So as we close in on selection Sunday, you're going to hear a lot more from people projecting the field talking about Quad Wins and NET ratings when discussing who they think will make the field. I thought I would take some time to try and break down the selection process and provide some context for those terms, at least from my understanding and try provide some basic insight into how I believe they (and other measurements) may be used by the NCAA Tournament Selection Committee to choose the field of At-Large teams for the NCAA Men's Basketball Tournament.
The Basics of Selecting the Tournament Field
Let's start at the very beginning. Here are the basics you must know about who makes up the field.
- 68 NCAA Division 1 Teams are selected for the NCAA Tournament from among the 353 Schools who compete at the Division 1 level.
- 32 Teams are awarded 'Automatic Bids' as Champions of their conference.
- Every Conference awards it's Automatic Tournament bid to the winner of their post season Conference Tournament. The winner of regular season (Team with the best record in their conference during the regular season) gains only the advantage of their conference tournament's #1 seed.
- The remaining 36 spots are considered 'At-Large' bids, and awarded to the teams selected by the consensus of the 10 person selection committee as the best 36 remaining Division 1 basketball teams who have not earned an automatic bid.
- Once all 68 Teams have been identified, the committee ranks the teams from 1-68 (Best to Worst) and sets out to place each team into the brackets in such a way that the higher the seed, the easier the theoretical path is to win the tournament.
- 8 Teams must play in Dayton on either Tuesday, March 19th or 20th and 'Play-in' to the field of 64 teams. These 8 teams are the 4 lowest rated Conference Champions (ranked 65th to 68th) and the 4 lowest rated 'At-Large' teams (ranked 32-36 of the At-Large Teams).
If you want to read more from the NCAA Tournament on this process, I have it linked up among my links on the left margin of the Blog.
The NCAA Evaluation Tool (NET) Rankings
The NCAA NET Ratings are a tool used by the selection committee for the purposes of trying to sort out who the best to worst teams in College Basketball are to assist in selecting the 36 At-Large Teams and then ranking the entire 68 team field from 1-68 as described above. The NET rating is essentially a formula designed to weight every win and loss by every NCAA Division 1 team to come up with a common measurement to be able to compare teams.
The NET Ratings are brand new for 2019 and replace the previous tool used by the selection committee called the Ratings Percentage Index, and better known as the RPI. The RPI has been retired and is no longer calculated and/or used by the NCAA to assist in the sorting and evaluation process.
For a deeper understanding of the NET Ratings, this is a good primer provided by the NCAA:
The general consensus in the past was the RPI was a flawed evaluation tool. The biggest complaint was that the tool was weighted in such a way that it rewarded teams too much for playing tougher games, even if the team won few, if any, of their toughest games. The hope in replacing it with the NET, and its formula, was that it would be easier for the Selection Committee to be able to evaluate teams who played in mid-major and small conferences more fairly against teams from the Power 6 Conferences (ACC, Big 12, Big 10, SEC, PAC12 and Big East) who have the advantage of playing considerably more games against high quality opponents than their smaller conference counter parts. Time will tell if this weighting tool is more effective than RPI or not.
Quad Breakdowns
The idea of 'Quads' were introduced into the evaluation process last season and are designed to provide the selection committee with a tool for evaluating the value (or lack thereof) of wins, and the significance of losses. The Quads attempts to equalize the value of a particular game by looking at who played the game (what were their NET ratings) and where was the game played.
The basic premise is that the higher the team is rated, the more it signifies the quality of the win. But not all wins are created equal, depending on where you play. If you beat the 65th rated team at home, its not nearly as impressive as beating them on their home floor, or on a neutral court.
To try and better quantify this, the NCAA developed 'Quads' ... Four to them to be exact. They then grouped the quads so that you could measure a teams wins and losses based on what quad the game fell into. Quad 1 represented the most impressive wins (Based on NET Rankings), while Quad 4 wins represent the least impressive wins. Conversely, Quad 4 losses are considered to be the 'Worst' losses, while Quad 1 losses are considered the least damaging. To put in terms of evaluating two teams side by side in simplest terms, a team that had a 20-1 record who had a Quad 1 record of 15-1 and was 5-0 against Quad 4 teams would be considered considerably better on paper than a 20-1 Team that hadn't played a Quad 1 game and was 15-1 against Quad 4 Teams.
Below is the Quad Breakdown being used this season by the NCAA:
| Home | Nuetral | Away | |
| Quad 1 | 1-30 | 1-50 | 1-75 |
| Quad 2 | 31-75 | 51 - 100 | 76 -135 |
| Quad 3 | 76-160 | 101-200 | 136-240 |
| Quad 4 | 161+ | 201+ | 241+ |
So, let's say team A is ranked 31st and Team B is ranked 65th and play a game:
- If Team A plays AT Team B's court and wins, Team A would be credited with a Quad 1 Win and Team B would be credited with a Quad 2 Loss
- If Team A Played Team B at Home and Lost, Team A would be charged with a Quad 2 Loss while Team B would be credited with a Quad 1 Win.
- If Team A and B played on a Neutral Court and Team A won, they would be credited with a Quad 2 Win, and Team B would be credited with a Quad 1 loss.
