Modeling March Madness 2019: Statistical Analysis of the Big Dance

John Reed / USA TODAY Sports

The 2019 NCAA men’s basketball tournament is finally here!

The process of creating a March Madness bracket can be exhausting as most basketball fans end up severely overthinking their picks. UMBC pulling off the impossible last year will cause people to second-guess their decisions even more. The analysts told you that a one-seed losing to a sixteen-seed was impossible — you can’t trust anybody!

When you’re finally complete with your bracket, you’ll probably be irrationally overconfident. Maybe you’ll think (for the fifth straight year) that this is finally “your year.” It’s a great feeling … until the team you picked to win the championship ends up flaming out in the Round of 32. Let’s avoid that.

I devised a model for predicting the results of March Madness just to make bracketology a little bit easier. After analyzing past tournaments, a total of thirty factors were found to be useful for predicting the outcome of the 67 games played during the Big Dance. These include a team’s offensive and defensive efficiency, strength of schedule, performance in the past ten games, shot-selection, ability to play away from home-court, and much more. However, there are a few important aspects of the game which are not accounted for. The most significant limitation of the model is the lack of injuries. Keep this in mind: the results of the model do not factor in injuries at all. Also, you can’t consistently forecast luck. I couldn’t use statistics to tell you if Gordon Hayward’s half-court prayer would go in against Duke in the 2010 National Championship. Don’t bank on any statistical model yielding a perfect prediction of the entire NCAA tournament.

Without further ado, let’s get to the results.

First Four

The model predicts that Belmont and Arizona State have a slight advantage in their respective matchups, while Fairleigh Dickinson over Prairie View A&M should be a safe bet. North Carolina Central is also given the edge over North Dakota State. These four teams will move on to face a higher-seeded opponent in the Round of 64. Do they have a shot at pulling off an upset? Let’s see.

East Region

Top Team: The one-seed Duke Blue Devils enter March Madness as heavy favorites to win it all. Our model predicts that while they will advance to the Final Four, it won’t be easy. Virginia Tech is given a reasonable shot (42.97%) at taking the Blue Devils down in the Sweet Sixteen, while their matchup against Michigan State in the Elite Eight is neck-and-neck. It’s worth mentioning that the model obviously includes Duke’s performance without Zion Williamson. There’s no quantifying the sheer star power which Duke possesses. At the same time, the best player in the nation doesn’t win the tournament all that often. It’s not as guaranteed as some people think it is that Duke will run away with the title.

Potential Disappointment: Louisiana State University ranks as the 29th best team in our model. This ranking is a bit unfair to the Tigers, but it does heavily suggest that the team is being overvalued as the third seed in the region. Yale has a solid chance at knocking them off in the first round and Maryland has an even better shot at taking them down in the second round. Don’t be surprised if the Tigers don’t make it to the Sweet Sixteen.

Sleeper Pick: No. 4 Virginia Tech is projected to advance to the Sweet Sixteen with a matchup against the Blue Devils. They won’t be favored, but their chances in that hypothetical matchup aren’t horrible. The two teams faced off in February where the Hokies won 77-72 while Zion was out with injury. The Hokies shot 30.8% from beyond the arc in the game, though, which is below their 39.4 percent rate over the course of the season (fourth-highest among tournament teams). Duke, meanwhile, hit 33.3% percent of their shots from downtown. Even with Zion back, it isn’t outside of the realm of possibility for the Hokies to get hot shooting the ball and advance to the Elite Eight.

Most Probable First-Round Upsets: No. 12 Liberty over No. 5 Mississippi State (45.60 percent chance), No. 14 Yale over No. 3 LSU (41.25 percent chance),

West Region

Top Team: No. 1 Gonzaga University is the favorite to make it out of the West Regional. While a hypothetical Elite EIght matchup against the Michigan Wolverines is projected to be a coin flip, the Wolverines have a much lower chance at getting to this game in the first place. Outside of the 15-seed Montana Grizzlies, every projected opponent in the regional is given at least a 40 percent chance of topping the Wolverines. According to these forecasts, Gonzaga has an easier road to the Final Four than every other one-seed in their respective regions.

