Madness Myths: You Can’t Pick All Number 1 Seeds
When it comes to Number 1 Seeds in the NCAA Tournament, evidence suggests that the public forces quotas on how many to pick to the Final Four.
Hunter Dickinson and Michigan will be one of the No. 1 seeds with a chance to win it all in 2021 (Photo by Adam Ruff/Icon Sportswire)
When it comes to making picks in the NCAA Tournament, a lot of bracket pool participants have a numbers problem. They get caught up in patterns and seed numbers, rather than just picking the teams. That’s why one of the biggest complaints we receive before the tournament starts is taking too many No. 1 seeds deep into a tournament.
In fact, we know that if our recommendation to a subscriber actually suggests someone select all four No. 1 seeds to the Final, we have a decent chance of hearing about it. Do we make that recommendation every year? No, every year is different, and every group of top seeds is different. But in some years, the four best options happen to be the top four seeds.
Back in 2008, when we recommended that subscribers take all four No. 1 seeds, and suggested Kansas over Memphis in the final, we heard the protests. “But there have never been four No. 1 seeds reach the Final Four!” Well, it happened that year, and when Kansas’ Mario Chalmers hit a big shot to send the game to overtime against Memphis, and the Jayhawks won, our subscribers were very happy.
In 2017, we again recommended all four No. 1 seeds to a lot of small and mid-sized pools. Two eventually made it, but they were the two that we suggested to the Final: North Carolina over Gonzaga. About 90% of our subscribers reported winning at least one prize that year.
So in this post, we are going to dispel the myth that you should avoid picking too many No. 1 seeds to advance deep into the tournament, by going through some facts.
On Number 1 Seeds and Patterns
All four Number 1 seeds have advanced to the Final Four only once in the same season, in 35 years since the tournament expanded to a 64 (or more) team field. So you could say it is a rare event. But everything is relative. Here’s something that’s true of every seed combination you could put in a Final Four:
They. Are. All. Rare. Events.
For example, if you just used 12 possibilities (the Nos. 1 through 12 seed in each region), that gives you over 20,000 potential combinations that you could pick as your Final Four field. If you got really wacky and used 16 possible teams for each region, that’s over 65,000 combinations.
You could play a NCAA Tournament every year for the entirety of human recorded history and not hit well over half of the potential, really rare combinations.
Yet, a lot of bracket contestants have issues with taking the 1-1-1-1 combination because it is “boring” or “everyone is doing it,” in ways that they would not be bothered when picking 1-1-2-3 as the combination of Final Four teams.
These issues stem from assuming that the No. 1 seeds are always the most popular choices (which we address later) or that somehow the other combinations are more likely than just picking the No. 1 seeds.
Remember: You Have to Pick the Specific Combination Order Correctly
Let’s do a thought experiment. Marty McFly, after traveling back in a time machine, tells you that two No. 1 Seeds, a No. 2 Seed, a No. 3 Seed will reach the Final Four. But he doesn’t give you the identity of the teams. Armed with that information, how should you pick the Final Four to maximize your expected number of correct picks? (For the sake of this hypothetical, let’s also assume we don’t have specific info on team strengths and ratings).
It’s not as easy as just going with a 1-1-2-3 combination. You have to get the order and sequence right, pairing the seeds correctly with the regions. There are actually 12 different combinations if you select two 1’s, a 2 and a 3 to your Final Four:
You have a 1-in-12 chance of nailing the exact correct seed/region pairings and getting four right. But your chances are double that (2-in-12) of getting zero right. And in four other cases, you get one right, while getting exactly two correct in the five other scenarios.
Thus, picking all No. 1 seeds, even knowing that a 2 and 3 would also advance, would result in a higher average number of Final Four picks correct (two every time) than trying to match the pattern (1.5 picks correct on average). And that’s in a hypothetical where you know the exact seed numbers for the Final Four participants, something we absolutely do not know at the start of any tournament.
The Actual Data for Who Has Reached the Final Four
So let’s go through how frequently each seed number has reached the Final Four from each quadrant. (Remember that whole “you have to pick the right sequence” thing). To do this, we looked at the Committee’s ranking of No. 1 seeds since 2005, when they began setting a top overall seed. From 1985 to 2004, we used the highest ranked No. 1 seed in the AP Poll, and assigned each quadrant by order of how the AP poll ranked top-seeded teams in the final release before the tournament.