Net Rankings are re-calculated daily, and team's NET ranking will rise and fall depending on who they play, who they beat or lose to, and where the games are played. As a result, what may have been a Quad 1 win 2 weeks ago, may be rated a Quad 2 or lower today if the team the team's NET ranking has dropped. Conversely. what may have been a Quad 3 Win a month ago, could be a Quad 1 or 2 win today if that team they beat has moved up in Net Ranking. It is all very fluid, and the outcome of every game played has impacts on every single team's NET ranking as a result.
To assist the committee, the NCAA provides a summary sheet for all NCAA Division 1 teams, which provides a wealth of data designed to aid in the comparing and sorting of teams for the purposes of ranking and seeding. At heart of the team sheet is the results of each game played in each quad, and a summarized view of those results. Below is an example:
This is the 3/4 team sheet for Wisconsin. Among the data provided is a summary of their record against each quadrant of teams, and further disection within each Quadrant of record at home, away, on a nuetral court, etc. I have highlighted this data in yellow on the top of the sheet. Below that is a breakdown of each game played in the quadrant, the opponents NET ranking, location, oppponent name, score and date played in descending NET ranking order. For example, the first game listed for Wisconsin was a Nuetral Court game played against 2nd Ranked Virginia which the Badgers lost on November 23rd. You will also notice that that within Quad 1 and Quad 2, the results are further grouped into tiers of games within the Quad. Highlighted in Yellow are the groupings of those tiers. Looking at tiers commonly referred to as Quad 1A and Q1B, you will see they are classified the following way. Quad 1 A wins would be Home wins against teams ranked 1-15, Nuetral court wins against teams 1-25, and Road wins against teams ranked 1-40. The idea is that a win against a 9th on your home floor (like Wisconsin beating Michigan) counts the same as a win on the road against the 40th ranked team (Iowa).
Without this tier of groupings, Quad 1 wins would be measured more broadly, such that a win at home against a 9th ranked team counted the same as a road win against the 70th ranked team (Xavier). These tiers allow the selection committee to easily identify higher quality Quad 1 or Quad 2 wins from lesser quality wins within the Quad. It would be hard to argue that beating Xavier on the road is as impressive a win as beating Michigan at home. However beating Iowa on the road is probably a better comparison.
Here is a further breakdown of the Quads which includes the A and B tiers found in Quads 1 and 2:
Confused yet? Me too ...
Putting the data into Practice
The development of tools like the NET rankings, and then establishing Quads to group wins and losses for each team, allows for the comparison of one team to other teams. It's far from perfect and in the end, the judgement of what the data actually means is completely subjective. But it does make the job simpler, and a bit more standardized, at least in theory.
Now, Lets take a look at what you often hear pundits talking about as we get closer to Selection Sunday ... something called 'Blind Resume's'. The blind resume' is a simple example of how you might compare data between teams, without knowing who they were, and attempt to draw a conclusion of who was more worthy of being ranked higher than the other. Take the following with data from approximately a week ago:
Each column represents the data for a single team. Each row is a specific data point to be used to compare teams.
The Blue team has a NET ranking of 34, a record against Division 1 teams of 22-6, a record in their conference of 9-3 and so on. All of this data can be gleaned from each team's NCAA Team Sheet. The final row is a data point known as 'Record vs. Field', which represents the record of the team against every other team who is under consideration for a bid and is designed to help the committee determine how a particular team has fared (or might fare) against other tournament quality teams.
In the end, what you may value as the most important data points, and what is the least important, are largely irrelevant when trying to project what teams will or will not qualify for the NCAA tournament. What matters is what the selection committee values. My goal is not to pick what I think are the best teams ... it's to pick who I think the NCAA Selection Committee will think are the best teams, and in what order they will rank them.
I am of the opinion these are among the most important data points, and so they tend to be the data points I look at first among the 100+ metrics I track. When I can't sort it out from these, I start looking at others to try and break ties. But in the end, each committee is unique, and will tend to emphasize some metrics one year (Road Wins against the best teams might be one) and the next, it may seem less important and what may be more important is the strength of the schedule a team played, and the record they earned doing it, particularly in non-conference games which is the portion of the schedule teams largely control. The big question they may say this answers ... 'Did the team challenge themselves?'. The committee has often punished teams with good records that piled up a bunch of wins in November and December playing poor opponents at home. Twenty wins used to be a bench mark for being considered 'Tournament Worthy'. Now, if you got 20 wins playing bad teams, you probably aren't going to get an at-large bid. That's why doing may be really frustrating one year and really rewarding the next. The one thing I never find it to be though, is boring. It's definitely an art, not a science, despite people using scientific sounding terms like 'Bracketology' to describe the activity.
To finish, take a look at the resumes above again and determine how you would rank those teams. Then look below to see who they were. Would you change how you rank them knowing who they are? I bet many of you would have ranked these teams differently if you had known from the beginning who the teams were. This is the point of trying to do the comparison blind as much as you can. It helps to filter out built in biases which influence rankings when a team's identity is known.
