Hoop Vision Weekly: Ball Screen Essentials (7/7/19)

Life, liberty, and the pursuit of 1.2 points per possession

Whether you were in the film room evaluating talent, in Las Vegas evaluating talent, or evaluating your neighbor’s talent in setting off fireworks…we hope it’s been an enjoyable and memorable July 4th week.

The past few weeks of the Hoop Vision Weekly have been devoted to Princeton and motion style offenses. This week it’s all about ball screens.

In this edition we’ll explore:

  • The data behind ball screen: efficiency, usage, and decision making

  • The team—and player—that faced the most aggressive ball screen coverage in the country this past season

  • A look back at a game from my New Mexico State days in which we decided to use the most passive ball screen coverage possible

Ball Screens, by the numbers

Not all actions and plays are created for the same purpose. A team might run an action for a lob immediately out of a timeout. If the lob’s not there, they move onto the next action.

Those quick hitter type “home run” plays are likely to be quite efficient, but that is largely due to selection bias. If the lob action is anticipated and defended properly, the lob likely won’t be thrown; thus we’d expect a team to score at a high rate on lobs, given the information that the passer actually decided to throw the ball.

This selection bias also contributes (at least somewhat) to the relative inefficiency of isolation plays. These plays tend to occur after the initial, desired action has stalled or been denied. An inefficient play type would become even more inefficient under the context of a late shot clock.

Ball screens are somewhere in between. You will certainly see them at the end of the shot clock as a last resort. But you will also seem them in transition (drag screens), where efficiency tends to be higher.

Just like the lob play example, ball screen rolls are almost inherently going to be efficient offense. Generally speaking, the handler/passer is only going to throw the pocket pass if it’s open.

With those caveats in mind, let’s jump into the data.

I went back through the 2018-19 season and calculated every team’s points per play (efficiency) in two different ball screen scenarios:

  • Any play ending with a ball screen that creates a shot for the ball handler

  • Any play ending with a ball screen that leads to the handler passing and in turn directly creating a shot for a teammate

Here’s the efficiency, by program, in those situations:

The teams above the line were more efficient when the ball handler passed out of ball screen situations. And vice versa for the teams plotted below the line. If the logo is right on that line, it means the team is just as efficient in each situation.

Overall, 250 of the 353 (71%) were more efficient in passing situations.

The top two teams in passing efficiency, North Carolina and Michigan, were both top 10 teams this past season.

The top team (by a wide margin) in handler efficiency was Charleston, paced by Grant Riller - who led the country in eFG% on shots out of ball screens at 63.2%.

Interestingly enough, there’s no correlation between the two efficiency metrics.

Now let’s take a look at the usage version of the same graph. While the above plot showed who was most efficient in those situations, we now look at which programs made ball screens an integral part of their offensive gameplan:

The first thing that jumps out is the correlation in usage. And that makes sense - being able to both pass and shoot out of ball screens is what makes them hard to defend.

There are some extremes on the graph though. The teams below the line featured shoot first lead guards with the ability to shoot long threes in ball screen situations: Hofstra’s Justin Wright-Foreman, Marshall’s Jon Elmore, and Marquette’s Markus Howard.

The most extreme of all is a team above the line: Siena. We’ll give them a full section down below.

If we combine the team-by-team data for an NCAA D1 average, we get the following table of ball screen decision making and efficiency.

While it is fairly safe to say that defenses should want to keep the ball in the handlers hand and force him to shoot in ball screen situations, it’s also oversimplifying. This isn’t a binary decision making process like deciding to run or pass in football or deciding to bunt or hit in baseball.

Instead of picking a discrete option, the defense picks a coverage. We’ve seen the NBA shift towards coverages (drop and ice) that increase the probability of the ball handler shooting.

In the two sections below, we’ll take a look at the opposite ends of the ball screen coverage spectrum.

The Most Blitzed Team in the Country

In Jamion Christian’s first (and only) season as the head coach at Siena, he implemented one of the most unconventional styles of play in the NCAA.

The first factor contributing to this unique style was personnel. Siena had a breakout freshman star in Jalen Pickett who led the way as the primary creator. Besides Pickett, there was an extremely limited amount of offensive creation on the roster.

That led to a surge in usage rate for senior big Evan Fisher, mostly by necessity. Fisher served as the secondary playmaker to Pickett - mostly on rolls, pops, and some post-ups.

