Is Experience Overrated? (Hoop Vision Weekly: 10/11/19)
A look at the relationship between returning minutes and yearly improvement
Welcome back — or for those who are new around here, welcome aboard! — to the Hoop Vision Weekly.
We have 25 days until the college hoops season officially tips off, with plenty of secret scrimmages and exhibition games leading up to November 5th.
Remember, remember the Fifth of November!!
And since we have 25 days until the season tips off, you may assume we’ll use the Weekly today to launch some gimmicky “Top 25” countdown. Well — spoiler alert — we are not doing that.
Still, we’re deep in the muck of preseason preview land; this week saw plenty of conference Media Day events, with the requisite preseason predictions and awards that come out of those events.
This week’s feature is related to those “prediction” mechanisms, taking some data and addressing some commonly-held beliefs, but more on that in a minute.
In this edition, we’ll also take a look back at the highlights from Hoop Vision PLUS — our premium subscription product which launched two weeks ago. As a bonus, we also unlocked one of the HV+ articles from Week 1, bringing it out from behind the paywall to give you a deeper feel of what subscribers are reading and viewing.
For those who have already subscribed to Hoop Vision Plus, thank you SO much for your early support. If you’re just hearing about HV+ for the first time now, click/tap here for the full explanation on what it is and why it’s happening.
Enough introduction. Let’s go!!
Returning Minutes
Nearly every preseason Top 25 list comes with a line for “key returners” and a line for “key losses.” The person generating the list uses intuition to determine — given those returners and losses — what to expect from that team in the next season.
That process usually leans far heavier toward art than science. But thanks to Bart Torvik’s self-titled website and its list of returning minutes for every team, we can take a slightly more data-driven approach.
Illinois-Chicago leads the way with 99.7% of last year’s minutes coming back for 2019-20. For the high-major conferences, there are just five teams with at least 80% of minutes returning:
Colorado (94%)
Notre Dame (89%)
Seton Hall (85%)
Maryland (83%)
Illinois (82%)
Colorado and Notre Dame lead the way in roster continuity — with nearly everyone coming back.
But among teams to earn a bid in last year’s NCAA tournament, Seton Hall and Maryland sit in a class of their own. The coaches appear to agree with that assessment. Seton Hall was picked to win the Big East by a single vote over Villanova. Maryland was picked second in the Big Ten — only behind likely national preseason #1 Michigan State.
The graph below plots returning minutes on the y-axis and last year’s adjusted efficiency (from KenPom) on the x-axis.
The graph pretty clearly shows the logic behind Michigan State as the (mostly) consensus preseason #1. The Spartans return the most minutes of any team from last year’s top 10 — with Purdue and Michigan nearly tied for second.
Obviously we like to go a little further with the numbers in this space. And when it comes to returning minutes, the natural question to ask is:
How much do teams returning the majority of their roster from the previous season actually improve?
To examine that question, we used KenPom’s “minutes continuity” metric that dates back to the 2007 season. In that time period, 425 teams have distributed at least 70% of their minutes to players that were on the team in the previous season.
The following graphs plot the offensive and defensive performances of those 425 team from the previous season (x-axis) to the next season (y-axis):
302 out of 425 teams (71%) improved offensively
253 out of 425 teams (60%) improved defensively
If a point is above the diagonal line, it represents a team which improved on that side of the ball in the following season. If a point is below the diagonal line, it represents a team that regressed on that side of the ball in the next season.
As you can see, there are two key differences between the offensive graph and the defensive graph.
Teams returning the vast majority of minutes are more likely to improve on offense than defense
Defense is less predictable than offense from season-to-season
In fact, projecting defense gets even more interesting if we only focus on “good” defenses. Of the 425 teams in the sample, 161 of them recorded an adjusted defensive efficiency of under 100 (e.g. allowing less than 1 point per possession) in the previous season.
Of those 161 teams — which were returning nearly everyone from the previous season — 56% of them actually declined on defense the next season.
That’s not exactly great news for teams like Seton Hall and Maryland who are looking to take a big step forward this season. Still, it certainly doesn’t mean a team returning “everyone” can’t take that step forward, but it is some evidence that we may have a tendency to overrate the power of an extra year of experience.
This week on Hoop Vision PLUS
We had three new pieces this week for HV+ subscribers, continuing the preseason cadence of going deep on the teams, trends and curiosities making an impact on college hoops at large.
As we do each week, we’ll clip up some excerpts from each HV+ piece and share them in this space.
On Sunday, we tried something a bit different, adding a layer of transparency to the HV+ operation with our first “business update” explaining a bit about how the launch week had gone, what the subscriber breakdown was looking like, and how Week 1 stories performed. (Click/Tap here for full post - “Week 1, By The Numbers”)
Then we said goodbye to the boardroom and got back to the basketball court on Monday, with a breakdown on shot selection:
Full Post: “Why the best teams don’t take the best shots”
At the college level, half-court shot attempts get classified into six different bins. Those six are listed below, along with their associated points per shot (using D1 averages from the 2018-19 season):
Reminder: All shooting statistics in this article are from half-court possessions only
Around Basket (not Post-Ups) — — 1.10 PPS (33% of shots)
Post-Ups — — 0.93 PPS (8% of shots)
Runner — — 0.72 PPS (6% of shots)
Short Jumper (<17') — — 0.75 PPS (7% of shots)
Medium Jumper (17' to <3 point line) — — 0.71 PPS (6% of shots)
Long Jumper (3 point jump shots) — — 1.03 PPS (40% of shots)
As expected, threes and layups are the two shot types with an expected value of over 1.0 points per shot. Post-ups are actually the third most efficient shot type — for reasons we outlined here.
Using these average expected values, we can calculate a “shot selection” metric for every NCAA team. This metric doesn’t consider the actual results of the shot, just the selection. For instance, all three-point attempts are universally counted as 1.03 points — regardless of the team or shooter.
The top five offenses in shot selection from last season are listed below:
Penn
Belmont
Winthrop
Quinnipiac
FIU
The list is full of mid majors, which isn’t necessarily unexpected. So here is the list of the top five high major offense in shot selection:
Pittsburgh (#6 overall)
Stanford (#9 overall)
Vanderbilt (#10 overall)
Auburn (#12 overall)
LSU (#21 overall)
Auburn and LSU were both very good last season, but overall this isn’t exactly a list of elite offensive efficiency teams. Which begs the question:
Why aren’t the best teams taking the best shots?
And on Wednesday, a video breakdown + notes on Purdue and the “Zoom” action which allowed the Boilermakers to be so good last year on the offensive end…and raises uncertainty around how they’ll look this year.
Full Post: Purdue “Zoom” Action
During last season, we briefly covered some of the key actions behind Purdue’s offense. Still, we probably didn’t give the #4 ranked offense in adjusted efficiency the spotlight it deserved. Today, we do just that.
The effectiveness of Purdue’s offense revolved heavily around a particular type of action — to such an extent that the action warrants its own vocabulary. We decided on the term “Zoom” action, which is explained in detail in the 6-minute voiceover video below covering Purdue’s offense.
This week’s unlocked HV+ post
Yesterday, we threw out this poll on Twitter, requesting votes on which Week 1 post should be unlocked; by the slimmest of margins (literally one vote), it was decided.
The winner: the Virginia Cavaliers — defending their national title with a FAR more important distinction as the inaugural winning subject of the HV+ unlock poll.
So, enjoy this now-unlocked (for one week) HV+ breakdown of the Virginia offense: