The maddest March: at last, a 16-seed upsets a number one

By D.R.

THE line separating the improbable from the impossible is hard to pin down. The annual single-elimination tournament to crown the champion of North America’s National Collegiate Athletic Association (NCAA) in men’s basketball is known as “March Madness”, thanks to the steady diet of upsets it produces. Every year, a few ragtag gangs of fresh-faced students from little-known universities, likely destined for mundane careers in accounting, sales or the like, somehow manage to topple a heavily favoured juggernaut packed with future National Basketball Association stars. Yet despite the event’s well-deserved reputation for giant-killing, it had never delivered the ultimate shock—a top-seeded team losing in the first round—until last night. In a game that was supposed to be little more than a tune-up before facing more formidable opponents later in the tournament, the top-ranked University of Virginia was stomped by the humble, 16th-seeded University of Maryland-Baltimore County (UMBC) in a 74-54 blowout.

From a statistical perspective, perhaps the only thing more surprising than the result was that it had taken so long for an upset like this to occur. The NCAA tournament is divided into four regions, each containing 16 colleges seeded based on their expected strength. From 1985, when the event adopted its current format, to 2017, schools given a number three seed—i.e., the ninth through 12th-best teams in the country—won 111 of their 132 first-round matchups against their 14th-seeded rivals. That 84% winning percentage works out to a “3-14” upset occurring a bit more often than once every other year. A “2-15” stunner is far rarer: second-seeded teams (the fifth through eighth-strongest in the event) had a won-lost record of 124-8 entering this year’s tournament, meaning that a 15-seed had only pulled off a victory once every four years on average. But a 16th seed beating a number one? Even for March Madness, that had proved a bridge too far. Over 33 tournaments, top seeds had gone undefeated in their 132 first-round matchups.

The persistent failure of 16-seeds to defy the odds has caused mild head-scratching among quantitative analysts of college basketball. Statistical models tend to give 16-seeds anywhere from a 1% to a 5% chance of pulling off such an upset, depending on the relative strength of the teams. If the average 16-seed has about a one-in-33 chance of victory, then around four of them should have managed to win by now; the odds that all 132 would lose are a paltry 1.8%. One possible explanation is that the methods used to produce such predictions are poorly suited for analysing extremely lopsided matchups—most regular-season games pair teams that are far closer in skill than the NCAA tournament’s 1-16 contests. Another is that the tournament features better officiating and playing conditions than earlier games, reducing the chance that the stronger team loses because of a few bad breaks. Even so, the sport’s Goliaths had almost certainly benefited from a long run of good fortune. If eight two-seeds had managed to go down in the first round, surely a number one would succumb eventually.

Even if it was about time for an unprecedented upset to occur, Virginia looked like one of the least likely potential victims. Not all 1-16 matchups are created equal: in many years, one region’s top seed is seen as unusually weak, and occasionally a 16-seed might be as strong as another region’s number 12 or 13. This year’s tournament featured two such outliers: bookmakers gave Xavier, number one in the West region, a lower chance to reach the tournament’s final four than North Carolina, the region’s second seed. Similarly, the University of Pennsylvania had a strong claim to be the best 16-seed in NCAA history. The Virginia-UMBC matchup, however, was a poor fit for the “weak one/strong 16” paradigm. In UMBC’s one regular-season game against a nationally competitive opponent, the Retrievers were crushed by Arizona, 103-78; the remainder of their schedule consisted of matchups against other minnows. The Cavaliers, meanwhile, were the top-ranked team in America, posted the NCAA’s best regular-season won-lost record at 31-2, and were the public’s most popular pick to win the title.

On the other hand, Virginia’s style of play did make them unusually vulnerable to weaker teams. The Cavaliers played at the slowest pace in all of college basketball this year, at just 59.4 possessions per 40 minutes. On offence, they play a half-court game that often does not yield a field-goal attempt until less than ten seconds remain on the 35-second shot clock; defensively, they rarely make gambles that risk yielding a quick, easy basket for their opponents. In general, better teams benefit from playing up-tempo, since each possession represents another opportunity for them to grow a lead or reduce a deficit with their superior ability. Conversely, underdogs tend prefer a slow, grinding game, in which luck matters more and skill less. By effectively making each game shorter, Virginia increased its chances of beating better teams and decreased its odds of defeating worse ones. And since there weren’t any better teams out there than the Cavaliers, their slow-paced style became a pure liability.

Virginia’s defensive strategy further reinforced this weakness. Playing a scheme known as the “Pack Line”, the Cavaliers generally kept their defenders well inside the three-point line, preventing opposing teams from gaining access to high-value regions close to the basket and daring them to shoot three-pointers. Just like a slow pace, the higher the share of shots in a game from three-point range, the more the outcome is determined by luck rather than skill. A typical team that makes half of its two-pointers and a third of its threes might score anywhere from, say, 48 to 72 points on 60 shots taken in the painted area in front of the hoop. But on the same number of three-point attempts, it could rack up anywhere from a paltry 30 points to a dominant 90. The Pack Line’s chief weakness is an opposing club happening to get hot from behind the three-point arc, and that was exactly what happened to the Cavaliers at the worst possible time: UMBC shot a blistering 12-of-24 from three.

Following such a remarkable, emphatic win, handicappers probably need to adjust their expectations for UMBC going forward. No matter how much fortune may have favoured them, the fact that the Retrievers knocked off the national number one is strong evidence that they must be a fairly decent squad. Moreover, they now inherit the relatively cushy path through the remaining early rounds of the tournament that had been intended for Virginia: they face ninth-seeded Kansas State tomorrow, and would play the winner of the game between fifth-seeded Kentucky and 13th-seeded Buffalo after that if they advance.

Nonetheless, the limited historical record does not show that the authors of astonishing early-round upsets continue to overachieve. Only one of the eight victorious 15-seeds managed to win their second-round game, and none of them have advanced to the tournament’s “Elite Eight” quarter-finals. Even after a hefty boost to UMBC’s rating, FiveThirtyEight, a statistical website, gives them just a 28% chance of victory against Kansas State, and less than a 1% shot at reaching the final four. In all likelihood, the Retrievers will have to content themselves with one lone night of basketball immortality.

(Photo credit: Getty Images)