March Madness Upsets by Seed: Every First-Round Matchup

Through the 2026 NCAA tournament, a No. 16 seed has beaten a No. 1 seed twice in 164 attempts. A No. 15 has beaten a No. 2 eleven times. From there the numbers climb fast: 14 seeds have won 23 first-round games, 13 seeds 33, and 12 seeds 58, which works out to better than one game in three. The 9 seeds have a winning record against the 8s.

Those are the men's Division I figures since 1985, the year the bracket grew to 64 teams and every seed from 1 to 16 got the same first-round opponent it has had ever since. The count covers 41 tournaments, because the 2020 event was canceled. The totals move every March, so each record below is stated as of the 2026 tournament, the last one played before the field expands again.

Every first-round matchup since 1985, in one table

Each of the four regions sends eight pairings onto the floor in the round of 64: 1 against 16, 2 against 15, and so on down to 8 against 9, with the two seed numbers always adding up to 17. Four regions and 41 tournaments make 164 games for each pairing.

Matchup	Higher seed's record	Lower seed's wins	Lower seed's win rate
1 v 16	162-2	2	1.2%
2 v 15	153-11	11	6.7%
3 v 14	141-23	23	14.0%
4 v 13	131-33	33	20.1%
5 v 12	106-58	58	35.4%
6 v 11	100-64	64	39.0%
7 v 10	100-63	63	38.7%
8 v 9	77-87	87	53.0%

One row adds up to 163 instead of 164. In 2021, with the whole tournament played inside a controlled environment in Indiana, No. 10 seed VCU was withdrawn hours before tip-off because of positive COVID-19 tests and No. 7 seed Oregon advanced without playing. The game is recorded as a no contest.

Look at where the rate jumps. Between the 13 line and the 12 line the lower seed's chance nearly doubles, from one win in five to more than one in three. Then it flattens out. A 12 seed, an 11 seed and a 10 seed all win at roughly the same clip, which says something about how little separates the teams the committee ranks between about 20th and 48th in the country.

16 over 11.2%

15 over 26.7%

14 over 314%

13 over 420.1%

12 over 535.4%

11 over 639%

10 over 738.7%

9 over 853%

Men's NCAA tournament, 1985 through 2026. Each figure is the lower-seeded team's share of first-round wins in that pairing; the 7 v 10 figure is out of 163 games because of one no contest in 2021.

Show the numbers

How often the lower seed wins in the first round

Item	Value
16 over 1	1.2%
15 over 2	6.7%
14 over 3	14%
13 over 4	20.1%
12 over 5	35.4%
11 over 6	39%
10 over 7	38.7%
9 over 8	53%

How the committee arrives at the seed list in the first place, and why the true seed matters more than the number printed on the bracket, is set out in our explainer on how the March Madness bracket is built.

What the NCAA counts as an upset, and what it leaves out

The word gets used loosely. A 9 seed wins and a studio host calls it an upset; a bracket contest awards bonus points for it. The NCAA's own record-keeping is stricter. It treats a tournament game as an upset only when the winning team was seeded at least five lines below the loser.

In the first round that draws the boundary between two pairings that produce almost identical results. An 11 beating a 6 is five lines and counts. A 10 beating a 7 is three lines and does not.

First-round result	Seed-line gap	Upset by the NCAA's definition
16 over 1	15	Yes
15 over 2	13	Yes
14 over 3	11	Yes
13 over 4	9	Yes
12 over 5	7	Yes
11 over 6	5	Yes
10 over 7	3	No
9 over 8	1	No

The choice of definition changes the totals a great deal. Add the six qualifying rows of the table above and the first round has produced 191 upsets in 41 tournaments, a little under five a year. Count every lower-seed win, the 10s and 9s included, and the figure becomes 341, more than eight a year. Both numbers get quoted. They are measuring different things, and a claim about "the average number of first-round upsets" is meaningless until it says which one it means.

The five-line rule also applies after the first round, where it catches games a casual count would miss. A 9 seed beating a 1 in the round of 32 is an eight-line upset. A 6 beating a 2 in a regional semifinal is four lines and is not. On that count no tournament has produced more than the 14 upsets of 2021, across all rounds. The 2026 tournament had six.

There is a fair argument that the rule is too generous at one end and too strict at the other. An 11 seed that came through the First Four as one of the last at-large teams is often a power-conference club with more talent than the 6 it beats, and that result is logged as an upset. A 10 seed from a one-bid league beating a 7 from the Big Ten is not. Seed gaps are a blunt tool for measuring surprise, but they are the only tool that has been applied the same way for four decades.

