Hope you listened when we told you to have a fire extinguisher on hand to deal with exploding brackets! WTF? Amid the chaos, here’s our match-by-match look at the potential for upsets in Sunday’s contests.

A few reminders: We study opponents separated by at least five seeds. Our analysis adjusts teams’ fundamental strengths according to how much they statistically resemble overdogs and longshots from previous tournaments and, where appropriate, by style matchups. And we can’t tell you exactly how to bet—it depends on how richly your pool rewards deep upsets and your tolerance for risk.

Second round – Sunday

No. 2 Marquette vs. No. 7 Michigan State

Chance of Upset: 38.9%

Slingshot says this one is going to be close. These teams are separated by about 3 points per 100 possessions in our base power ratings. Marquette is stocked with excellent shooters and outstanding at forcing turnovers, but to repeat our earlier analysis, that doesn’t make the Golden Eagles a mighty behemoth. Even while handily taking care of business against Vermont in the first round, Marquette grabbed just 5 offensive rebounds. And while the Spartans don’t make it a point to push for turnovers or crash the offensive boards themselves, they play a slow-paced style, focused on limiting opponents’ shooting quality and punctuated by the occasional three, which will likely frustrate the Golden Eagles.

This is most evident in our similar game analysis. Of the 10 games closest to this matchup in our database, five produced upsets – the most for any giant vs. killer matchup in the second round. One of these (Gonzaga vs. St. John’s in 2011) was a grotesque mis-seeding. The other four were dead ringers for each other and for what this game will look like if Michigan State wins: Ohio State over Iowa State in 2019, Rhode Island over Creighton in 2017, North Carolina State over Villanova in 2015 and Murray State over Vanderbilt in 2010. All of these favorites were among the top shooting teams in the country. But in those upsets, their long shots held them to about 30% shooting on threes, limited them to taking about a third of their missed shots as offensive rebounds and gave them unusually frequent turnovers.

You can probably think of several matchups similar to these. In 2017, Middle Tennessee went 31-5, tried to play this disciplined underdog style against Butler and held the Bulldogs to just 2 offensive rebounds, but collapsed at the rim. Last year, Michigan State was swept by Duke. It is by no means infallible, just an efficient way to create a stroke where the odds even out. Overall, the 10 most similar Marquette-Michigan State matchups had an average margin of victory that favored the Giants, but by just 3.02 points per 100 possessions. And that’s essentially what our model sees here.

No. 3 Xavier vs. No. 11 Pittsburgh

Chance of Upset: 27.8%

We were tough on Pitt in our first round of the Midwest, but the Panthers played their best 10 minutes of the season by opening a 22-2 lead to start the game against Iowa State, becoming the only 11-seed to advance.

Slingshot still isn’t too impressed. Pittsburgh’s 3-point shooting boosts its offensive efficiency to 114.7 adjusted points per 100 possessions, the 27th-best rate in the country. But the Panthers don’t add many possessions through turnovers or rebounds, and they don’t have a top-100 defense. However, our model does not see Xavier as a particularly strong favorite either. The Musketeers’ strength is also rooted in their shooting efficiency, especially inside. And their first-round win over Kennesaw State was their 10th win of the season by five or fewer points. To use a precise statistical term, that’s a lot. Xavier is a good team. They are not as good as their record, or their seeds.

Boosts Pitt’s chances a bit: Our research shows that when long shots force opponents to play very slowly, as the Panthers do, it especially improves their upset chances against overdogs who aren’t dominant offensive rebounders. Xavier ranks 91st in OR%. Pushing them back down: When sharpshooting killers like Pitt have faced tournament giants focused on defensive rebounding like Xavier, only 19% of the matchups have resulted in upsets.

Put it all together and we end up very close to where we started, with a difference of around 8 points per 100 possessions between these teams.

No. 9 Florida Atlantic vs. No. 16 Fairleigh Dickinson

Chance of Upset: 13.4%

“They limited my touches in the post,” Zach Edey said after his Purdue team took a slingshot in the face from FDU on Friday night.

Yeah, well, they would, wouldn’t they? The Eagles would have quintuple Edey if they knew their pressure plus the spotlight plus what’s in the water in Columbus would so cripple the Boilermakers that Steve Lappas screamed on national TV that neither of them wanted to shoot the ball. Going down to a 16 seed takes a lot of work, and Purdue struggled to the max, shooting just 19.2% from behind the arc, throwing the ball away on one of four possessions, and refusing to attempt a zone while Sean Moore played his , or anyone else’s, game of life. Everything makes us want to say Virginia, come home, all is forgiven. Well, not really.

After this huge upheaval, our statistical model faces two key questions looking forward. Firstly, should it give full credit to FDU for its style of play? Usually, we hesitate to compare 16-seeds to teams much higher in the tournament, because favorites and longshots have to be at least comparable in quality for their giant and killer attributes to matter. Howard was 32nd in the nation this year in offensive rebounding, 34th in 3-point shooting and 56th in forcing turnovers, and what did that buy the Bison? A 28-point blowout to No. 1 Kansas.

But obviously, FDU plays high risk/high reward in ways that can have a meaningful impact on tournament play. Our model credits the Knights for their killer qualities – and for the fact that when teams like FDU have played generic giants like Florida Atlantic, which have no special statistical characteristics that indicate safety from underdogs, 29% of matchups have ended in upsets. And honestly, it’s impressive stuff!

It’s a further question whether FDU was underestimated severely enough that we or anyone else should have seen the Eagles coming. It can be a tough call when an underdog seems to be improving right before our eyes, like Saint Peter’s was last year. But we cannot agree here. The Knights played the fourth weakest schedule in DI and lost 15 games, including contests not particularly close to the likes of Stonehill and Hartford. They gave up a whopping 118.4 adjusted points per 100 possessions in the regular season, third worst in the nation. That’s 10 points worse than Idaho State, 30 worse than Alabama.

FDU plays hard and smart and may have turned a corner in this tournament. But as far as we can tell, the best measure of a team’s fundamental strength is the difference between its season-long offensive and defensive efficiency, adjusted for pace and strength of schedule. That’s what we still use as the basis for the Knights’ rating.

FAU protects the ball. The Owls have a deep rotation, four members of which are shooting 39% from long range. And they’re flying high, too, after their frantic last-second win over Memphis. None of this makes them a particularly strong top dog, but it all signals a better game that Purdue put up. Slingshot doesn’t see them letting lightning strike twice.

Thanks to John Harris, Kevin Hutson, and Liz Bouzarth of Furman University for research assistance.