Simulating the 2024 NCAA Men’s Basketball Tournament
An Objective Look at the Odds and The Final Four
March Madness has come and gone. Mad it was. April brings us the reality of the Final Four, and the hopes of finding a perfect bracket are left in the dust. Is your bracket totally busted? Even the public top scoring brackets on major outlets appear to be (Mine certainly is, even at the 99.7% percentile on ESPN), so the answer is almost definitely yes. That’s no surprise as upsets have been on the rise, with the NCAA writing that the average is 8.5 per year covering the past 38 seasons.
There’s a major catch: The most upsets that have previously happened in a single bracket is 14, which happened twice. In 2021 and 2022. The transfer portal has definitively changed the landscape and the types of lineups most contending teams are able to put on the court. On top of that, more veteran talent is available at the top of player rotations with the added COVID year of eligibility. 2024 (So far) has seen 15 lower seeds win, already setting a new record. It’s likely the NCAA’s advice against choosing 15 or more upsets for your bracket didn’t account for this recent trend.
Here stands an unlikely group that contains what I like to call a pseudo-Cinderella: 11-seed N.C. State still resides in a power conference. N.C. State’s listed roster contains 4 Seniors and 6 Juniors, a dearth of veteran talent. KenPom.com has their D-1 experience average listed at 3.05, good for 14th in the nation. The other 3 teams that are still alive are all ranked in the 60s. The Wolfpack had the 8th most experienced lineup in this field, which may lead some to believe they were under-seeded. The objective seeding of the tournament seen here shows the opposite, with N.C. State as a 12-seed. Veteran teams don’t always win; 7 other tournament teams who have been knocked out were ranked even higher for D-1 experience, with TCU ranked as high as 3rd.
Of UConn, Purdue, and Alabama, it’s UConn’s talent that leads the pack as undeniable, with the Huskies sitting atop most metric leaderboards out there. Could they still be vulnerable tonight, playing against a slightly more experienced lineup? The Huskies have 3 players listed as Seniors and just one labeled a Junior. Alabama looks like they’ll have 5 seniors in the rotation, if the news that Wrightsell is healthy enough to play is true (They also have one Junior on the roster). We’ll soon find out how much experience matters on this biggest of stages.
The Field so Far
A lot of fans and pundits alike hemming and hawing over their bracket opinions and pre-conceived bias. Some are leaning hard into confirmation bias, “knowing” all along that certain conferences were over- or underrated. Some are even buying into conspiratorial fantasies about teams controlling the metrics to their advantage. What’s true is that teams who play stronger schedules and perform well against them always bubble to the top of stat leaderboards. Power conference schools lining up non-conference schedules full of cupcakes is hardly anything new, and it happened in the 80s and 90s as much as it does today. If any ‘gaming’ of the system is taking place, it’s well before the season even starts — blaming the metrics themselves, like the NET, is like blaming 2+2 for equaling 4.
There’s a tendency to want to believe your conference is the best, that your team is better than all the rest. Going in with that mindset leads a lot of people to cherry-pick information and formulate entirely subjective numbers that are every bit as silly as what the AP voters come up with during the season. This is why I prefer full fields of data wherever possible to eliminate bias towards individual teams.
It’s true that in KenPom’s AdjEM ratings the Big XII got largely pounded, taking several early losses and watching a roster-neutered version of Houston slog ahead. It’s also true that even after that, the Big XII is still the top conference. The Big East and SEC took their lumps, with the SEC having a terrible first round, and the results of the NIT / CBI also playing their part. The ACC has closed a wide gap after several of their teams went on deep runs, but none of the power conference rankings have changed from the start of the tournament to today.
The Field’s Odds
I am not concerned with the myriad opinions or entertaining any of the bias mentioned above. This is a purely objective look at predictive odds and finding out how likely, or unlikely this Final Four was.
As Ken Pomeroy has pointed out in his writing, efficiency remains a historically better predictor of outcomes than judging teams by their Q1 wins and losses. Plenty of teams with solid Q1 wins offset that merit badge by the times they were blown out, and plenty of teams with a heap of Q1 losses are clearly better performers to date (Hi, Alabama!). Picking individual-win-darlings for a resume may swoon then flawed NCAA Committee, but doesn’t do much to predict how teams will perform come bracket gametime. This is where a simulator projecting teams by their efficiency scores comes into play.