16 over 1: UMBC, Fairleigh Dickinson and the near misses before them

Going into the fourth and last 1 v 16 game of 2018, No. 1 seeds had won all 135 meetings. Then, on the night of March 16 in Charlotte, the University of Maryland, Baltimore County beat Virginia 74-54.

Virginia was not just a 1 seed. It was the top overall seed in the field, 31-2, the champion of the ACC regular season and tournament. The score at halftime was 21-21. UMBC scored 53 points in the second half against a defense that had been the best in the country, Jairus Lyles finished with 28, and the margin of 20 was not a fluke of the final minutes. Virginia was beaten comfortably.

Five years and one day later it happened again. Fairleigh Dickinson, 63, Purdue, 58, in Columbus on March 17, 2023. FDU's presence in the bracket was itself an accident of the rulebook: the Knights had lost the Northeast Conference title game to Merrimack, which was ineligible for the NCAA tournament because it was still reclassifying to Division I, and the automatic bid passed to the runner-up. FDU then had to win a First Four game in Dayton just to reach Purdue. By average height it was the shortest team in Division I, and it was playing a front line built around 7-foot-4 Zach Edey.

What the two games had in common

Both favorites played slowly. Virginia ran the most deliberate offense in the country and Purdue fed the post, and a low-possession game is the best friend an underdog has, a point taken up further down. Both underdogs were small, quick and guard-led, and neither had won its own conference's regular-season title. Both 1 seeds were missing something: Virginia had lost De'Andre Hunter to a broken wrist days earlier, and Purdue's guards shot poorly under pressure all night.

Neither winner got any further. UMBC lost to Kansas State two days later, 50-43, and FDU lost to Florida Atlantic, 78-70.

And both losers answered. Virginia won the national championship in 2019. Purdue reached the title game in 2024.

The games that almost got there first

The 1989 first round had two. Georgetown beat Princeton 50-49 in Providence, with Alonzo Mourning blocking the Tigers' last attempts, and Oklahoma beat East Tennessee State 72-71. A year later Murray State took Michigan State to overtime before losing 75-71, the only 16 v 1 game to need an extra period. In 1996 Western Carolina had two shots in the closing seconds to beat or tie Purdue and lost 73-71. Even the first year had a scare: Michigan beat Fairleigh Dickinson 59-55 in 1985.

• 164Games played
• 2Wins by the 16 seed
• 20UMBC's margin of victory
• 5Fairleigh Dickinson's margin

Men's tournament only. Margins are the final scores of the two wins by No. 16 seeds, UMBC over Virginia in 2018 and Fairleigh Dickinson over Purdue in 2023.

15 seed over 2 seed history: eleven winners, most of them recent

Year	No. 15 seed	No. 2 seed	Score
1991	Richmond	Syracuse	73-69
1993	Santa Clara	Arizona	64-61
1997	Coppin State	South Carolina	78-65
2001	Hampton	Iowa State	58-57
2012	Norfolk State	Missouri	86-84
2012	Lehigh	Duke	75-70
2013	Florida Gulf Coast	Georgetown	78-68
2016	Middle Tennessee	Michigan State	90-81
2021	Oral Roberts	Ohio State	75-72 (OT)
2022	Saint Peter's	Kentucky	85-79 (OT)
2023	Princeton	Arizona	59-55

Richmond was first, under Dick Tarrant, a coach who made a habit of this sort of thing; his Spiders had already beaten defending champion Indiana as a 13 seed three years earlier. Santa Clara's win in 1993 had a freshman guard from British Columbia named Steve Nash making free throws at the end. Coppin State won by 13 and was never threatened late. Hampton's one-point win in Boise produced the picture of coach Steve Merfeld being carried across the floor, legs kicking.

Then nothing for eleven years, and then two in one day. On March 16, 2012, Norfolk State beat Missouri in the afternoon and Lehigh beat Duke that night, with C.J. McCollum scoring 30.

The split is stark. Four wins in the first 27 tournaments of the 64-team era. Seven in the 14 tournaments from 2012 through 2026.

1985 to 20114wins

2012 to 20267wins

First-round wins by No. 15 seeds over No. 2 seeds in the men's tournament. The first period covers 27 tournaments and the second covers 14, so the difference in rate is larger than the bars suggest.

