A Hybrid Solution to Seeding the NCAA Tournament – update

Last year, I proposed a hybrid solution to seeding the NCAA Tournament, combining the aspiration to both reward strong resumes and maintain the integrity of the bracket–that is, actually giving higher seeds the better draw.

That post was lacking in clarity a bit, so here is an update aimed at making things a bit clearer and easier to understand.

The Goal

As a recap, our goal should be that teams with the best resumes get the most favorable draws. If the 64 tournament teams were ranked the same in both resume and team quality, things would be easy and we could just seed teams in that order. However, when those two rankings do not line up, we have to choose between ordering teams by resume versus giving better resumes a more favorable draw.

Here’s an example: let’s say we have a 4-team league. The Dominants went 3-0, the Pretty Goods went 2-1, the Luckys went 1-2 and the Bad Beats went 0-3. The Luckys beat the Bad Beats, but the Bad Beats are a much higher-quality team and neither the Dominants nor the Pretty Goods want to play them over the Luckys. What do we do? We have two choices:

  • Seeds teams by the resume/record: This would put the Dominants #1, the Pretty Goods #2, the Luckys #3, and the Bad Beats #4.
  • Create the most favorable path: In this scenario, the Dominants deserve the best draw and so we would move the Luckys to #4 and the Bad Beats to #3.

In the first scenario, the Luckys are rewarded with the higher seed over the Bad Beats, deservedly so as they finished with a better record. However, the Dominants are hurt in this scenario (with the Pretty Goods the beneficiary)–they’d prefer to be the #2 seed!

In the second scenario, we fix that–now the Dominants get what they deserve, the easiest draw. However, now it’s the Luckys who are unjustly hurt–they have to take the toughest draw even though they deserve to have an easier path than the Bad Beats.

As you can see, there’s no solution to satisfy everyone, so we have a decision to make. I don’t think there’s an objectively correct answer, it’s just a matter of subjective preference. But in my mind, I lean towards protecting the top resumes, meaning I’d opt for option 2.

The Methodology

Ok, onto the NCAA Tournament and a real-life example. (NOTE: I’m making one small tweak from my last post for simplicity and removing the 1st Round safeguard against teams moving more than 3 seed lines; this can easily be added back in.)

Each round will work like this:

  1. Rank teams/pods by their resume.
  2. Divide the pods in half.
  3. The top half of remaining pods are protected and stay in the order of their resume.
  4. The bottom half of pods are re-sorted by their (predictive) true strength. The easiest of these pods is paired with the best resume team (the #1 overall seed) down to the toughest pod being matched up with the worst of the protected pods.
  5. These matchups are now locked in and create new pods.
  6. Repeat for each round.

This will be clearer as we progress through an example. I’m using my own Achievement Score as the resume metric (it is a version of “Parcells“, from the “WAB” family of metrics, for those that follow Seth Burn) and my predictive ratings for team’s true strength.

Initial Ranking

Initial ranking of teams, sorted by resume

We start with a ranking of teams from 1 to 64 based on their resume. Already, we can see some flaws–I doubt Ohio State is very happy about having to play Arizona (#9 true strength) while the team below them, Illinois, gets to play Stephen F. Austin (#62).

1st Round Matchups

So let’s do our first adjustment. The teams on the left (the top 32 resumes) are protected and we sort the teams on the right now by their team quality.

Okay, now Ohio State gets a much easier 1st Round matchup in #45 Northern Iowa, and a slightly better one than Illinois gets (#43 Richmond). Stephen F. Austin bears the brunt of this, as they drop all the way down a 16-seed to play Baylor since the Lumberjacks are the 3rd-worst team in the field.

These pairings are now locked in to what I’ll call “pods”. They cannot change, but the pods can move around, affecting their matchups in later rounds.

2nd Round Matchups

Initial 2nd Round Matchups
Adjusted 2nd Round Matchups

The first chart shows our initial 2nd Round matchups. The 1st Round matchups are now locked in, but we will now adjust the 2nd Round pairings. Again, the top half of remaining “pods” are protected–that is the top 4 seed lines in this iteration. The other pods (seeds 5 through 8) are re-ordered based on their pod strength (defined as the predictive ranking of the best team in the pod).