After running 1000 simulations of the tournament field based on team efficiency scores, as well as offensive and defensive scores, the results have been largely accurate. The ‘chalk’ bracket, where the statistically-higher ranked team always advances, is still in the 94.4th percentile of ESPN’s Bracket Challenge —not bad for chalk. Changing the seed of a team doesn’t seem to have much impact upon the rates at which they make the Final Four, either. To investigate this further, I took my previously created objectively-seeded bracket and used that as control, also running 1000 simulations.
Here are those two resulting fields of odds:
The changes in odds are not drastic or sweeping. Sorry to disappoint those who may have been searching for a smoking gun. A statistically balanced seeding of the bracket, if anything, may block some paths of teams who were previously under-seeded. Let’s highlight New Mexico, whose resume received the biggest slap from the committee as an 11-seed that was statistically effecient enough to be a 6-seed. Their chances to win it all actually went down as a 6-seed, as the Lobos won 7 of 1000 times as an 11-seed and just 3/1000 as a 6. Getting past #11 James Madison, and then the high win-percentage teams of #3 Alabama, and #2 Arizona, ended up being a very tough path.
Auburn was certainly a beneficiary of their objective boost to a 1-seed. That opening game with a 16-seed really is a huge boon for future win percentages.
It may be easy to criticize or even get lost in anger over what’s really happened in the tournament when looking at these numbers. Sure, Auburn lost early. That now looks like the worst Round of 64 loss that any team endured. It doesn’t mean Auburn was overrated, it just shows that any seed lower than a 2 is inherently vulnerable, no matter how good they are (Auburn is still ranked #4 on KenPom following the tournament action to date). It doesn’t discredit what they’ve achieved in their body of work; the Tigers were simply the big ol’ disappointment this year.
Similarly, the Big XII and Houston haters out there might be in fits over seeing which team was deemed most likely to win it all. This is where it’s important to remember that the statistics used to simulate the tournament are based on team performance, not individual performances. Team numbers have been far more valuable for predictions (Not to mention easier to assemble), but unfortunately have the drawback of not accounting for injuries. Houston’s injuries were well-documented and while they were deserving of a 1-seed, they may have been the easiest #1 to pick against, depending on your methodology.
The Final Four
With all perfect brackets long gone, how likely was the field of Connecticut, Alabama, Purdue, and N.C. State? For starters, there was only a 1.8% chance of the simulation correctly predicting 3 of the 4 teams. Out of the 1000 simulations run, none correctly predicted this particular quartet. This is largely because of the …no, not Elephant, but the Wolfpack in the room, who made the Final Four just 1.1% of the time. They didn’t make the Championship game in any of those simulations, which may not bode well for the matchup against Purdue. It’s worth noticing that N.C. State did win it all 1 time in the objective bracket simulations as a 12-seed, while making the final game 3/1000 times. As always, anything can happen.
So how likely was this Final Four? Well, let’s look at the odds of each team getting here:
- Connecticut: 37.2%
- Alabama: 11.8%
- Purdue: 38.7%
- N.C. State: 1.1%
Combine these odds — P(A∩B∩C∩D) — and we see that there was a 0.01868653% chance of these 4 teams getting here. If you played the tournament 10,000 times, you’d see this Final Four array roughly twice. If these were the teams you picked on your bracket, you were seriously lucky! Still, this result is very probable when compared to the odds of getting a perfect bracket.
Going by tonight’s odds, we’re seeing numbers such as UConn being 78% likely to advance and Purdue winning 82% of the time. Combining those odds, there’s a 69.36% chance we’ll have a Championship showdown between two 1-seeds. That’s the most likely finals scenario that didn’t include Houston. Is there one more upset remaining to push the new record to 16? Or even 17?! Personally, I hope we’ll have at least one more big shock.
Whoever or whatever you’re rooting for tonight, I hope you’ll enjoy the games.
If you enjoyed this piece, please consider sharing it with a friend, reposting wherever you like to spend your time online, or by giving this a bunch of claps. Have a question or want to see me cover something in particular? Drop a comment or reach out and let me know!