Show the numbers

Wins by 15 seeds, before and after 2012

Item	Value
1985 to 2011	4wins
2012 to 2026	7wins

Why the 15 line woke up

Several explanations are offered, and none can be proved from eleven games. The most persuasive concerns age. The NBA's draft age limit, in force from 2006, pushed the best freshmen through college in a single season, so the programs that recruit them have often been younger than the conference champions they meet in March. That dynamic is covered on our page about the one-and-done rule.

A second is style. The three-point shot became central to college offenses over exactly this period, and a team that takes 25 or 30 of them has a wider range of possible scores on any given night than one that plays through the post.

A third is simply exposure. A player at Saint Peter's or Oral Roberts has now watched hundreds of games involving the opponent's league, and has film of every possession the favorite played all season.

The honest caveat is sample size. Seven events in fourteen years is a cluster, and clusters happen by chance. No 15 seed won in 2024, 2025 or 2026.

14 over 3 and 13 over 4: frequent, but seldom twice in one year

Combine these two lines and a 13 or 14 seed has won 56 first-round games. That is more than one a year on average. What is uncommon is a bracket that breaks in several places at once.

The 14 seeds

Twenty-three wins, spread across about half of the 41 tournaments. Two 14 seeds have won in the same year only three times: in 1986, when Cleveland State beat Indiana and Arkansas-Little Rock beat Notre Dame; in 1995, with Weber State over Michigan State and Old Dominion over Villanova in triple overtime; and in 2015, when UAB beat Iowa State and Georgia State beat Baylor a few hours apart.

The memorable ones tend to involve a blue-blooded victim. Bucknell over Kansas, 64-63, in 2005. Mercer over Duke in 2014. Weber State over North Carolina in 1999, when Harold Arceneaux scored 36. Oakland over Kentucky in 2024, with Jack Gohlke making ten three-pointers and attempting almost nothing else.

Only two 14 seeds have won a second game. Cleveland State did it in 1986 and came within a point of the regional final, losing 71-70 to Navy and David Robinson. Chattanooga did it in 1997. Nobody seeded 14th has reached the Sweet 16 since.

The 13 seeds

Thirty-three wins, a rate of one in five. A 13 seed that wins has a real chance of winning again, because its next opponent is a 5 or a 12 and not a 2. Six of them have reached the second weekend: Richmond in 1988, Valparaiso in 1998 on Bryce Drew's buzzer-beater against Ole Miss, Oklahoma in 1999, Bradley in 2006, Ohio in 2012 and La Salle in 2013. None went further.

One recent pattern cuts against the idea that the bottom of the bracket keeps getting more dangerous. In 2025 and again in 2026, every team seeded first through fourth won its opening game. That had never happened in consecutive tournaments since the field went to 64.

12 over 5, the upset everybody expects

No pairing has a bigger reputation. "Pick a 12" is the first piece of advice anyone filling in a bracket hears, and the numbers support it: 58 wins in 164 games through 2026, a rate of 35.4 percent.

An average of 1.4 wins a year hides a lumpy distribution. In six tournaments, 1988, 2000, 2007, 2015, 2018 and 2023, all four 5 seeds survived. In five others, 2002, 2009, 2013, 2014 and 2019, three of the four lost. The typical year sits between those extremes with one or two.

The usual explanations for 12 seeds beating 5 seeds

The first is about who lands on each line. The 12 seeds are, in most years, the best automatic qualifiers from outside the major conferences: teams that won 25 or 28 games, dominated their league and then won its tournament. The path to those bids is described in our guide to conference tournaments and automatic bids. The 5 seeds are the 17th through 20th teams on the committee's list, good but flawed, and frequently the fourth- or fifth-best team in a deep league.

The second is that the last at-large teams have sometimes been placed on the 12 line too, which means a 5 seed may be facing a power-conference opponent that the committee itself considered only a dozen places worse.

The third explanation is the least flattering to the legend. Look at the table again. The 12 seeds win 35 percent of the time, the 11 seeds 39 percent and the 10 seeds just under 39 percent. The 12 line is not unusually volatile. It is the lowest line that behaves like the middle of the bracket, and it gets noticed because the number looks so much bigger than 5. The real cliff is one line down, between 12 and 13.

A 12 seed that wins is also well placed to keep going, since a 4 seed is beatable and a 13 sometimes clears the way. Two have reached a regional final, Missouri in 2002 and Oregon State in 2021.

