Final March Madness Projections 2016

Just in time for tip-off of game number one, here are the region-by-region probabilistic projections for the 2016 tournament.

This model, like many others that can be found around the internet at this time of year, uses power ratings as the core of its projection algorithm.  But it also includes three additional adjustments based on game location, tournament experience on the roster, and the coach’s track record of tournament success.

2016_projections_south

2016_projections_west

2016_projections_east

2016_projections_midwest

  • The model likes three times as title favorites: Kansas, North Carolina, and Michigan State.  North Carolina.
  • In two regions, the (2) seed has a better chance of reaching the Final Four than the (1) seed.
  • Oregon is the most vulnerable (1) seed.  Xavier is the most vulnerable (2) seed.
  • This year, (9) and (11) seeds are particularly strong.  All four (9) seeds are favored to win their first game.  One (11) seed, Gonzaga, is a healthy favorite, while two more, Michigan and Wichita State, are virtually even in their first round matchup.
  • Yale is the most likely (12) or higher seeded team to make it through the first round.
  • (14) seeds look a little more dangerous than (13) seeds this year.

Happy March Madness to all!

Coaches in the 2016 NCAA Tournament

Based on pick data from ESPN’s and Yahoo’s free bracket challenges, Kansas is the most popular choice to win the title this year, fitting for the number one overall seed. In second place, though, is Michigan State, a team that is not even a (1) seed in its own region. One possible reason for this much love for Sparty: Tom Izzo.

Coaches matter in the NCAA tournament. Obviously, any coach who makes it to the NCAA tournament is a successful coach. But historically, some coaches, like Izzo, are able to consistently navigate the bracket successfully, while others tend to come up short in the postseason. So let’s take a look at how this year’s tournament coaches have done in the past.

The goal of this article is to isolate a coach’s tournament performance. To do this, we compare each coach’s results to what might be expected given his team’s seed. As seen in my earlier post on conference performance, I constructed a model to project outcomes based only on the seeds of the two teams playing, and the distance each team traveled to the game location. The model uses data from every tournament game since 1985.

With this model, we can project the win probability and expected point differential given average coaches on both sides. This leads to two complimentary metrics: Wins Added and Points Added.

Wins Added estimates the extra number of wins a coach has accumulated over his career compared to an average coach. For example, if the model gives a team a 40% to win a game, a win would give the coach +0.6 Wins Added for that game, while a loss would cost the coach -0.4 Wins Added.

Points Added follows the same logic. If the model projects an expected point differential of -3, a win by 5 would give the coach +8 Points Added, a loss by 1 is +2 Points Added, and a loss by ten would be -7 Points Added.

So let’s take a look at how this year’s coaches stack up based on these two metrics, region by region.

South Region Coaches

South, Wins Added

South, Points AddedObservations

  • The region is marked by two coaches who made impressive debuts: Connecticut’s Kevin Ollie and California’s Cuonzo Martin. Ollie took the (7) seed Huskies to the NCAA title in 2014. Meanwhile, Martin guided (11) seed Tennessee from the First Four to the Sweet Sixteen in 2014, winning each of his first three games by double digits. It will be interesting to see how these two coaches follow up in their second tournaments.
  • Wins Added does not love the top two seeded coaches, Kansas’s Bill Self and Villanova’s Jay Wright, although Self does significantly by the Points Added metric.
  • By these measures Fran Dunphy, might be the worse tournament coach in the field. Out of all 2016 coaches, he is last in Wins Added by more than one full win, and is third to last in Points Added.
  • The matchup between Gregg Marshall and Sean Miller is yet another reason that Wichita State vs. Arizona is my favorite first round matchup this year.

 

West Regions Coaches

West, Wins AddedWest, Points AddedObservations

  • No surprise that Coach K is on top by both measures.
  • One thing to watch is how Shaka Smart’s past success translates in his first year at Texas.
  • Dana Altman is an interesting case, as he is ranked third in the region in Points Added, but has a negative Wins Added rating.
  • Oklahoma is a popular pick to take this region, but Lon Kruger is ranked last in the region by both metrics.
  • Overall, the West as a region has the most coaches with negative Points Added (seven), and is tied for the most coaches with negative Wins Added (six).

