Saturday, February 9, 2019

Top 16 Projection for 2/9/19 Reveal

This afternoon the NCAA Tournament Selection Committee will be releasing their third annual top 16 team projection. The committee’s goal is to shed light on the selection and seeding process, letting the public know what factors they deem most important in determining where teams are ranked. Below is my attempt to mimic what their ranking will be, as well as what region the teams will play in. Note that this projection is reflective of results through Friday 2/8.

Teams below are ranked in order by seed.

1 seeds: Duke, Virginia, Tennessee, Michigan
2 seeds: Gonzaga, Kentucky, North Carolina, Michigan State
3 seeds: Kansas, Houston, Marquette, Purdue
4 seeds: Louisville, Wisconsin, Nevada, Villanova

Teams just missing the cut (aka the 5 seeds): LSU, Iowa State, Texas Tech, Iowa

Last season the committee implemented a four-tier stratification that adjusts for location of the game, and that process is being used again this season. The only change (and it is a big one) is that the metric used to rank the teams has changed from the RPI to the NET. The NET is a more modern model that was developed using machine learning. It includes factors such as offensive and defensive efficiency, margin of victory (capped at 10 points), and location of the game. The old RPI metric only considered the winning percentage of the opponents on a team’s schedule and adjusted for location. The NET is a much improved model and is fairly aligned with other popular rankings that are publicly available such as KenPom, BPI, and the Sagarin rankings. Below is the four-tier stratification previously mentioned:

NET Quadrant 1: Home (1-30)                    Neutral (1-50)                   Away (1-75)
NET Quadrant 2: Home (31-75)                  Neutral (51-100)               Away (76-135)
NET Quadrant 3: Home (76-160)                Neutral (101-200)             Away (136-240)
NET Quadrant 4: Home (161-351)              Neutral (201-351)             Away (241-351)

Below is an explanation of my rankings.

Duke, Virginia and Tennessee are the consensus top three teams – a case can be made for ranking them in any order. After that, I believe there are three teams fighting for the final one seed: Michigan, Gonzaga, and Kentucky. Below are the Cliff Notes versions of each team’s NET profile:

Michigan – Michigan is tied for the most combined Quad 1 and Quad 2 wins in the nation with 13, with five of those wins coming against Quad 1. They have two losses, both in Quad 1 on the road and to teams ranked in the top 25 of the NET. Last year the committee placed emphasis on road wins – Michigan is 5-2 on the road in Quad 1 and Quad 2 games and have a neutral court win in Quad 2. The biggest blemish on their profile is their strength of schedule (SOS), which ranks 69th nationally despite playing in the Big Ten conference. This is driven by their non-conference SOS being ranked 184th nationally (out of 353).

Gonzaga – The Zags once again have a solid team that will win the West Coast Conference regular season and tournament titles and will likely be rewarded with a top two seed come Selection Sunday. The hang up, as always, is that their conference is far weaker than all the other teams that are considered for one and two seeds. Despite this, Gonzaga has four Quad 1 wins and four Quad 2 wins. They own one of the best wins any team has this season – a neutral court win over Duke, arguably the best team in the nation. However, the quality of wins drops off thereafter, with the next best wins being against Washington (ranked 25th) San Francisco (two wins, ranked 52nd) and Creighton (ranked 55th). Their two losses are to Tennessee and North Carolina, and both of those teams will also likely be ranked in the top two seed lines come Selection Sunday. Gonzaga’s SOS ranks 38th nationally with a non-conference SOS ranking 50th.

Kentucky – The Wildcats have the most Quad 1 wins of the teams under consideration for the final one seed with six, and that includes wins over North Carolina on a neutral court, at Louisville, and at home versus Kansas. Kentucky’s SOS ranks 50th nationally and their non-conference SOS ranks 24th. Their blemish is a neutral court loss to Seton Hall, who ranks 67th nationally and is considered a Quad 2 loss, something Michigan and Gonzaga do not have on their resumes.

A solid case can be made for any of these teams, but I think Michigan will get the nod due to the quality of their best wins (North Carolina, Purdue, and at Villanova) and lack of a bad loss.

After those teams I think that North Carolina and Michigan State will round out the two seeds. These teams own a high volume of Quad 1 and Quad 2 wins, they just have more losses than the teams above them.