So Quadrant 1 here represents the region where the top overall No. 1 seed was positioned, and Quadrant 4 represents the final No. 1 seed by either Committee Rank (since 2005) or AP Poll Rank (before 2005). That gives us 35 seasons of data.
|Seed Number of Team||Quadrant 1||Quadrant 2||Quadrant 3||Quadrant 4||Overall|
- The Number 1 Seeds have made up nearly half of the Final Four teams (45.7%) from the first three Quadrants.
- The Number 1 Seeds have reached the Final Four at least twice as frequently as any other individual seed in each of the first three Quadrants.
- The 2 Seed has made the Final Four more frequently than the Number 1 Seed since 1985 out of Quadrant 4. However, much of that was in the first decade, when six of them made it. Since 1995, the Number 1 Seed has also been the most frequent seed to reach the Final Four out of Quadrant 4 as well.
- Number 1 Seeds have reached the Final Four overall about twice as frequently as 2 Seeds.
- Number 1 Seeds (57 times) have reached the Final Four more than teams seeded 3 to 11, combined (54 times).
So, past history has shown that Number 1 Seeds are by far the most likely seed to reach the Final Four, even as compared to No. 2 Seeds, and picking a team besides a Number 1 to advance carries a decent amount of increased risk.
Public Pick Rates and Number 1 Seeds
One of the considerations for our bracket picks is value. Usually (not always) the No. 1 seeds are the best teams (i.e., least risky) in the tournament, which is obviously something you want for your picks. But they can also be among the most popular choices, which can be a negative when you’re trying to make choices that give you an edge over your opponents.
But let’s take a look at the actual data on pick popularity by seed number. This chart is based on a decade’s worth of picking data (2010 to 2019) and shows the average number of teams at each seed level picked to win in each round.
|Seed Number Of Team||1st Round||2nd Round||Sweet 16||Elite Eight||Final Four||Champ Game||Overall Wins||Wins Per Team|
Here’s a quick explanation for some of that information, taking a look at the first row as an example. The average public entry has picked 3.92 No. 1 Seeds to advance out of the First Round. In the Elite Eight, the public averages picking 1.83 No. 1 Seeds to advance to the Final Four. For the overall tournament, the average bracket is picking the No. 1 Seeds, cumulatively, to win 13.76 games. Since there are four No. 1 Seeds each year, that means the average No. 1 Seed has been picked to win 3.44 games over the last decade.
So it’s true that the No. 1 seeds collectively are the most popular choices, being selected to win more games in every round compared to any other seed line. (Keep in mind these are averages, and don’t necessarily apply to every year and every No. 1 Seed.)
Now, let’s give some context to that information, and the value propositions, by comparing to historical results.
It’s the No. 2 Seeds that have been most overvalued by the public
The chart above shows the average pick rate by seed for the last decade. Let’s compare those public picking rates to the actual win results for all seeds over three periods:
- The last decade;
- The last 20 years; and
- All tournaments since 1985, since the NCAA Tournament went to 16 seeds in each region.
|Seed Number Of Team||Average Public Rate||Avg. Wins (L10 Years)||Avg. Wins (L20 Years)||Avg. Wins (Since '85)|
So let’s talk about what that data shows. You would expect the better teams in the NCAA tournament to be picked at a more frequent rate by the public, and that’s what the data shows. There’s logic to this, because value alone will not win you a bracket pool. You have to nail the teams that advance deep into the tournament and win titles in most scoring formats.
But you would also think, if there were no other biases at play, that the Number 1 Seeds would be the most overvalued group looking at popularity versus performance, since they are the group of teams to most likely advance.
But the data shows it’s actually the No. 2 Seeds who are more overvalued. For example, over the last ten years, Number 1 Seeds have averaged 3.13 wins. Over that same span, the public pick rates had Number 1 Seeds winning 3.44 games on average, about 0.3 more than the actual win rate.
Meanwhile, Number 2 Seeds (2.23 average wins over the last ten season, versus 2.76 expected wins based on public pick popularity) show the largest dropoff between the public’s expectations and actual results (about 0.5 wins).