San Diego St. is our #1 overall team. To start this iteration they were paired with USC (#44 predictive ranking) and Arizona (#9). They say, no way we want to play Zona and instead look down the list and get paired up with the 5-12 matchup between ETSU (#35) and Yale (#47), which now becomes an 8-9 matchup. Note, however, that while ETSU gets moved to a worse seed here, they still keep their reward of being an original 5-seed and get to play Yale instead of Arizona, who USC (an original #8 seed) has to play.

Sweet 16 Matchups

Initial Sweet 16 Matchups
Adjusted Sweet 16 Matchups

Same as before, we’re down to 16 pods of 4 teams each. The top 8 pods (those headed up by the #1 and #2 seeds) are protected and the bottom 8 are re-ordered. Creighton is the top #3 seed (#9 overall resume), but they are no longer protected and are the weakest of the unprotected pods, so they move into San Diego State’s region. Nobody wants to play Michigan St., my #3 team in predictive rankings, so they (and their pod) move into FSU’s region as the Seminoles are the last of the protected pods.

Again, we see the benefits of the rolling matchups. Auburn is the 11th-best resume and now has to join the region of the #1 quality team, Kansas, while a worse resume team, Oregon, gets to draw Kentucky’s much more favorable pod. However, Auburn still has the easier 1st Round matchup compared to Oregon (#54 Texas State instead of #49 Liberty) and the easier 2nd Round matchup (#14 Rutgers instead of #5 Ohio State), so they do derive some reward from their better resume.

Elite 8 Matchups

Initial Elite 8 Matchups
Adjusted Elite 8 Matchups

Same thing here: San Diego State gets to swap out a pod with two top-6 teams and replace them with one where #9 Arizona is the best team, making their Elite 8 opponent a bit weaker. Dayton as the #5 overall resume is “hurt” here (they actually aren’t hurt in this exact scenario, since San Diego State is not as strong of a team as their resume indicates) but still has an easier draw in each of the first 3 rounds than #8 resume Florida State.

Final Four Matchups

Initial Final 4 Matchups
Adjusted Final 4 Matchups

Finally, we match up regions for their Final Four pairings. The top 2 overall seeds, SDSU and Kansas, are protected. In this example, Baylor and Gonzaga’s regions are flipped, allowing SDSU to choose the (slightly) more favorable of the two regions, thus avoiding Gonzaga and the two Michigan schools, all rated among the top 6 toughest teams.

The Final Bracket

Above are the initial bracket, seeded solely by resume, and the final “hybrid” bracket resulting from the above method.

Here’s a view of the progression, starting with the initial bracket in gray and moving inward. Teams in green are “protected” teams for that round, while teams in yellow are ones that were moved to a different pod during that iteration. You can follow SDSU and see how they were able to swap out their initial pod for an easier one at each stage:

  • First, they went from Siena to Saint Francis (PA) as their 1st Round opponent.
  • Next, they eliminated Arizona from their potential 2nd Round opponents and instead draw ETSU and Yale.
  • For the Sweet 16, Ohio St. is a dangerous team, so they get moved out in favor of Creighton and Illinois.
  • In the Elite 8, they have Michigan, Michigan St. and FSU in their path, but here get to swap them out for Arizona, Iowa, and Dayton.
  • Finally, in the Final Four, they get two avoid Gonzaga’s region.

Creighton provides another good example of the balancing act going on here. They have the 9th-best resume, but rate just 26th in the predictive rankings. In the Sweet 16 iteration, they get moved from the best 3-seed to worst 4-seed. However, they get to keep their easy 1st round matchup vs New Mexico St. (the 56th-best team in the field) and 2nd round matchup against Illinois (#19), making their path to the Sweet 16 much easier. Michigan St., meanwhile, gets moved into Creighton’s slot as the top 3-seed, but has to play #50 Akron in the 1st round and potentially #6 Michigan in the 2nd.

