Six styles of team offense in men’s college basketball

Six styles of team offense in men’s college basketball

Using a clustering algorithm to sort teams based on the ways they try to score points

Styles make fights.

It’s a boxing axiom but you hear people apply it to basketball as well. The assertion is that the relative quality of two opponents – in an abstract sense – is less important than how they match up against each other. Nobody wants to be the hulking slugger who is too slow for his counter-punching challenger, or the glass-jawed finesse artist stuck in the ring with a brawler. The question is – do our weaknesses align with our opponent’s strengths? And it’s a concern that is heightened in a single-elimination tournament where one nightmare matchup can have your team packing its bags and heading back to campus.

What happens if a big, plodding, post-heavy team faces off with a diminutive, nonstop, motion offense – who has the advantage?

I’m not actually sure. We might have some intuition about how this (or other) style face-offs tend to play out – and we could test those hypotheses using historical data – but, first, we need to define which styles of offense are out there in the current basketball landscape.

To help explore the different approaches that college coaches are using on offense, I sorted the 358 NCAA Division I men’s teams based on the actions they have used to try to score points this season (using the Gaussian Mixture Models clustering algorithm via the mclust package in R). I also included data from the 30 NBA teams in the sorting process, as a point of reference. Specifically, I used Synergy’s play-type and shot-type frequencies (i.e, “%TIME” stats) to characterize each team’s half-court offensive style.

How often has this team tried to score in isolation? How much of their offense has come directly from passes out of the post? How many unguarded catch-and-shoot jumpers have they taken?

I ended up with 27 of these style-defining features that, collectively, offer a thorough description of how a team tries to score. I fed these 27 inputs to the clustering algorithm and let it sort the offenses into groups. Here’s one way we can visualize the different styles:

In these charts, teams with lots of pink use pick-and-roll actions more than their competitors – to set up the ball handler for his own shot; to open the screener for rolls to the hoop, pops, and slips; or to create a scoring chance for another teammate who is cutting or spotting up. Teams with an abundance of purple in their circles use isolation action more than others – with the ball handler either trying to score himself or setting up a teammate for a shot. Teams with big green bars use a lot of off-ball movement – creating scoring chances with off-ball screens, handoffs, and cuts. Teams with broad swaths of blue get more spot-up chances than the rest. And teams flooded with orange tend to use lots of interior actions – scoring and passing out of the post, making flash cuts, and crashing the offensive glass more than other teams.

In this post, we are going to look at what happens when you ask the computer to create six groups of teams with six different offensive styles. There is nothing definitive about the number six. We could have asked the algorithm for 16 offensive styles (or 60!) and it would have returned different cluster assignments delineating smaller groups of teams with more refined similarities. But six is good enough. It gives us a sense for some of the bigger distinctions we might draw between the prevailing offensive approaches and sheds some light on differences in style of play that exist between the 32 DI conferences as well as emphasizing how offensive styles differ between the collegiate and NBA levels.

Here’s a look at the average team from each of the six clusters:

You can see the defining characteristics of each group in the charts above. Cluster 1 has lots of green (off-ball movement) and not much pink (PNR actions); Cluster 2 has healthy chunks of orange (scoring and passing out of the post) and blue (spot-ups and threes) and very little purple (short- and off-the-dribble jumpers); Cluster 3 is sort of the mirror image of Cluster 2 – with lots of purple (iso scoring, short- and dribble jumpers) and not much blue (not many open catch-and-shoot jumpers); Cluster 4 has lots of pink (PNR passing to rolls, pops, and slips) and blue (spot-ups, catch-and-shoot, threes) but not much orange (post); Cluster 5 is dominated by orange (scoring and passing out of the post, flash cuts, and putbacks); and Cluster 6 has the biggest pink (PNR) and purple (iso) bars with stubby orange (post) and blue (spot ups) ones.

We can check out a few of the teams from each group, starting with Cluster 1:

The movement offenses

You can see right away that – although these teams are all in the same group – none of the five examples is a perfect match with the cluster average and no two charts are exactly alike. Indeed, there are several different offensive principles lumped together in this group. Whether it’s the Princeton offense, 5-out motion, mover-blocker, continuity or flow – what all these teams have in common is that they generate more of their looks with off-ball movement (screens, handoffs, and cuts) than most other teams. The other side of the coin is that these five offenses use on-ball screens less than most teams, with the lack of PNR-ball handler scoring particularly noticeable.

It’s fun to see the Golden State Warriors screen-heavy offense pop up here next to 65 college teams in the movement offense group. All the other NBA teams were clumped together in a different group – more on that in a minute.

