2023 NCAA Tournament Upset Picks and Favorites

The next week contains 4 of the 6 best sports days of the year, the first 2 rounds of the NCAA tournament (the other 2 are the Super Bowl and Sunday at the Masters). I always like to celebrate by firing up a model and writing about my 2023 NCAA Tournament upset picks and favorites.

The goal of this article is to help you plan your 2023 March Madness strategy. From likely first round upsets to the most probable champion, it’s best to go in informed. All of the results presented here are the result of our own model. The model is designed to take in all available information, place emphasis on more recent performance, and to appropriately incorporate strength of schedule. You can read an in depth description of the model at the end of this article.

We’ll begin with overall team ratings. Then, we’ll look at our estimated title favorites. Finally, we’ll pick a few upsets in the early rounds to be on the lookout for. To see how we did last year, check out the article here!

To receive email updates when new articles are posted, use the subscription form below!

March Madness Best Teams

Our model rates each team’s quality by assigning them a number. This number represents “points better than an average team”. The title contenders typically have ratings near 20 while the worst tournament teams are near 0.

The difference between two team’s ratings is designed to be a prediction for the margin of victory were they to play. For example, a team rated 15 should be expected to beat a team rated 10 by 5 points.

The table below shows the ratings for the top 16 teams in the field!

Team	Rating	Team	Rating
Alabama	23.1	Arizona	18.6
Houston	21.6	Purdue	18.3
UCLA	21.1	Baylor	17.7
Texas	20.8	Marquette	17.3
Gonzaga	20.3	St. Mary’s	16.9
Tennessee	20.0	Creighton	16.5
Kansas	19.9	Kansas St.	16.3
UConn	19.6	Xavier	16.2

This list shouldn’t be terribly surprising, though perhaps the order that the teams show up is. Most notably, Kansas is at 7th and Purdue is at 10th overall even though they’re 1 seeds. Our model thinks the 2 seeds UCLA and Texas should have replaced them.

Our model also thinks that:

• Marquette is over seeded
• Tennessee is dramatically under seeded (lots of models think this, check out what KenPom thinks)
• Gonzaga is under seeded
• UConn is under seeded

But how do these ratings transform into title probabilities?

2023 NCAA Tournament Favorites

It isn’t enough to be the best team in order to win the title. Another important factor is how difficult their region is. It’s possible for the best team to not have the best odds because they got a bad draw.

While most analysts look at the entirety of a region to determine its difficulty, determining an easy region is actually much simpler. The factor which most determines a region’s difficulty is the quality of the single strongest opponent you’ll have to play. For example, when evaluating the region difficulty for a 1 seed, typically all we need to do is look at how good the 2 seed is. For everyone else, the strength of the 1 seed is really all that matters.

While this provides good intuition, we also wanted to run the numbers. To determine the NCAA Tournament favorites, we computed the probability that each team wins the title. This was accomplished by using our model to determine how good every team is then feeding this into 100,000 Monte Carlo simulations. The top 10 title favorites along with their odds are listed below.

Team	Odds
Alabama	22.9%
Houston	14.9%
UCLA	10%
Texas	9.4%
Tennessee	7.5%
Gonzaga	6.4%
Kansas	5.8%
UConn	4.2%
Arizona	3.6%
Purdue	3.3%

After this, Marquette, Baylor, and Kansas St. all have above a 1% chance to win while everyone else falls below this threshold.

Notice that the probabilities follow roughly in the same order as the team rankings. This shouldn’t be terribly surprising but it does mean that the region you’re in doesn’t matter as much this year.

2023 NCAA Tournament Upset Picks

We now want to give some upset picks. To do this, we’ll go through seeds 10-15 and look at who is most likely to win in the first round and make it to the sweet 16. According to our model, none of the 16 seeds stand above a 1% chance.

In the following, the numbers in the parentheses are a team’s rating.

The 14 and 15 Seeds

The four 15 seeds are UNC Asheville (1.4), Vermont (3.6), Princeton (2.7), and Colgate (3.8). These are all pretty common teams that show up in this range. Though Colgate is the best team by rating, Vermont actually has the best odds to win in the first round – just over 5%. None of the teams have measurable probabilities of making the sweet 16.