Alternative structures

There are plenty of adjustments you could make to this structure. Again, it’s mostly a matter of personal preference. Here are some potential ones:

  • Let protected teams pick their pod matchup instead of automatically assigning them. This is my favorite.
  • As I mentioned in the previous article, you may want to calculate pod strength as the weighted true strength of the entire pod instead of just using the shortcut of the best team. This would be slightly more accurate.
  • Also as mentioned earlier, you may want to provide a little bit of protection for the unprotected teams by limiting how far a team can fall (or rise) from their initial seeding. Any adjustment here will (slightly) hurt the better resume teams, so it is a trade off.
  • You can, of course, use a different resume metric (ESPN’s Strength of Record (SOR), Wins Above Bubble (WAB), etc.) or predictive metric (KenPom, etc.). I just used my own as it was easier for me to code. You could even use a combination of metrics, just don’t mix resume metrics and predictive metrics together please (like the NET does)!
  • You may need to include some safeguards to ensure some of the bracketing principles aren’t violated, such as two teams from the same conference playing early in the tournament.
  • After the final bracket is set, you could then let teams also draft their location for the regional and the first weekend.

There are lots of minor tweaks that could potentially improve upon this. However, I think the main framework is the best way to combine rewarding strong resumes and maintaining the fairness and integrity of the bracket by giving the best resumes the most favorable path.

Posted in College Basketball, descriptive, March Madness, predictive, team evaluation | Leave a comment

A Hybrid Solution to Seeding the NCAA Tournament

I’m on record many times as being in favor of a purely objective selection criteria. However, when it comes to seeding, things are a bit trickier. While selection is a purely binary process–you’re either in or you’re out–all teams are intertwined when it comes to seeding.

I’ve previously offered up an alternative where teams are selected by their resume, but seeded by their true strength. While this balances the bracket correctly, it completely abandons the reward-based nature of the tournament, which we’d like to preserve as much as possible.

There is no system to both perfectly reward teams for their body of work and preserve the integrity of the bracket so that higher-seeded teams are never given a tougher draw than a team seeded lower than them. Every time you move a team to help with bracket integrity for one team, you by definition hurt another.

Despite this impediment to perfection, we can do better than either a purely reward-based or a purely quality-based seeding system. It necessarily involves trade offs, and you have to balance maintaining the reward for some teams against hurting/helping others.

Continue reading
Posted in College Basketball, descriptive, March Madness, predictive, team evaluation | 2 Comments

The Case for Objective Selection

Every year around this time, the 350+ Division I basketball teams wrap up their regular seasons and–save for the 32 conference champions who receive automatic bids–leave their fates in the hands of the aptly-named Selection Committee. Later today, this committee will finish pouring over each team to select the 36 at-large teams to complete the field of 68. It’s a supposedly painstaking process that takes many days and long nights for the committee members to arrive at their selections.

It’s also completely unnecessary. Actually, worse than that. It’s detrimental to the process. I’m here to tell you why.

Continue reading
Posted in College Basketball, descriptive, March Madness, predictive, team evaluation | 3 Comments

Conference Tournament Predictions 2019 – All Tournaments

All conference tournament predictions are now in. You can follow along with this year’s contest by using this google spreadsheet link: predictions. TeamRankings currently holds a commanding lead and is trying to win their 3rd Conference Tournament Predictions Contest. Final results to be posted once all the games are completed.

Posted in College Basketball, Conference Tournament predictions, March Madness, predictive, team evaluation | Leave a comment

Conference Tournament Predictions 2019 – Early Tournaments

I’m once again tracking conference tournament predictions between my own system, KenPom, and TeamRankings. I discuss the scoring system in this post, but essentially for each team result that actually happens (e.g. Loyola wins the MVC, Indiana loses in the 2nd Round, etc.) we grade the predictions based on how likely they thought that result was to happen, punishing predictions more severely the less likely a system predicted it to happen.

You can follow along with this year’s contest by using this google spreadsheet link: predictions. In the very early going, TeamRankings is off to an early lead, followed by my Predict the Madness system, and then Ken Pomeroy’s in 3rd.

I will try to post another update early next week once all of the conference tournament brackets have been set. Until then, enjoy the early conference tournament action!

Posted in College Basketball, Conference Tournament predictions, March Madness, predictive, team evaluation | Leave a comment

Conference Tournaments 2018 – Final Results

Before I get to this year’s predictions, I realized I never posted last year’s final results, but my predictions did pull out the victory, narrowly holding off a strong TeamRankings run. KenPom finished 3rd.

