We pulled 25 years of NCAA tournament data — over 1,000 games from 2002 to 2025 — and ran every statistical angle we could think of to find what actually predicts upsets.

The short version of the findings is upsets are driven by surging underdogs with elite defense facing overseeded favorites who are fading, inexperienced, or undersized — while popular narratives like tempo, three-point shooting, and free throw percentage have zero statistical significance across 25 years of tournament data. (Click here to jump to the full methodology)

Then we applied that model to every first-round matchup in the 2026 bracket and scored it out of 100 based on our methodology. Here's what it's telling us.

Upset Tiers

Tier 1: Highest Upset Probability (50+)

1. (8) Villanova vs (9) Utah StateWest Region · Score: 62.4

This is the closest matchup in the entire bracket by the numbers. These two teams are separated by just 3 spots in adjusted efficiency margin. Utah State fits our Cinderella profile with a top-45 defense and solid ball security, and they're the more experienced team (rank 19 vs. Villanova's 78). The Aggies also rank 14th nationally in effective field goal percentage. This is a coin flip that Vegas will probably agree with, and we'd lean Utah State.

2. (8) Clemson vs (9) IowaSouth Region · Score: 62.0

Iowa's adjusted efficiency margin rank is actually better than Clemson's — 25th vs. 36th nationally. The Hawkeyes rank 16th in eFG% and fit the Cinderella mold with a top-31 defense. The one flag? Iowa has been fading a bit since February (dropped from 18th to 26th in the Barttorvik rankings), while Clemson has held steady. Still, the raw talent gap favors the "underdog" here.

3. (6) North Carolina vs (11) VCUSouth Region · Score: 61.2

This is the game the model is most excited about relative to the seed line. VCU has been on a tear — they've climbed 21 spots in the national rankings since February 1st (69th to 48th). UNC, meanwhile, is one of the least experienced teams in the tournament at rank 220. VCU fits the Cinderella profile, they're more experienced (134 vs. 220), and they're riding serious momentum into March.

4. (6) Louisville vs (11) South FloridaEast Region · Score: 59.4

South Florida has surged 16 spots since February and matches the Cinderella profile — top-40 defense plus top-77 turnover discipline. They also have a significant edge in turnover rate (rank 77 vs. Louisville's 159). Louisville's defense is solid (25th), so this isn't a vulnerability story as much as it is a "the underdog is legitimately good and getting better" story. USF's offensive rebounding rate (6th nationally) could be the X-factor that extends possessions and keeps this close.

5. (6) BYU vs (11) NC StateWest Region · Score: 56.4

NC State has the 9th-best turnover discipline in the country, top-10 three-point shooting, and they're significantly more experienced than BYU (rank 38 vs. 137). The teams are only 11 spots apart in efficiency. The one thing holding NC State's score back: their momentum has been slightly negative, slipping 11 spots since February. But the profile match is strong.

6. (5) St. John's vs (12) Northern IowaEast Region · Score: 52.9

Here's your classic 5-vs-12 upset candidate and it's a surprise given St. John's took the regular and tournament Big East championships. Northern Iowa has climbed 28 spots in the rankings since February 1st — from 98th to 70th. They fit the Cinderella profile perfectly: top-24 defense, top-44 turnover discipline. Historically, 5-vs-12 matchups flip 35% of the time anyway, and Northern Iowa is playing their best basketball of the season right now. St. John's has the better team on paper (AdjEM gap of 55), but momentum matters in March, and the Panthers have it.

7. (7) Kentucky vs (10) Santa ClaraMidwest Region · Score: 52.9

Only 7 spots separate these teams in efficiency. Santa Clara has the better offense (rank 23 vs. Kentucky's 39) and ranks 19th nationally in offensive rebounding. Kentucky has the defensive edge, but this is the kind of 7-vs-10 matchup that flips 39% of the time historically when the gap is this small.

Tier 2: The Sleepers (Score 41–50)

These games are in the "don't be shocked" range — not the most likely upsets, but ones where the data says the underdog has a real shot.

Kansas (4) vs California Baptist (13)East · Score: 41.4

Kansas has been one of the biggest fallers in the national rankings, dropping 9 spots since February. Meanwhile Cal Baptist has climbed 20 spots. That +29 momentum differential is one of the biggest in the bracket. Kansas is still the better team, but they're a 4-seed playing like a 6.

Alabama (4) vs Hofstra (13)Midwest · Score: 41.2

Hofstra is one of the hottest teams in the tournament — they've surged 32 spots since February 1st. Alabama has gone the other direction, fading 3 spots. A +35 momentum differential in a 4-vs-13 that already upsets 21% of the time historically? Put this one on your radar.

Vanderbilt (5) vs McNeese (12)South · Score: 49.3

Vanderbilt is overseeded by about 16 spots relative to their efficiency, and McNeese fits the Cinderella profile while climbing 13 spots since February. Classic 5-vs-12 territory.

The Momentum Watch: Teams Surging and Fading

The most underrated edge in bracket building is knowing which teams are peaking at the right time. Here are the biggest movers since February 1st among tournament teams:

Surging (climbing the rankings):

Fading (slipping in the rankings):

What Actually Predicts Tournament Upsets (And What Doesn't)

What matters:

The single strongest predictor of upsets is late-season momentum — teams that are climbing the national rankings in the final weeks before the tournament upset at significantly higher rates than teams that are coasting or fading. We measured this by comparing every team's Barttorvik ranking on February 1st to their ranking at Selection Sunday.

After that, the biggest factors are how close the teams actually are in efficiency (regardless of seed), whether the favorite is overseeded relative to their true quality, and the underdog's defensive efficiency. Good defense travels in March — that's not just a cliche, it's statistically significant at p<0.001 across 25 years.

We also found a specific profile more susceptible to an upset — a higher seed with mediocre defense that's also either inexperienced or undersized. When this profile triggered historically, the upset rate was 72.7%. That's not a typo.

What doesn't matter (despite what you'll hear on TV):

Tempo? Statistically irrelevant (p=0.956). Three-point shooting reliance? Doesn't matter (p=0.733). Free throw percentage? Nope (p=0.848). Conference affiliation barely misses significance. And the overall upset rate of about 26% has been remarkably stable for over two decades — there's no "year of the upset" trend.

The Model: How We Score Each Game

We built a composite upset score from 0 to 100 for every first-round matchup, weighting eight factors:

A score of 65+ is a High Upset Alert. 50–64 is Elevated Risk. 35–49 is Moderate. Below 35 is Low Risk.

This analysis was built from 25 years of NCAA tournament data (1,088 games, 2002–2025) using statistically validated predictors. Momentum data sourced from Barttorvik T-Rank, Feb 1 vs. March 15 2026 snapshots.

Want the raw numbers? Check out the interactive dashboard at coverodds.live.