The 14 seeds are Grand Canyon (3.4), Montana St. (3.4), UCSB (3.2), and Kennesaw St. (2.8). Though none of these teams are demonstrably better than Colgate, they all have better odds because their opponents are much weaker.

Montana St. has about a 6.6% chance at beating Kansas St. and roughly 1% to make the Sweet 16. Kennesaw St. similarly has about a 6% chance over Xavier. The others are around 3%.

My best guess is that there are no upsets in this seed range. But in the next range…

The 12 and 13 Seeds

This is usually where most of the fun upsets come from. This year, the 13 seeds are Iona (8.2), Louisiana (4.7), Furman (5.8), and Kent St. (7.6). Our model thinks there is a 45% chance that at least one of these teams win.

Though Iona has the best rating, they unfortunately got matched up against UConn and only advance 8% of the time. Louisiana also only advances 3.5% of the time being the worst of the 13 seeds. However, Furman beats Virginia 20% of the time and Kent St. beats Indiana a whopping 23% of the time!

The 12 seeds get even more interesting. Usually the 12 seeds are the most popular pick for upsets. This year we have Drake (9.3), College of Charleston (9.8), VCU (10.3) and Oral Roberts (11.4). Our model says at least one of these teams wins 76% of the time.

Each of these teams have between a 20 and 33% chance at winning their first round matchup. VCU has the hardest path, beating St. Marys only 21% of the time and making the Sweet 16 3.8% of the time.

College of Charleston beats San Diego St. 25% of the time – if you want to get cute you could pick a Charleston v. Furman R32 matchup!

Oral Roberts wins 1/3 of the time in the first round against Duke but only makes the Sweet 16 5% of the time. This is because they get buzz-sawed by the model-loved Tennessee in the second round.

Finally, Drake is the best bet in this group. They beat Miami in 35% of simulations and make it to the Sweet 16 in roughly 1/8 brackets.

10 and 11 Seeds

The 11 seeds this year are Nevada/Arizona St. (~9 each), Pitt/Mississippi St. (roughly ~9 each), Providence (11.4) and NC State (11.0). Compared to the 12 seeds, these ratings are unimpressive this year. In fact, their odds of advancing are comparable to the 12 seeds.

• Providence beats Kentucky 36% of the time and makes the Sweet 16 11%
• NC State beats Creighton in 26% of simulations and makes the Sweet 16 in 6%
• The winner of Nevada/ASU beats TCU 24% of the time and makes the Sweet 16 hardly ever because the model is very high on Gonzaga
• The winner of Pitt/Mississippi Iowa St advanced only 21% of the time and makes the Sweet 16 hardly ever

Finally, the 10 seeds are Penn St (10.8), Boise St. (12.8), USC (12.5), and Utah St. (14). This is where things start getting really fun.

• Utah St. is actually favored in the first round against Missouri – winning a whopping 64% of matchups. They make the Sweet 16 20% of the time and the Elite 8 almost 8%. These are great odds for a 10 seed
• USC is a tossup with Michigan St. and makes the Sweet 16 in 15% of runs.
• Boise St. is a tossup with Northwestern but only makes the Sweet 16 8% of the time because they have to play UCLA.
• Penn St. is the odd one out beating Texas A&M only 20% of the time.

Other Notes

There are a few other oddities we wanted to note rapid fire in bullet point form:

• Purdue loses in the second round in 1/3 simulations and makes the Elite 8 only 32% of the time.
• Both UCLA and Texas make the Final Four nearly a third of the time – roughly as likely as Marquette losing in the second round.
• Xavier has the toughest road as a 3 seed, making the Elite 8 under 20% of the time.
• Tennessee is over 8 times as likely to make the elite 8 as Virginia is: 47% compared to 6%.

Model Notes

Our model is built using the data from this Kaggle competition. Our model uses Bayesian estimation to estimate each team’s overall quality (the ratings we’ve provided). The notes on this estimation problem include:

• The prior is given by a system average of all other models available in the data. That means our model takes into account Massey ratings, KenPom data, etc. and uses it to inform our own.
• Our model places higher emphasis (3 times as much) on more recent games than on games at the beginning of the season.
• We also use a form of bootstrapping. For example, a team that has a good night and steals a victory from the #1 team in the country will get a huge ratings boost. Bootstrapping uses random sampling of the data so as not to overemphasize the effects of any individual game.

To receive email updates when new articles are posted, use the subscription form below!