That gives two victories apiece for myself and TR, with KenPom having one victory. Full results (PTM are my own predictions):

  • 2013: (1) TR, (2) KP, (3) PTM
  • 2014: no contest
  • 2015: (1) TR, (2) KP, (3) PTM
  • 2016: (1) PTM, (2) TR, (3) KP
  • 2017: (1) KP, (2) TR, (3) KP
  • 2018: (1) PTM, (2) TR, (3) KP

It’s been an especially impressive run for TeamRankings, who has 2 first-place finishes and 3 second-place finishes, while never finishing in last.

The full results (Google spreadsheet) can be viewed by clicking here.

Posted in College Basketball, Conference Tournament predictions, March Madness, predictive, team evaluation | Leave a comment

Conference Tournament Predictions 2018 – Early Results

As I’ve done for most of the last few seasons (I missed 2014 for some reason), I’m tracking conference tournament predictions between Ken Pomeroy, Team Rankings, and myself. In the previous 4 years I’ve tracked, Team Rankings has 2 victories to 1 each for both KenPom and my own predictions.

Now that all the conference tournament fields have been set, here are the full predictions for all 3 systems this year (with results through Sunday): predictions. I discuss the scoring system in this post, but essentially for each team result that actually happens (e.g. Loyola wins the MVC, Indiana loses in the 2nd Round, etc.) we grade the predictions based on how likely they thought that result was to happen, punishing predictions more severely the less likely a system predicted it to happen.

In the early going, it’s my Predict the Madness system that is off to a strong start, but there are still a lot of games to be played as the majority of conferences will kick off tournament play this week. I’ll be back early next week with the final results.

Posted in College Basketball, Conference Tournament predictions, March Madness, predictive, team evaluation | 1 Comment

Conference Tournaments 2017 – Final Results

Quick post today as the big tournament has started, but time to wrap up the Conference Tournament prediction contest. For the first time, Ken Pomeroy takes home the title. Team Rankings still leads with 2 overall titles and my own rankings have 1 title as does KenPom now.

Team Rankings came in a close 2nd while it was a tough year for Predict the Madness, who came in a distant third.

The spreadsheet with this year’s results can be found here.

Posted in College Basketball, Conference Tournament predictions, predictive, team evaluation | Leave a comment

Conference Tournament Predictions 2017 – Late Tournaments

Last week, I posted the predictions for the 13 early tournaments. Now the other 19 (that’s right, the Ivy League has joined the party) tournaments are set and most are underway, so I’ve updated the Google spreadsheet with all the predictions. KenPom is off to the early lead and looking for his first ever victory in the prestigious Conference Tournament Predictions contest that he doesn’t know he’s in. TeamRankings, the two-time champion, is in 2nd place, while my predictions (the reigning champs!) are trailing the pack.

Cinderella Index – Early Tournaments

In last week’s post, I introduced the Cinderella Index as a way to try and quantify which teams to root for during chamionship week. Let’s see where we stand with those Cinderellas: Continue reading

Posted in College Basketball, Conference Tournament predictions, March Madness, predictive, team evaluation | 1 Comment

Conference Tournament Predictions 2017 – Early Tournaments and the Cinderella Index

Thirteen of the conference tournaments kick off early and are all under way. With the lone exception of the America East tournament, they will conclude early as well beginning with the Ohio Valley on Saturday and ending with the Patriot league title on Wednesday. The rest of the tournaments kick off next Monday or after and conclude either next Saturday or on Selection Sunday itself including the first ever Ivy League tournament.

As I’ve done in years past (2013, 2015, 2016), I’ll be tracking my conference tournament predictions against those of both KenPom and Team Rankings. The predictions for the early tournaments are up here on Google drive and I’ll add the remaining tournaments in once their brackets are set.

But while we wait and see how the projections will fare, I want to explain what my rooting interests will be during Championship Week.

Introducing the Cinderella Index

Continue reading

Posted in College Basketball, Conference Tournament predictions, March Madness, predictive, team evaluation | 2 Comments