I’m projecting the three seeds to be Kansas, Houston, Marquette, and Purdue. I don’t anticipate much change here. Kansas has the #1 raked SOS, both overall and in the non-conference. As such, they have compiled eight wins in Quad 1, which is only behind Michigan State for the most nationally. They do have two loses in Quad 2 and six losses overall, but the volume of Quad 1 wins was important last year and I imagine that to be the case yet again this season. 

Houston only has three Quad 1 wins, the lowest of any team up to this point in my projection, but they have eight Quad 2 wins. This is a function of them playing in a slightly weaker conference than the other teams around them in the ranking and them scheduling a lot of home games – 16 of their 22 wins have come at home. The edge they have over other teams is they only have one loss, and it is not a bad loss (at 50th ranked Temple). 

Marquette owns seven Quad 1 wins which is impressive, but just two Quad 2 wins. They are also not as favorably ranked in the NET or KenPom rankings relative to the other teams around them in this projection, likely due to them having seven wins against Quad 4 teams which is reflective in their non-conference SOS being ranked 106. Quad 4 games are essentially games that should never be lost by a team selected to the NCAA tournament. 

Purdue owns the #2 SOS nationally, thanks in part to them playing in the Big Ten and their non-conference SOS being ranked 28th. They have six Quad 1 wins (one them being right on the cut line, at Penn State) and four Quad 2 wins. Their blemish that knocks them down is a Quad 3 loss to Notre Dame, which is also right on the cutline of being just a Quad 2 loss.

The final four teams I have making the projection are Louisville, Wisconsin, Nevada, and Villanova. I think that Louisville and Wisconsin are safely in. Louisville owns three wins over top 10 teams in the NET rankings – the most of any team. They also have the 5th rank SOS nationally. Their hold up is having just eight total Quad 1 and Quad 2 wins. Wisconsin has seven Quad 1 wins and five Quad 2 wins for 12 combined wins in that cohort – that is the second most nationally. Wisconsin’s blemish is a road loss to Western Kentucky, who ranks 123rd nationally. 

After that we come to the most intriguing resume of all the teams in this projection: Nevada. The Wolfpack have zero Quad 1 wins – that’s right, no “quality” wins. They do own eight Quad 2 wins, however, with six of those wins coming in the non-conference. That shows they attempted to schedule tough, but unfortunately for them the Pac-12 conference is way, way down this season (in fact it is likely just a one bid league). They played, and beat, Arizona State, USC and Utah. Those teams typically are in the hunt for an NCAA tournament berth, but this year that is not the case. Nevada also won at Loyola-Chicago, the Cinderella team from last season’s NCAA Tournament that made a run to the Final Four. Loyola-Chicago is projected to win their conference, but they rank 124th nationally. The other problem Nevada has is their only loss came against a Quad 3 opponent, and that loss will likely remain in Quad 3 all season. They have the worst SOS of any team considered for this projection (108th nationally) but their non-conference SOS ranks 43rd. I think the committee will factor in the combination of only having one total loss and Nevada’s attempt to schedule tough in the non-conference and reward them with a spot in the top 16, but I would not be surprised to see them just outside the top 16.

Villanova, winner of two of the past three national titles, rounds out the projection. They only have three Quad 1 wins, but they have nine Quad 2 wins. That combination of 12 wins in that cohort is what separates them from the teams beneath them. What holds them back relative to the teams above them is having two Quad 2 losses and the lack of Quad 1 wins – three total is the second lowest only ahead of Nevada.

The four teams that I have outside of the top 16 that I could see making the projection are LSU, Iowa State, Texas Tech and Iowa. LSU has a very similar profile to Villanova, but the key difference I think between them is the total number of Quad 1 and Quad 2 wins. LSU has ten to Villanova’s twelve. And, let’s be honest, there is likely to be some name recognition bias that comes into play here. Iowa State, Texas Tech, and Iowa also have similar profiles to each other – poor non-conference SOS, five total losses, similar number of Quad 1 and Quad 2 wins. The blemish on all three of these teams relative to the teams ahead of them in this projection is the non-conference SOS. The committee has historically penalized teams for scheduling a poor non-conference schedule, but that is also correlated with the lack of opportunities to pick up quality wins.

This is the second best day of the year in the bracketology universe, only behind Selection Sunday, and I’m looking forward to learning what the committee values as that will not only help rank these team but also help shape the bubble picture in determining which teams to include and which teams to exclude in rounding out the field of 68.

No comments:

Post a Comment