Potential Disappointment: The two-seed Michigan Wolverines have a very plausible chance at reaching the Final Four — it wasn’t easy to pick a potential flop in this region. However, both Florida and Texas Tech are given decent shots at defeating the Wolverines. Again, though, Michigan is a very good team with a solid shot at another Final Four appearance. Their outlook is not as bleak as the aforementioned LSU Tigers.

Sleeper Pick: No. 5 Marquette University has a very similar outlook as the Florida State Seminoles, but the model gives a very slight edge to the Golden Eagles. While the gap isn’t meaningful, Marquette would have a solid chance at topping the Gonzaga Bulldogs in the Sweet Sixteen if they were able to put away the Seminoles. They would then be given a more favorable matchup in the Elite Eight with a prime chance at a Final Four berth.

Most Probable First-Round Upsets: No. 10 Florida over No. 7 Nevada (51.58 percent chance), No. 11 Arizona State over No. 6 Buffalo (42.21 percent chance)

South Region

Top Team: The one-seed Virginia Cavaliers have the best shot at coming out on top of the South region. While their fans won’t look past their first-round matchup, they have good odds at making it to the Sweet Sixteen where they would be projected to play against the Wisconsin Badgers, who they beat 53-46 earlier this season.

Potential Disappointment: The Kansas State Wildcats are projected to drop out of the playing field in the Round of 32, but this model does not account for injuries. K-State star forward Dean Wade has been listed as “doubtful” to play in the tournament. The Wildcats don’t have great chances of making a deep run to begin with, but once you consider the key loss of Wade, there’s a solid chance of them flaming out in the first round against UC Irvine.

Sleeper Pick: The five-seed Wisconsin Badgers aren’t heavy favorites in any of their games, but they’re also given a good chance at winning in every game. The model gives Oregon a decent chance at upsetting Wisconsin in the first round, but it also gives the Badgers a plausible shot at a Final Four run. It’s difficult to project how far into the tournament they’ll make it, but don’t be surprised if they make a run.

Most Probable First-Round Upsets: No. 10 Iowa over No. 7 Cincinnati (49.78 percent chance), No. 12 Oregon over No. 5 Wisconsin (42.03 percent chance), No. 11 Saint Mary’s over No. 6 Villanova (41.69 percent chance),

Midwest Region


Top Team: The five-seed Auburn Tigers are given the best chance of advancing out of the region into the Final Four. The North Carolina Tar Heels are the only one-seed team which is not the favorite within their region. Auburn is projected to face stiff competition throughout their entire run, but the model gives them the edge in every game. The Kentucky Wildcats are forecasted to be the Tigers’ opponent in a tight Elite Eight matchup.

Potential Disappointment: The one-seed North Carolina Tar Heels are the one-seed most likely to not advance to the Final Four. In all of the past nine tournaments, there has been a four seed or higher that won their region. The Midwest might be the region for that trend to continue this year. The Tar Heels are slight favorites over the nine-seed Washington Huskies in the Round of 32 and slight underdogs in an eventual Sweet Sixteen matchup against the aforementioned Auburn Tigers.

Sleeper Pick: The three-seed Houston Cougars are another squad that could bust a few brackets on their way through the Midwest Region. The aforementioned Auburn Tigers are the more obvious sleeper pick, but the Cougars have a legitimate shot at making a run as well. They’re projected to go out in the Sweet Sixteen in a very tight loss to the Kentucky Wildcats. They could very well make it all the way to the Elite Eight. Their path is similar to the Wisconsin Badgers in the sense that they have a legitimate chance in all of their games but they’re not heavy favorites in any of their games either.

Most Probable First-Round Upsets: No. 11 Ohio State over No. 6 Iowa State (44.53 percent chance)

Final Four

The model forecasts Virginia coming out of the tournament as champions of the college basketball world, a fitting redemption story after their humiliating defeat last year. The win probabilities for these games are unsurprisingly very tight, though, especially the National Championship Game. Don’t read too much into the single percentage gap between Duke and Virginia.

2 Replies to “Modeling March Madness 2019: Statistical Analysis of the Big Dance

Leave a Reply

Your email address will not be published.