Mercer's Seniors Beat Duke, But Does Experience Matter In March Madness?

Last Friday, No. 14 seed Mercer shocked No. 3 seed Duke, 78-71, in the opening round of the NCAA tournament. Mercer started five seniors, while Duke started freshman and presumed-lottery-pick Jabari Parker alongside three sophomores and a lone senior. Quickly a narrative emerged: Mercer won because it was more experienced than Duke, suggesting that the era of one-and-done will also be an era of Cinderellas.

There's a seductive logic here, but is it true? Are the one-and-done-reliant teams—or at least those teams not coached by John Calipari—secretly vulnerable against older, more seasoned opponents? The answer, it turns out, is no.

We calculated the average "age" of each NCAA tournament team—weighted by each player's minutes played during the season—for every tournament dating back to 2007. Data were from Sports Reference and ESPN; you can see the full results here. Freshmen were marked as 1, sophomores at 2, etc., so a team that gave all the minutes to freshmen would have an average "age" of 1, a team of all seniors would have an average "age" of 4, and a team that split time equally between all grades would have an average age of 2.5.¹

Across all 512 tournament teams since 2007², the median team age was 2.77, meaning that squads were slightly skewed toward older players. It turns out that higher seeds are typically a bit younger than lower seeds (although No. 8 seeds have been the youngest of all; see below), and this divide has been fairly consistent since 2007. That makes sense: One-and-dones are a real phenomenon, and those players are more likely to go to the top schools.

Mercer's Seniors Beat Duke, But Does Experience Matter In March Madness?

This split can lead to some pretty dangerous traps. Here's a list of all 489 tournament match-ups since 2007. There have been 131 basic tournament upsets, and 55 percent were won by the older team. There have been 70 "big" upsets like Mercer-Duke—in which the winning team was at least four seeds lower than the losing team—and 57 percent were won by the older team. Are older teams better at upsets, or is this just showing that lower seeds are older, and lower seeds, definitionally, get more upsets? Maybe we need to approach the matter from a different angle: In games where one team is significantly older than the other, has the older team actually tended to outperform its seed?

Mercer had a weighted age of 3.5 years, making it the second-oldest team in the 2014 bracket, while Duke had a weighted age of 2.4 years. That 1.1-year gap was large enough to draw all this attention in the first place, so let's start with games with a gap of at least one year, the equivalent of a team of all seniors playing a team of all juniors.

There have been 33 such tournament games since 2007. We can use historical data on seeding matchups to estimate the expected wins for the older teams in these games; 11 seeds have beaten 6 seeds in 33 percent of matchups, so if an older 11 is playing a younger 6, the "expected wins" would be 0.33. We have to remove two historically rare seeding matchups³, but in the table below you'll find the remaining 31 "one-year gap" games. The older team was expected to win 13.4 games and actually won 12 games.

If we want to widen our sample a bit, the average age gap for a game was 0.43 years, so games with a gap of 0.86 years had twice as large a split as average. There have been 57 such games since 2007; removing three rare matchups yields 54 games. The older team was expected to win 21.9 games and actually won 19 games

It would appear that older teams aren't overperforming their seed; they're actually slightly underperforming.

There's one extra wrench, though: In the 31 one-year gap games, John Calipari coached the younger team six times, including this year's Kentucky squad that ranks as the youngest team in our sample (1.34-year average). In the 54 0.86-year gap games, he coached the younger team 11 times. In those 11 games, Calipari's teams were expected to win 7.3 games and actually won 10. Is it possible that older teams generally do have an advantage, but John Calipari just kicks ass at coaching young teams?

If you take out the Calipari games, the older teams no longer underperform, but, alas, they perform almost exactly in line with expectations. In the 25 remaining one-year gap matchups, the older teams won 12 games vs. an expected win total of 11.0. For 43 remaining 0.86-gap matchups, the older team won 18 games vs. an expected win total of 18.1.

Mercer's Seniors Beat Duke, But Does Experience Matter In March Madness?

This year's tournament has a ton of young talent, so we get two Sweet 16 matchups with large age gaps: Tennessee vs. Michigan (Tennessee is 1.2 years older), and Louisville vs. Kentucky (Louisville is 1.5 years older). Should either Tennessee or Louisville win, expect to hear plenty about the importance of experience. Take it with a grain of salt.

1. This was a simple methodology that let me calculate team ages without intensive research on individual players, but it has a couple weaknesses that are worth mentioning:

  • Injuries: If starters missed significant time during the regular season but returned for the tournament, the minutes-weighted team age might not be perfectly representative of the team on the floor for critical tournament games.
  • Redshirts: The data I used didn't separate out redshirt players, so the "age" of squads with key redshirts is being underestimated (since those players have been with the team one season more than their listed class year would indicate).
  • Transfers: The Mercer argument hinges on the idea that the seniors had a leg up because they had spent years playing together. Transfers haven't spent that sort of time with their teams, even if they're older, so the "age" of squads with key transfers is being overstated a bit, in terms of the effect we're talking about.

2. We ignored play-in losers. Sorry guys.

3. I decided to throw out games in which the seeding matchup had occurred less than 10 times in tournament history. The three games removed in this way were Michigan (4) over Florida (3) in 2013 (Florida was older by 1.4 years), Connecticut (3) over Butler (8) in 2011 (Butler was older by 1.0 years), and Florida (7) over Norfolk State (15) in 2012 (Norfolk State was older by 0.9 years). You'll notice that the older team lost all of these games, so in removing them we're being extra generous to the "older is better" theory.