Next up: Cluster 2:

The post-and-space offenses

Aside from Wyoming, these teams do not possess an overwhelming amount of low-post scoring but they do each have at least one productive post-up threat – with guys like Eric Dixon, Jason Carter, Ben Vander-Plas, Matthias Tass, and Ryan Davis each racking up 50+ scoring chances in the post already this season. Moreover – while they may not score a ton in the post – these teams do play through the post quite a bit. In fact, Wyoming, Saint Mary’s, Ohio, and Villanova are all top-5 in the number of scoring chances created directly from a pass out of the post this year. And the defining feature of this group is how these teams mix that post play with three-point jump shooting, as four of these five post-and-space teams – Vermont, Villanova, Wyoming, and Ohio – are also around the 90th percentile in 3PA rate.

You might have noticed that I haven’t said anything yet about pace. That’s because I didn’t include a team’s fast-break rate as a style-defining feature, here. I was more interested in characterizing how teams try to score once it’s time to run the offense, in half-court situations. As a result, the average pace of play across the six clusters was pretty consistent; but the post-and-space group was slightly slower than the rest. Wyoming and Villanova, for example, are both bottom-10 in transition possessions per game.

Next up: Cluster 3:

The guard-heavy offenses

I wasn’t sure what to call this group. It’s true that there is a slew of high-scoring guards here – with perimeter players like Jordan Shepherd, Taz Sherman, Sean McNeil, Jordan Walker, Quan Jackson, Matt Bradley, Trey Pulliam, Lamont Butler, Caleb Murphy, Javon Greene, Jamir Chaplin all scoring double digits for one of the five highlighted teams. But guard dominance is not really what sets this group apart from the rest of college basketball. The defining feature of this cluster is the balance of 3-point vs. 2-point jumpers. These five teams are all in the bottom-50 for three-point attempt rate among DI teams and they are 85th percentile or higher for two-point jumper rate.

Next up: Cluster 4:

The drive and kick offenses

Here we find five teams who get more offense from pick-and-roll passes than most other DI teams – each one ranking 97th percentile or higher in that category. The characteristic pink and blue pattern of this group suggests the use of a lot of spread ball screens. Alabama was one of five SEC teams that landed in this drive-and-kick group.

Next up: Cluster 5:

The post-heavy offenses

Stephen F. Austin, Utah Valley, and Indiana are all top-15 nationally in post scoring this year (at 10+ per game). Arizona and Memphis post-up less frequently but they get lumped in here because of the other orange stuff they do – with their high rates of flash cutting and offensive rebounding. Seven different Big Ten teams were in this post-heavy group.

Finally, Cluster 6:

The pro style offenses

This group included the Dallas Mavericks, the Brooklyn Nets, the Portland Trail Blazers, and 26 other NBA teams (with the Warriors – who we saw earlier in the movement offense group – being the one exception). The pro style offenses are characterized by relatively high amounts of pick-and-roll actions and isolation. NBA teams also tend to use handoffs more than their collegiate counterparts. Alongside of all these NBA teams, there were also 13 college teams in Cluster 6, including Detroit and Kansas State.

We can line up the five examples from each group in one big summary graphic to get a sense for the similarities and differences in offensive styles within and between the six clusters.

One fun takeaway from this sorting process is that the mix of offensive styles varies between conferences. Half of the teams in the Big Ten were in the post-heavy group, half of the Ivy League was in the movement group, and half of the ASUN Conference was in the drive-and-kick group. Moreover, there were some offensive styles that were unevenly distributed between the big programs and the smaller ones. Only 3 of the “power 6” teams (if you want to call them that – I’m talking about the 76 teams from the ACC, Big East, Big Ten, Big XII, Pac-12, and SEC conferences) were in the post-and-space group (that’s just 4%), whereas 44 of the 282 non-power 6 teams (16%) fit in there. Here’s the full breakdown of cluster counts by conference.

OK – we have succeeded in sketching out some loose definitions of six different offensive styles being used by college basketball teams this season. And styles make fights, remember. So, our next step will be to look at how these different styles have performed against each other in the past to help predict what might happen in future matchups. But that’s a topic for a future post!


Todd is building tools to help coaches, scouts, and players find winning team strategies as part of Synergy’s Analytics and Insights Team. He creates inviting infographics, engaging charts and interactive displays that make data compelling and accessible. Follow him on Twitter @crumpledjumper.


Synergy NBA Heat Check


Synergy Sports Backing Elite Women’s Wheelchair Basketball