 

East Region Coaches

East, Wins AddedEast, Points Added

Observations

  • This region is characterized by a few very strong coaches (North Carolina’s Roy Williams, Kentucky’s John Calipari, and Michigan’s John Beilein) and a few very weak coaches (Notre Dame’s Mike Brey, Pittsburgh’s Jamie Dixon, and Indiana’s Tom Crean).
  • I was pretty shocked at how Roy Williams’s ratings turned out. He and Izzo are the runaway top two coaches in the tournament this year.
  • Andy Enfield’s back! Can he recapture the Dunk City magic this year?
  • West Virginia’s Bob Huggins, like Dana Altman in the West, does really poorly in Wins Added, but is above average by Point Added.

 

Midwest Region Coaches

Midwest, Wins AddedMidwest, Points Added

Observations

  • Looking at the coaches makes it pretty clear why Michigan State is such a favorite to come out of this region. Not only is Tom Izzo one of the best pure tournament coaches of all time, but top-seeded Virginia’s Tony Bennett is ranked last in the region by both metrics.
  • The potential second round matchup between Izzo and Dayton’s Archie Miller would be fun. Miller has made quite the impression in his first two tournaments, but he’s going to have a tough time getting two wins for the third straight year.
  • I was not expecting Tubby Smith to be rated so highly, considering it has been eleven years since he made it out of the first weekend of the tournament. But prognosticators might want to take a second look before they jump on the Butler bandwagon in the first round.

 

Data from Sports Reference College Basketball

Data organized and compiled using MySQL

Visualizations made with Tableau Public.

 

Conference Performance in the NCAA Tournament, Part 2

This is part two of a two-part series examining conferences in the NCAA tournament. Part one is here.

In part one, we examined the historical performance of college basketball conferences in the NCAA tournament. However, many of these conferences have changed significantly over the years. In part two, we will attempt to examine this topic in greater detail.

Simple Wins Per Year vs. Current Team Wins Per Year

To measure the effects of conference re-alignment, I compared two simple metrics. The first is the original Wins Per Year metric, as seen in part one, which measures the average number of wins each conference accumulated each year. The second metric, Current Teams Wins Per Year, is the total tournament wins produced by the current schools in the conference since 1985, divided by 31 (the number of tournaments since 1985). This comparison shows the shift in teams over time, by comparing where they were when they won games in the tournament, to where they are currently.

Power Conference Wins Per Year

Win Migration, Power.png

As we might expect, most power conference teams are winners by this measure. We see that the ACC, already the leader in wins per year, is only getting stronger after adding tournament regulars Syracuse, Pittsburgh, and Louisville to their roster. Meanwhile, the Big East took the biggest hit, losing about 40% of its yearly wins. Overall, four of seven conferences have higher CTWPY than their original WPY.

Mid-Major Wins Per Year

Win Migration, Mid-Major

Where did all those wins come from? Well, these power conferences raided the best teams from the Mid-Major ranks. We can see that most of the Mid-Major teams have lost yearly wins, with Conference USA and the WAC being hit the hardest. Conference USA has lost about 60% of its yearly wins, while the poor WAC has been almost wiped out, losing 93% of its yearly wins.

Single Bid Conferences Wins Per Year

Win Migration, Single Bid

But things look even worse for the single bid leagues. From part one, we identified the Colonial and the Horizon as the single-bid conferences that most exceeded expectations based on their seeds. However, both have been crippled by re-alignment, losing most of their yearly wins. In fact, of the six single bid leagues who have historically produced more than 0.25 wins per year, five have lost more than 60% of their yearly wins. Only the MAC has been able to maintain their lineup of successful tournament teams.

While these metrics give a good sense of how modern conferences differ, one shortcoming is that they fail to account for the changes in size for each of the conferences. For example, is the ACC getting better, or just bigger?

 

A Measure of Conference Quality

To examine this idea, I used tournament wins since 1985 as a proxy for team strength. Using this metric, we can look at the average tournament wins per team in each conference as a measure of conference strength.

Win Density

As we can see, the ACC is still the top dog. The average ACC has won just under 27 tournament games since 1985. The Big 12 and Big Ten are roughly even for second place, with about 21 tournament wins per team in these conferences. Rounding out the top seven are the SEC at about 18 wins/team, the Big East at just under 17 wins/team, the Pac-12 at 15 wins/team, and The American at 13 wins/team.

There is a large drop off after the power conferences, with 8.5 wins separating the average team from The American from the average Atlantic 10 team. We also see that the WAC, classified as a Mid-Major based on bids per year, has clearly fallen to the level of a Single Bid conference.

Conference Quality Over Time

Finally, we take a look at how this measure of conference strength has changed over time.

Wins Per Team, Power

For most of the power conferences, we see a downward trend in wins per team. This suggests that the growth of the power conferences is diluting the quality of the average team in the conference. Indeed, only the Big 12 has increased its average wins per team, and it is also the only power conference that has decreased in size.

Wins Per Team, Selected Other

When we look at conferences outside the top seven, we see the toll that re-alignment has taken on them. This is especially true over the last five years, as we see a large drop in almost every conference sometime between 2011 and 2014. Conference USA, once just short of a power conference itself, has been completely picked dry, falling from almost 13 wins per team in 1996 to slightly over two wins per team now. The Horizon and the WAC have seen similar levels of destruction, falling from peaks of a healthy 9.4 (Horizon) and 5.6 (WAC) wins per team, down to 1.2 and 0.3 wins per team in 2016, respectively.

In the first section of this article, we saw that the Mountain West improved by the Wins Per Year metric, but it appears that improvement was due to growth in size, not quality. Its average team currently has almost exactly half the number of tournament wins as the teams that founded the conference in 2000.

Overall, we see a downward trend in this metric across all categories of conferences. The shifting of teams in the top conferences has a trickle-down effect that hurts everybody. The power conferences are absorbing the best Mid-Major teams, diluting their own product in the process. The Mid-Majors are losing their best teams, and replacing them with weaker teams from Single Bid leagues.

While this year has been celebrated as a year of parity, this data suggests that as conferences shift, this parity is more likely to be confined within the best conferences, as opposed to parity between power conference and non-power conference teams.

Data from Sports Reference College Basketball

Data organized and compiled using MySQL

Visualizations made with Tableau Public.

Conference Performance in the NCAA Tournament, Part 1

This is part one of a two-part series examining conferences in the NCAA tournament. Part two can be found here.

The 2016 NCAA Tournament field has been released, and this year’s field is heavily saturated with teams from the top conferences in college basketball. The ACC, Big Ten, Big 12, and Pac-12 all placed seven teams in the bracket, and overall only nine conferences have multiple teams in the field. But historically, how have the conferences performed in the tournament? Let’s take a look.

All data in this article covers tournaments beginning with 1985, the year the NCAA expanded the field to 64 teams.

 A Simple Measure – Wins Per Year

Perhaps the easiest way to measure tournament success in a conference is by a simple average of NCAA tournament wins per year of the conference’s existence.

Wins Per Year

With the fracturing of the once proud Big East, the ACC now stands alone at the top of the charts, with about 9.5 wins per year. For this article, I will be treating the new Big East as a different entity than the original, although I have included the performance of both combined in the chart above.  The Big Ten is a clear second, followed by the Big 12 and SEC. As we move down the line, we see pretty much what we would expect: a progression from the power conferences, to the stronger mid-majors, and then into the single bid leagues.

One issue with measuring teams this way, though, is that conferences that receive more bids have more chances at wins. Similarly, conferences that get seeded higher receive get better matchups. To correct for this, we can examine how the conferences perform compared to what would be expected from their seeds. Who lives up to their seed assignments, and who falls short?

Comparing Against Expectations Based On Seeding

Using data from every tournament game since 1985, I constructed a model to project outcomes based only on the seeds of the two teams playing, and the distance each team traveled to the game location.

