Coming up next month is the NCAA’s March Madness, one of the most important sports events all year, where 68 men’s college basketball teams compete for a championship. However, the winner of this tournament isn’t necessarily the best team in the NCAA.
The problem stems from the fact that March Madness is made up solely of one-game series, which means that there is a small, forty-minute sample size used to determine which team in the series is better.
Because this sample size is so small, there is a greater chance that the worse Team A wins because there are fewer chances for the better Team B to outperform them even if, on average, Team B is better than Team A. To put it in math terms, there can be greater variance from the mean. This might matter less in a series where the two teams have wildly different talent levels, like a 1-16 matchup, but it can matter a whole lot when it comes to closely matched teams.
In theory, this should be common knowledge. Any basketball fan knows that it’s not all that uncommon for one team to have a game where they wildly over-perform, hitting every shot and locking down the other team, or have a game where they wildly underperform, going cold from the field and being unable to stop anyone on defense.
But in reality, fans are quite willing to accept that if Team A beats Team B during March Madness, then Team A is the better team, especially if the game is a blowout.
Take, for instance, the 1990 championship game between UNLV and Duke, where UNLV beat Duke 103-73: a thirty-point margin.
“UNLV made its point in an emphatic way, wiping out the final remnants of doubt that the Runnin’ Rebels are the finest college basketball team in America,” sportswriter Phil Axelrod wrote in the Pittsburgh Post-Gazette the next day.
Many similar conclusions were drawn by other sportswriters and fans alike, including sportswriter Thomas Bonk who wrote in the LA Times that the players for UNLV “were clearly superior.”
UNLV had an incredible game, shooting 62 percent from two-point range and 57 percent from three (admittedly, on fourteen attempts). Meanwhile, Duke paled in comparison, shooting 50 percent from two and 9 percent from three (again, on just eleven attempts).
Duke would go on to win the next two championships and consistently make deep tournament runs while UNLV would go on to have a Final Four appearance the next year, followed by a fifteen year stretch in which they only made the tournament twice, both times exiting in the first round.
Now, that’s not to say that the UNLV team of 1990 wasn’t better than the Duke team of 1990. Especially in college basketball, teams tend to change a whole lot from year to year, and Duke’s success in future seasons doesn’t guarantee that they were the best team in 1990. The point is that there’s no definitive way to figure out which team was the best at the time, and the NCAA Men’s championship game doesn’t give much of an insight.
Yet, as evidenced by the reactions of sportswriters at the time, many fans thought that UNLV was head-and-shoulders above Duke based on this one game.
Statistician Ben Taylor calls this “sample-size insensitivity,” which he defines as “a tendency to consider the given sample size as sufficient for reaching a conclusion.”
In his book, Thinking Basketball, Taylor compares the reactions to this game between UNLV and Duke to the reactions to Game 1 of the 1985 NBA Finals where the Boston Celtics blew out the Los Angeles Lakers 148-114, dubbed the “Memorial Day Massacre.”
The name notwithstanding, the reactions to that 1985 game were much more subdued, pointing out that the Celtics played phenomenally, but never saying that the Celtics were far and away the better team. That’s because this was Game 1 of a seven-game series; the Celtics still had to win three more games before they would be crowned the champions (and they never did, only winning one more game in the series, losing in six to the Lakers).
But even a seven-game series can’t always determine which team is better in a close matchup.
“In a matchup where one team would win 66 percent of the games, it would take a best-of-23 series for that better team to win the series at least 95 percent of the time,” Taylor wrote, referencing author Leonard Moldinow, “Such a team would lose a best-of-seven series about 20 percent of the time.”
This shows the randomness of basketball and sports in general. Obviously, the solution is not to make every series in March Madness 23 games long—that’s equivalent to more than half the length of the average NCAA college basketball season. But there are some things that could be done. For one, there doesn’t need to be 68 teams in March Madness.
Since the tournament expanded in 1985, the number one seeds have just two fewer championship game appearances than all the other seeds combined, and they have nine more championship wins than all the other seeds combined. In fact, seeds 9-16 have never once so much as appeared in the championship game, and seeds 4-8 each have three or fewer appearances and no more than one championship win.
On top of this, according to sports website SB Nation, as of 2017, a massive 41 percent of teams—122 out of 296–to qualify for March Madness since 1985 have never won a single game in the tournament. Part of this is a result of the fact that many of these teams have only appeared once or twice. However, the “Big Four” pro sports leagues (the MLB, NFL, NBA, and NHL) combined only have 2.5 percent of teams—3 out of 122–that have never had a playoff win, a striking difference.
Furthermore, well over half of teams in the tournament have never reached the third round, or the Sweet Sixteen, even more have never reached the Elite Eight, and still more—almost 85 percent—have never made it to the Final Four. Just twelve teams—4 percent—account for 58 percent of all Final Four appearances, according to SB Nation.
Those numbers are absolutely staggering, and bring about the obvious conclusion that, so far as determining the best men’s college basketball team goes, there is no reason for March Madness to be so enormous. The tournament would be better served by going back to the sixteen-team or even eight-team format and playing more games in each series.
The top teams in the country would participate in that tournament, and the lower-ranked teams would be divided up in a similar fashion and participate in their own tournaments where they would have a real chance of winning.
Although this would by no means guarantee that the better team won all the time, it would greatly improve the odds of that happening. This means there would be fewer upsets, and it would reward teams for being good, not for being lucky.
However, this is unlikely to happen. March Madness has high ratings almost every year, and games are always well attended (in normal years), making the NCAA a whole lot of money. Besides, it’s all right to have a tournament that doesn’t produce the best team—few tournaments can consistently do so anyway. The point of March Madness is to have fun watching basketball, and the increased randomness makes it all the better. Watching the underdog upset the juggernaut is always a cause for celebration. However, it’s important to go into the tournament with no illusion that the team that emerges victorious is by any means the best.