For the record, at the time I put this together, this is how I ranked them against each other:
This is the 3/4 team sheet for Wisconsin. Among the data provided is a summary of their record against each quadrant of teams, and further disection within each Quadrant of record at home, away, on a nuetral court, etc. I have highlighted this data in yellow on the top of the sheet. Below that is a breakdown of each game played in the quadrant, the opponents NET ranking, location, oppponent name, score and date played in descending NET ranking order. For example, the first game listed for Wisconsin was a Nuetral Court game played against 2nd Ranked Virginia which the Badgers lost on November 23rd. You will also notice that that within Quad 1 and Quad 2, the results are further grouped into tiers of games within the Quad. Highlighted in Yellow are the groupings of those tiers. Looking at tiers commonly referred to as Quad 1A and Q1B, you will see they are classified the following way. Quad 1 A wins would be Home wins against teams ranked 1-15, Nuetral court wins against teams 1-25, and Road wins against teams ranked 1-40. The idea is that a win against a 9th on your home floor (like Wisconsin beating Michigan) counts the same as a win on the road against the 40th ranked team (Iowa).
Without this tier of groupings, Quad 1 wins would be measured more broadly, such that a win at home against a 9th ranked team counted the same as a road win against the 70th ranked team (Xavier). These tiers allow the selection committee to easily identify higher quality Quad 1 or Quad 2 wins from lesser quality wins within the Quad. It would be hard to argue that beating Xavier on the road is as impressive a win as beating Michigan at home. However beating Iowa on the road is probably a better comparison.
Here is a further breakdown of the Quads which includes the A and B tiers found in Quads 1 and 2:
Confused yet? Me too ...
Putting the data into Practice
The development of tools like the NET rankings, and then establishing Quads to group wins and losses for each team, allows for the comparison of one team to other teams. It's far from perfect and in the end, the judgement of what the data actually means is completely subjective. But it does make the job simpler, and a bit more standardized, at least in theory.
Now, Lets take a look at what you often hear pundits talking about as we get closer to Selection Sunday ... something called 'Blind Resume's'. The blind resume' is a simple example of how you might compare data between teams, without knowing who they were, and attempt to draw a conclusion of who was more worthy of being ranked higher than the other. Take the following with data from approximately a week ago:
Each column represents the data for a single team. Each row is a specific data point to be used to compare teams.
The Blue team has a NET ranking of 34, a record against Division 1 teams of 22-6, a record in their conference of 9-3 and so on. All of this data can be gleaned from each team's NCAA Team Sheet. The final row is a data point known as 'Record vs. Field', which represents the record of the team against every other team who is under consideration for a bid and is designed to help the committee determine how a particular team has fared (or might fare) against other tournament quality teams.
In the end, what you may value as the most important data points, and what is the least important, are largely irrelevant when trying to project what teams will or will not qualify for the NCAA tournament. What matters is what the selection committee values. My goal is not to pick what I think are the best teams ... it's to pick who I think the NCAA Selection Committee will think are the best teams, and in what order they will rank them.
I am of the opinion these are among the most important data points, and so they tend to be the data points I look at first among the 100+ metrics I track. When I can't sort it out from these, I start looking at others to try and break ties. But in the end, each committee is unique, and will tend to emphasize some metrics one year (Road Wins against the best teams might be one) and the next, it may seem less important and what may be more important is the strength of the schedule a team played, and the record they earned doing it, particularly in non-conference games which is the portion of the schedule teams largely control. The big question they may say this answers ... 'Did the team challenge themselves?'. The committee has often punished teams with good records that piled up a bunch of wins in November and December playing poor opponents at home. Twenty wins used to be a bench mark for being considered 'Tournament Worthy'. Now, if you got 20 wins playing bad teams, you probably aren't going to get an at-large bid. That's why doing may be really frustrating one year and really rewarding the next. The one thing I never find it to be though, is boring. It's definitely an art, not a science, despite people using scientific sounding terms like 'Bracketology' to describe the activity.
To finish, take a look at the resumes above again and determine how you would rank those teams. Then look below to see who they were. Would you change how you rank them knowing who they are? I bet many of you would have ranked these teams differently if you had known from the beginning who the teams were. This is the point of trying to do the comparison blind as much as you can. It helps to filter out built in biases which influence rankings when a team's identity is known.
For the record, at the time I put this together, this is how I ranked them against each other:
- Temple
- Oklahoma
- Utah State
- Georgetown
I could make a case for ranking them about 10 different ways, and so could most of the people who looked at this data and had some knowledge of the teams. The point is though, that this ranking is based on what I think the Committee would do when asked to rank them. That is what bracket projection is when it's boiled down to it's simplest form.
So there you have it ... a primer on NETS and QUADS, blind compares and using the data the NCAA makes available to try and pick the field, rank it, seed it, and put it into brackets. There's a ton more ... we didn't even touch on KenPom efficiency metrics, the Sagarin ratings, the 'eye test', unbalanced conference schedules, or host of other factors that are known to play some kind of role in how the committee evaluates and compares teams. Hopefully if you have an interest in understanding any of this stuff better, it helped. The links I have provided on the right margin of the blog may help too.
If you took the time to read this and have thoughts or questions I would love to hear them.
No comments:
Post a Comment