Then, I used two complimentary comparative metrics. How often did teams from a conference win compared to what would be expected? And how did teams from a conference do in terms of point differential, compared to projected point differentials? Win percentage against expectation is a better descriptive measure of performance, looking only at the outcome of games. Point differential captures more information about relative performance, which is helpful for spotting trends.

Finally, for this article, I will be classifying conferences into three different categories. Power conferences are the strongest seven conferences, which each send many teams to the tournament every year. The Mid-Major conferences are a group of six, who average between 1.5 and 2.9 bids each year. The rest are the “Single Bid” group, averaging less than 1.5 bids per year.

Power and Mid-Major Conferences

Against Expected Win %(1)

Against Expected Point Differential(1)

Shade of bars coincide to total number of games played by teams in the conference.

From looking at the data, we see that power conference teams tend to live up to the hype, and the mid-major teams tend to fall short of expectations.

For the power conferences, the new Big East and The American have the biggest deviations from their expected performance. In the first two years since the split, teams from the American have far exceeded expectations, while teams from the new Big East have under-performed. The American’s numbers are a bit inflated by Connecticut’s improbable title run in 2014, but it will be interesting to see if these trends continue in the future. The old Big East was one of the two most dominant conferences over the 29 years from 1985-2013, but the new Big East is not the same conference. It’s plausible that the committee overrates teams from the current iteration, due to the Big East name.

Among the other, more established power conferences, the SEC historically has the best performance against expectation, while only the Big 12 under-performs by both measures.

Meanwhile, the data suggests the selection committee consistently overrates teams from the Mid-Major conferences.   Indeed, no mid-major conference exceeds expectations in both wins and point differential. The Mountain West and Conference USA are the biggest underachievers by both measures, which does not bode well for Fresno State or Middle Tennessee this year.

Single Bid Conferences

Against Expected Win %, Single Bid

Against Expected Point Differential, Single Bid

Finally, are you looking for a Cinderella for your bracket pool? Maybe you should start your search with the five single-bid conferences that exceed expectations by more than a point per game. The Atlantic Sun, Colonial, Horizon, Ohio Valley, and Southern have historically produced scrappy underdogs. However, notice that only the Colonial and the Horizon have translated these improved point differentials into improved win percentages, winning 10% (Colonial) and 9% (Horizon) more often than average for their seed. These are the leagues that produced some of the most memorable runs in recent tournament history: George Mason in 2006, Butler in 2010 and 2011, and VCU in 2011.

But don’t jump on the UNC Wilmington/Green Bay bandwagon too quickly. As I’m sure some of you noticed, all three of those teams have moved to higher tier conferences in the past few years. Conference re-alignment has shaken up the landscape of college basketball – a topic we will examine more closely in part two.

Data from Sports Reference College Basketball

Data organized and compiled using MySQL

Visualizations made with Tableau Public.

Tournament Experience for 2016

One narrative that tends to emerge around March Madness every year is the value of experience come tournament time. When we look at the data, there is little evidence of the value of overall experience. However, tournament experience does appear to matter, as seen in this piece at TeamRankings.com, and here at the Harvard Sports Analysis blog.

So with that in mind, here is the breakdown of tournament experience for each of this year’s teams.  Seen below are total minutes logged by active players on the roster of each tournament team, accounting for injuries.

South Region(1)

East Region(1)

Midwest Region(1)

West Region(1)

Some Observations:

  • The one team that jumps out at me based on experience is Wichita State in the South. Not only are they the third most experienced team in the field, but they will get another full game under their belt by playing in the First Four (assuming they advance).
  • North Carolina is the only (1) seed that is the most experienced team in their region.
  • The top five most experienced teams are, in order, Michigan State, North Carolina, Wichita State, Dayton, and Virginia.
  • In every (8) vs. (9) matchup, the (9) seed is the more experienced team.
  • Two other first round matchups have underdogs that are more experienced than the favorites: (10) VCU against (7) Oregon State in the West, and (11) Gonzaga against (6) Seton Hall in the Midwest.
  • The West Region, considered by many to be the weakest region,  is also clearly the least experienced.

 

Data from Sports Reference College Basketball

Visualizations made with Tableau Public.