2024 Women's March Madness predictions
South Carolina is undefeated again. But Caitlin Clark is in their way again.
Thanks to all of you who newly subscribed to Silver Bulletin for my men’s NCAA tournament predictions. Usually Substack traffic is fairly predictable, but the numbers really blew away my expectations and are a reminder that I need to work sports into the mix more often here. They’re also an extremely encouraging sign for similar projects down the line.
To reward all the interest, I’ve also run the numbers for the women’s NCAA tournament. Most of this post is going to be free for everyone, with the exception of two downloadable Excel files at the end that I’ll put behind the paywall. If you’re interested in that level of detail, I figure you’re using these projections for gambling purposes and I don’t mind asking you to pay up. But paid subscribers will also get access to the men’s projections, plus all other paid content, plus the option value of locking in current pricing so long as you remain subscribed. You can sign up for a paid or free subscription right here:
The women’s NCAA basketball tournament didn’t used to be known for big upsets, #16 seed Harvard defeating #1 Stanford in 1998 notwithstanding. (It’s hard to think of anything involving Harvard succeeding as too much of an upset.) The top schools are often utterly dominant; the average margin of victory in last year’s tournament was 17 points, which is down slightly from a peak in 2016-2018 but still notably higher than the men’s tournament, which averaged an 11-point victory margin last year.
The Final Four last year, however, was a spectacular exception. South Carolina came into the tournament undefeated, but then lost in the national semifinal to Iowa and Caitlin Clark, the all-time NCAA Division I scoring leader and the most popular player in college basketball, men or women. Then Iowa lost to LSU, a #3 seed, in the final.
South Carolina is undefeated again this year and overwhelmingly the highest-ranked team in every power rating that you can find. So the question is not whether they’re the favorite but just how much of a favorite. My model, it turns out, is more conservative about this than the consensus. It has the Gamecocks with “only” a 41 percent chance to win the national title, which is still roughly twice any team’s odds in the men’s tournament — but less than Vegas, which implies their chances are more like 55 or 60 percent instead.
So should you literally bet against South Carolina? I don’t know. It’s still tough to win six games in a row in any basketball tournament. And last year’s Gamecocks aren’t the only undefeated upset victim; UConn lost in the national semifinals as an undefeated team in both 2017 and 2018. But any kind of outlier can be hard to model. One assumption my model makes is that teams get credit as they win games throughout the tournament (provided that it’s by more than the implicit point spread that the model calculates). What that means is that if and when South Carolina reaches the Final Four again, they’ll be facing off against teams that are by definition “hot”, having won at least 4 or 5 games in a row. It’s not trivial to get through that gauntlet. Still, we’re talking about a team that’s lost only once in the past two seasons.
The methodology here is similar to the men’s bracket and similar to the former FiveThirtyEight women’s tourney model, which I designed. However, there is a relative paucity of women’s data as compared to men’s data. In 2019, we introduced our own home-cooked version of women’s Elo ratings to make up for that, but I’ve just got to be honest that I just don’t have enough time to pull all together all the data and code to run the Elo numbers still get these projections out to you in a timely fashion. So instead, I’ve used the NCAA’s NET ratings as another system. Thus, the overall mix is as follows:
Massey ratings (2.25x weight)
Sonny Moore ratings (2x weight)
LRMC ratings (1x weight)
NET ratings (1x weight)
Preseason rankings (1.25x weight)
S-Curve seed (1x weight)
The S-Curve seed is supposed to be a team’s overall seed from 1 to 68; however for some reason the NCAA doesn’t publish this data for women as it does for men. It does rank the top four seeds, however — South Carolina, unsurprisingly, is the overall #1 — so I’ve used those, and then broken ties for the other seed lines using NET ratings, the NCAA’s official system.
The reason for the extra weight placed on the Massey and Moore ratings is that these are the only systems that actually include an actual rating (e.g. Iowa is a 92.4) as opposed to merely a ranking (e.g. Iowa is 3rd overall), which allows for more precision, especially when evaluating the top teams. Now, the assumption I use for the ranking systems is that there’s still a fairly big gap between the #1- and #2-rated teams, so this turns out not to affect South Carolina all that much, but it still isn’t ideal.
I also made a small adjustment to the women’s ratings to slightly reduce the travel distance penalty (although note that this is still a fairly big deal since the first two rounds are played on campus sites hosted by the top seeds) and the aforementioned bonus that the model assigns to winners once the tournament is underway. We found that these adjustments had originally been set slightly too high, but I’m not sure we ever incorporated that change into the women’s model as we had for the men. Note that the women’s model does not include an injury adjustment so if you actually are planning to bet these games, you should do some due diligence there.
Onward, then. The NCAA gives the women’s regions rather clunky names; instead of e.g. “East” and “Midwest”, there’s “Albany 1” and “Albany 4”, indicating that the Sweet Sixteen and Elite Eight games involving those teams will be played in Albany, New York and that they involve the #1 and #4 overall seeds, respectively. Similarly, there’s “Portland 2” and “Portland 3” with the regionals set to be played in Oregon. So let’s start with South Carolina’s Albany 1 bracket:
This is about as much chalk as you can get; literally every higher seed is favored in literally every game. Note that #2 Notre Dame is a particularly heavy favorite over #3 Oregon State, in part because of the travel distance penalty the Beavers face, so this probably isn’t the place to go looking for big upsets in your bracket.
Next, Portland 4, where the winner will face the Albany 1 champion in the national semifinal:
This is at least a little more active. The 8-9 and 7-10 games are pretty much pure tossups. And #2 seed Stanford is just barely ranked behind #1 Texas, making for a potentially compelling Elite Eight matchup.
Back across the country to Albany 2. (Seriously, can we pick better names here? The last time I heard so much about Albany was when Sam Bankman-Fried was arrested in the luxury apartment complex named Albany).
This is Caitlin Clark’s region and it’s not a super easy one. The main problem is the presence of last year’s national champion LSU, which was also the top-ranked team in preseason polls. The composite power ratings have LSU as the 5th-best overall team, a tough out for a 3-seed.
Finally, Portland 3:
Here the model is going defiantly against the seeding committee. The various power ratings, including the NET ratings, mostly have UConn as the 2nd best team in the country. And yet they Huskies received only a #3 seed. This is just a bit weird; UConn faced a bunch of injury problems earlier in the season, but they’re coming into the tournament on a 9-game winning streak.
For paid subscribers, you can find spreadsheets with all these numbers and the underlying ratings below.