Christian decided to play slow. Siena was dead last in the country in average possession length. They slowed the game down and ran a set almost every possession. Christian said at one point their playbook contained 362 plays.

The majority of those sets were some type of false action to ultimately get a Pickett-Fisher ball screen. They were slow developing - opting for good spacing and allowing Pickett to make reads instead of pace.

Pickett averaged 21 ball screens per game that directly led to a Siena shot.

You can certainly argue that the pace allowed Siena to maximize their personnel, but the nature of the offense made it very easy on opponents to get into their ball screen coverage. That’s essentially because there was no element of surprise… a ball screen was coming for Pickett.

No other player saw more aggressive ball screen coverages (blitz, trap, hard hedge) than Pickett. The image below features the results of four ball screens (all from different games) that were set around the three-point line.

That two-hand fadeaway jump pass became a common sight. The idea here by the defense is pretty simple: Get the ball out of Pickett’s hands.

As a result, over 10% of Siena’s offensive plays were ball screen rolls/pops. Not only was that the highest in the country this season, but the highest overall since these things have been tracked (and, given the increase in ball screen usage, likely the highest of all time).

That blitzing certainly played a factor in Evan Fisher’s usage jumping from 17% to 31%. And is why Jalen Pickett led the country in ball screen assists to the roll man.

The Game We Played Ball Screens 2-On-2

Going back to my time on staff at New Mexico State, our second game of the 2017-18 season was a road game against Saint Mary’s — featuring All-American center Jock Landale in his senior season. The Gaels were loaded with shooting around him, plus All-WCC point guard Emmett Naar.

Saint Mary’s was another slow-paced offensive team looking to space the floor and use ball screens and post-ups to find efficient shots; they finished 11th in the country in adjusted offensive efficiency that season.

Nearly every role player they had was an above average three-point shooter; all were players that you need to take away the attempt from. Which, of course, would put defenses in nearly impossible situations when Naar and Landale were running ball screens.

We (NMSU) had a 5-foot-9 point guard, usually preventing us from just switching the ball screen. Landale (6-foot-11) was maybe the best back-to-the-basket player in the country that season; if we we put a 5’9 defender, on him we were going to have to eventually send two to the ball and give up threes.

Naar was an extremely good player, but had a pass-first approach with funky shooting form. Granted, he shot 41% from three in his career—but he took nearly twice as many twos than threes.

We felt like Saint Mary’s was at their best when they were using their talent to generate open threes for others. With switching out of the question, that meant trying to force Naar to beat us as a scorer in ball screens.

Our ball screen coverage for that game called for the three players NOT involved in the ball screen to essentially give NO help. No tagging. No squeezing the roller.

The results? Well, if the goal was simply forcing Naar to beat us, the approach was certainly successful. He took 15 two-pointers, the second most in any game of his Saint Mary’s career.

Unfortunately for us, he made nine of those shots and Saint Mary’s scored 1.41 points per possession.

Three of those nine baskets were simply straight line drives to the rim for Naar, almost completely unimpeded. But even when our guards did a better job of fighting through the screen and staying even, Naar made us pay, as he used his size to finish (and never bailed us out by settling for mid-range looks).

There are some things we could’ve tweaked in our coverage—and did tweak in the second half, before eventually playing some zone. We tinkered with going over/under the screen and putting more size on Naar.

But we also only have a 40-minute sample to go off. The bottom-line results were decidedly poor, but who’s really to say if our process and philosophy were truly sub-optimal?

The larger point to be made here is on a macro level, around creating gameplans. Whether you are using analytics or traditional scouting, it can be easy to determine what not to do. We didn’t want to give up spot-up threes or Landale rolls.

You learn quickly that telling a coach what not to do gets you nowhere. There need to be solutions immediately to follow. But if those solutions are extreme enough, they can take you out of sample entirely, without any relevant film or data to analyze.

In Case You’re Not Sick of Ball Screens Yet…


Part of the goal of launching this newsletter is to gather feedback from you. We’ll hope to engage this community in the occasional poll, contest, or something else along those lines.

If you have any feedback on what you’d like to see from this, any general thoughts on the Twitter/YouTube/Podcast coverage...or anything really, please fill out THIS FORM.