11 over 6 and 10 over 7: how thin the middle of the seed list is

The 6 seeds are 100-64. The 7 seeds are 100-63. A team that the committee rated among the top 28 in the country loses its first game about four times in ten to a team rated around 40th.

That should change how the middle of a bracket is read. The committee is ranking teams whose seasons are separated by a game or two, a late injury, or one bad afternoon in a conference tournament. A seed line in this range is a statement about a résumé, and a résumé is not a prediction.

The 11 line has an added complication. Since 2011, when the field went to 68, the last four at-large teams have played each other in the First Four in Dayton, and the winners have usually been dropped onto the 11 line. Those teams are often from major conferences, they arrive having already won under tournament pressure, and they have done disproportionate damage. VCU went from Dayton to the Final Four in 2011. UCLA did the same in 2021. Tennessee in 2014 and Syracuse in 2018 reached the second weekend, and in 2026 Texas beat NC State in the First Four, then BYU, then Gonzaga, before losing to Purdue by two in the Sweet 16.

So an 11 seed is one of two quite different things: a mid-major champion, or a power-conference team that barely got in. The record does not distinguish them.

The 10 seeds get less attention and nearly identical results. Their 63 wins include one curiosity already mentioned, the 2021 no contest, and their ceiling has been lower than the 11s': Syracuse in 2016 is the only 10 seed to reach a Final Four, against six 11 seeds. Why 11s have outrun 10s so consistently is partly the First Four effect and partly bracket geometry. A 10 that wins meets a 2 seed next. An 11 that wins meets a 3.

8 v 9: the pairing that is even by design

After the 2026 tournament the 9 seeds led 87-77. They swept all four games that year, TCU over Ohio State, Iowa over Clemson, Utah State over Villanova and Saint Louis over Georgia, to stretch a lead that had been 83-77.

Nothing about that should be surprising. The eight teams on these two lines are ranked 29th through 36th on the committee's list, and the difference between the 32nd and 33rd team in the country is not something any committee can measure. The split has wobbled around even for forty years and will keep doing so.

The more interesting fact is what happens next. The winner plays a 1 seed in the round of 32, almost always two days later and often in a building chosen for the 1 seed's convenience. Most lose. That makes the 8 line arguably a worse place to be than the 10 or 11 line, where a first-round win is harder to get but leads to a 2 or 3 seed instead.

Some do get through. Iowa beat top-seeded Florida 73-72 as a 9 seed in 2026 and reached a regional final. North Carolina beat Baylor as an 8 in 2022 on its way to the title game. Loyola Chicago beat Illinois in 2021, Wisconsin beat defending champion Villanova in 2017, Wichita State beat Gonzaga as a 9 in 2013, and Kentucky's 2014 team, an 8 seed full of future professionals, ended Wichita State's unbeaten season.

Here the two lines diverge in an odd way. The 9 seeds win the first-round series, yet the 8 seeds own nearly all the deep runs: four appearances in the national championship game, and the only title, Villanova's in 1985. No 9 seed has played for a championship.

Why upsets happen in a single 40-minute game

Give the better team enough chances and it wins. The whole structure of the NCAA tournament works against giving it enough chances.

Possessions. A college game is 40 minutes, played under a 30-second shot clock since the 2015-16 season. A typical team gets somewhere between 65 and 70 possessions. An NBA team, with 48 minutes and a 24-second clock, gets around 100, and then plays the same opponent up to seven times. Fewer trials means more variance, and a favorite that chooses to play slowly, as Virginia did in 2018, shrinks the sample further. The clock's path from 45 seconds to 35 to 30 is traced in our history of the college shot clock.

The three-point shot. It is worth half again as much as a two and goes in far less reliably from one night to the next. A 16 seed that makes 12 of 24, as UMBC did, has added 36 points from a shot the favorite's size cannot contest. The men's line has been at 22 feet, 1¾ inches since 2019-20; the distances at each level are compared on our page about the college three-point line, and the wider effect on the sport is the subject of our piece on how the three-point line changed basketball.

Age and continuity. The best low seeds are often built around fourth- and fifth-year players who have run the same offense together for several seasons. The high seed may start two freshmen who will be in the NBA by June.

Neutral floors, mostly. No team plays a tournament game in its home arena. Top seeds are placed near home where possible, so the crowd is not always neutral in fact, but the underdog's shooters are at least not walking into the favorite's own building. Compare the women's tournament below, where that protection is missing in the early rounds.

One game. No series, no second leg, no replay. A single bad shooting night ends a season that took five months to build, and the format has no way of correcting for it. That is a design choice. It is the reason the event is watched.