What’s Going on Here?
This is a theory, because we don’t have access to each individual bracket from the public to analyze, only overall pick rates. But our guess is that the public buys into a myth that they need to avoid selecting too many Number 1 Seeds to the Final Four.
And who are the most likely replacements when you force a quota? It’s the Number 2 Seeds, who end up being over-selected as a group relative to their expectations.
The average public Final Four picks (based on data from the last ten years) feature:
- 1.8 1 Seeds
- 1.0 2 Seeds
- 0.8 3 or 4 Seeds
- 0.4 5 Seeds or worse
Rounding that even further, then, the typical bracket will have two No. 1 Seeds, a No. 2 Seed, and a No. 3 or No. 4 Seed, with some percentage of brackets getting even wilder than that.
Championship Games and Public Quotas
We see evidence for a similar artificial quota being enforced in championship game picks. The public averages picking 1.07 No. 1 Seeds to the Final–just barely over one per bracket on average. That suggests it is rare for most brackets to actually have two No. 1 Seeds matched up in the Title Game. Even though that’s the single most likely outcome!
There are 8 possible combinations matching up two No. 1 Seeds in the title game, and picking a winner. On average, based on our pick data for each No. 1 Seed, you would expect the public to have picked each specific 1 vs. 1 matchup and outcome in 3.6% of brackets on average.
Unless you pick the overall most popular championship pick to go against the most popular choice from the other side, there’s a decent chance in small pools that no one will have your specific final game combination. And that’s true even if you have two No. 1 Seeds in the final!
In 2017, we had North Carolina over Gonzaga in most of our recommendations, and neither was the single most popular choice (that was Villanova). Gonzaga was being selected at a lower rate to the title game than three teams who were not No. 1 Seeds. By doing the “obvious” thing of taking two No. 1 Seeds in many brackets, 90% of our subscribers reported winning a prize that year when that’s what actually resulted.
In 2019, both Virginia and Gonzaga were actually being picked by the public to reach the Final Four at a lower rate than their projected actual odds to get there. So both had value as Final Four (or deeper) picks. So don’t assume that all top seeds are overvalued. The evidence shows that the public may sometimes avoid picking too many No. 1 Seeds, creating value on the ones they spurn.
Pick Teams, Not Seed Numbers
We’ve shown you some historical trends and demonstrated that, overall, the No. 1 Seeds outperform other teams and are the safest category to pick to the Final Four. But we don’t make pick recommendations based on these historical seed trends. In fact, some “hot trends” that come from seed-based results are often the path to poor predictive advice.
Teams are not defined by their seeds, and it’s important to make picks based on the actual team quality. Before adapting a strategy to your scoring system, or your pool size, or public pick rates, the initial goal should be to pick the most likely outcomes. Then, you can deviate from that based on those other considerations, taking on more risk if the point system dictates it, if the pool size requires greater risk taking, or if great values present themselves.
Often, there is a high correlation between No. 1 Seeds and the best teams in college basketball. But sometimes a team will get seeded as a No. 1 Seed based on some close results or key outcomes, but not predictively look like one of the best. Other times, one of the best teams may lose a few key results and be seeded lower.
When No. 1 Seeds have also been in our Top Four in the predictive power ratings at the conclusion of Selection Sunday, they have averaged 3.5 wins in the tournament since 2010. Seven of them won national titles over the last decade. The eleven No. 1 Seeds we did not have rated that highly over the last decade averaged only 2.1 wins in the tournament, and only one reached the Final Four. So when we make recommendations on teams, they aren’t blindly done on seeding.
What should be clear, though, is that forcing predictions to fit quotas or pre-set notions is buying into a myth. If you intentionally avoid selecting three or four No. 1 Seeds in a small pool, you may be making choices that hurt your chances of winning the pool.
Why should you be okay with picking a team if they happened to lose in their conference tourney and drop to a No. 2 Seed, but be troubled picking the exact same team if they won their conference tournament and were the No. 1 Seed in that region instead, just because you were already picking the No. 1 Seeds in two other regions?
We want our opponents to think that way, because it creates more value opportunities for us. But we want our readers to understand that buying into that mentality is not the best way to win.