder to compare teams from different leagues without crossover opponents. These models are far more accurate in predicting conference games where there are far more data points for comparison.

Bottom line: Who knows?

Probably not the computers, at least not to the level they project Michigan to handle Alabama as the sun sets on the San Gabriel Mountains on Jan. 1 in Pasadena.

Maybe that question wasn’t that simple after all.

Michael Casagrande is a reporter for the Alabama Media Group. Follow him on Twitter @ByCasagrande or on Facebook.