After the first round: low seeds that won again

A first-round upset is one game. The harder trick is the second, played 48 hours later against an opponent that has had a day to prepare and has watched what just happened.

Four 15 seeds have managed it. Florida Gulf Coast was first, in 2013, beating San Diego State two days after Georgetown. Oral Roberts followed in 2021 and Princeton in 2023. Saint Peter's, in 2022, went one further, beating Murray State and then Purdue to become the only 15 seed to play in a regional final.

The second weekend almost always has at least one team seeded 10th or lower. When Texas reached the Sweet 16 in 2026 it extended a run of 18 consecutive tournaments with a double-digit seed among the last sixteen.

Beyond that point the story changes character. Winning two games as a low seed needs a hot weekend. Winning four needs a team that was underseeded to begin with, and winning six has never been done by anyone below the 8 line. Which seeds have reached the Final Four, which have played for the title, and why the trail goes cold where it does are all on our page about the lowest seed to win March Madness.

The women's tournament by comparison

The women's NCAA tournament uses the same 16 seed lines and produces far fewer early upsets. The main reason is structural. Since 2015 the top 16 overall seeds, the teams on lines 1 through 4, have hosted the first and second rounds on their own campuses. A 13 seed in the men's tournament plays a 4 seed in a neutral city. A 13 seed in the women's tournament plays a 4 seed in the 4 seed's gym, in front of its crowd, after its opponent slept at home.

A second reason is the distribution of talent. For most of its history the women's game has had a steeper drop from the top handful of programs to the rest of Division I than the men's game has, though that gap has been narrowing. First-round wins by teams seeded 13th or lower have been rare enough to remember individually.

One of them stands above the rest. On March 14, 1998, No. 16 seed Harvard beat No. 1 seed Stanford 71-67 at Maples Pavilion, Stanford's home floor. Stanford had lost two starters, Vanessa Nygaard and Kristin Folkl, to knee injuries in the days before the game. Harvard had Allison Feaster, the nation's leading scorer, who put up 35 points. It was the first 16-over-1 result in either NCAA tournament and it stood alone for twenty years, until UMBC.

It is still the only one on the women's side. The format and site rules for that event are explained in our guide to the women's NCAA tournament.

What the 76-team field changes from 2027

The table at the top of this page is about to close. In May 2026 the NCAA announced that the men's and women's fields will each grow from 68 teams to 76 beginning with the 2027 tournaments. The First Four is replaced by a 12-game March Madness Opening Round.

1. 176 teams selectedThirty-two conference champions qualify automatically and the committee picks 44 at-large teams.
2. 224 sent to the Opening RoundThe 12 lowest-ranked automatic qualifiers and the 12 lowest-ranked at-large teams play first.
3. 3Six games among automatic qualifiersThe winners fill the bottom of the bracket, on the 16 line and part of the 15 line.
4. 4Six games among at-large teamsThe winners take the last at-large places, most likely around the 11 and 12 lines.
5. 5Round of 64Twelve winners join 52 teams already placed, and the familiar 1 v 16 through 8 v 9 pairings are played.

The structure the NCAA announced in May 2026 for the 76-team field. Which exact seed lines the at-large winners occupy depends on each year's seed list.

The pairings in the round of 64 keep their numbers. What changes is who is standing in them.

From 2027, every 1 seed will face an opponent that has already won a tournament game, and that got there by beating another team from the bottom of the list. Under the format used from 2011 through 2026, only two of the four 16 seeds came through a play-in. The 16 line will be, on average, slightly stronger and noticeably more battle-tested, and the weakest automatic qualifiers will mostly be gone before Thursday.

The middle of the bracket shifts too. With six Opening Round games among at-large teams instead of two, more of the places around the 11 and 12 lines will be held by teams that had to win their way in. The evidence from the First Four years suggests those teams are dangerous. More of them means the 5 v 12 and 6 v 11 records could tilt further toward the lower seed, or the mid-major champions who used to hold those lines could be pushed down to 13, changing that row instead. Nobody knows yet.

What can be said is that 1985 through 2026 forms one consistent series: 64 teams in the main bracket, the same eight pairings, 41 tournaments. Results from 2027 onward should be kept beside it, not added to it. A 16 seed's record against a 1 seed after 2027 will describe a different kind of 16 seed.

More on the sport away from the bracket is on our basketball hub, and the origin of the event's name is told on our page about why it is called March Madness.