The Strickland Archives: Highway to the Floater Zone

Digital archeologists this week discovered what was believed to be lost research from a contributor of The Strickland who goes by the name of Drew, aka “Doug.” What makes this article fascinating about this document is that it was originally supposed to be published back in May, roughly five months before a report was published about the New York Knicks focusing on “floaters.” Did the Knicks analytics and coaching department hack into The Strickland’s secure database to steal Drew’s research, or did much smarter people who work for the team coincidentally come to a similar conclusion? No one will ever know. Here is that document.


May 30, 2022

This is not the first time I have discussed nor will it be the last. Shooting zones are kind of my thing. May not be the best “thing” to have but at least I have some knowledge in one specific aspect of basketball analytics. Maybe? I’m not really sure at this point anymore, really. Nevertheless, let’s talk about shooting zones. 

Back in the Posting & Toasting days, I made a nifty little stat called The MoreyBall Coefficient. The purpose of the statistic as well as its companion metric, The MoreyBall Shot Coefficient, is to measure how efficient, effective, and diverse a player’s offensive shooting production is. It rewards three-pointers and shots at the rim and penalizes shots in the long midrange area. I strongly recommend taking a look at the article as it provides a significant amount of background information on the topic of linear regression, which is going to be an important factor later in this article. 

At the time, I took the five-year totals of shots made and attempted in the designated shot zones via NBA Stats from 2014-19 and regressed those variables to Five-Year Luck Adjusted Offensive RAPM of the same timeframe. What really caught me off guard about that article is that I was still contributing at Posting & Toasting at the beginning of the COVID-19 pandemic and that Allonzo Trier had the best MBC on the team. Remember Allonzo Trier and the early days of the pandemic where we all lost our minds watching Tiger King?

Two years later, I have fine-tuned the methodology and analyzed the most recent five-year sample. This time, however, I am not going to use a regression equation to calculate an estimated RAPM value based on shooting zones. This analysis is going to go in another direction. A direction that points us straight to the floater zone, baby!

Methodology

The most important change that I made to the methodology of the study was using shots made and shots missed as the variables for each zone compared to shots made/shots attempted (NBA Stats data) and percentage of shot taken/percentage of shots made (Basketball Reference data). MBC was paired with the “MoreyBall Shot Coefficient” (MBSC) as a way to control for where players primarily take their shots. For the three-point analysis, using Basketball Reference’s “percentage” stats for team data was a simple and effective way to have standardized data that measured volume and efficiency for shots in specific locations. Unlike my “MoreyBall” metric/analysis and the influence other shots on the court have on three-point percentage, I wanted to examine the consequences of taking a shot in different areas of the court at the player level. Another way to frame this is that I wanted to determine the risk and reward of taking specific shots in specific court zones. 

The data collected comes from PBP Stats (the data here is pulled from NBA Stats play-by-play data) and NBA Shot Charts, spanning the 2017-18 through the 2021-22 regular seasons. The sample is the top 500 players in total minutes over this five-year span. PBP Stats does not provide figures for every single player during this five-year span and that is fine as 500 players who actually received minutes is a very good representative sample to draw conclusions from. But before we get into the meat of this analysis, I want to provide everyone a visual of where the shots are located. PBP Stats breaks down shooting zones into the following sections:

  1. At the Rim

  2. Short Midrange

  3. Long Midrange

  4. Corner 3s

  5. Above the Break 3s

  6. Non-heave Above the Break 3s

  7. Heaves (3s)

The exact distances for each zone are not provided by PBP Stats; however, if you look at the Player Shooting Data on NBA Stats, the breakdown is somewhere between Shooting Zones and the 5ft Range if you convert the sections to less-than-5 feet, 5–14 feet, 15 feet–three-point line, and 24+ feet/3-pointers. To get a better feel for the exact locations, I took the Knicks shot data from this season and broke it down into different sections based on the values NBA Stats provides.

Let’s get back to the study. So, I performed a basic OLS regression using 5-year Offensive RAPM (ORAPM) as the dependant variable, and the following as the explanatory variables:

  1. At Rim Shots Made

  2. At Rim Shots Missed

  3. Short Midrange Shots Made

  4. Short Midrange Shots Missed

  5. Long Midrange Shots Made

  6. Long Midrange Shots Missed

  7. Corner-3 Shots Made

  8. Corner-3 Shots Missed

  9. Non-Heave Above the Break Shots Made

  10. Non-Heave Above the Break Shots Missed

  11. Usage Rate

Each variable’s measurement is per 100 possessions and standardized. The reason I selected ORAPM rather than Luck-Adjusted ORAPM is because the luck-adjusted version is based on players’ career shooting numbers and not the specific 5-year sample. We are measuring exact figures and their impact on actual results and not estimated/controlled results. With that said, at the end of the article, you can view the results of the OLS regression with Luck-Adjusted ORAPM as the dependent variable. Below is the results of the regression:

What we have here is a statistically significant regression where the explanatory variables (shooting zones) explain roughly 49 percent (0.4867 r-squared value) of the variation of the dependent variable (5-year ORAPM). We also have low standard error values across the board, which is always a good thing. Let’s translate the results:

  1. The two best shots in basketball are shots at the rim and free throws. For every unit increase in shots made at the rim, a player’s ORAPM increases 0.571 while a unit increase in shots missed at the rim decreases the player’s ORAPM by -0.111. Apply the same language and logic with the estimate values for free throws.

  2. Okay, so if you want to take the sums of the estimates for each zone (ex. 0.571 + -0.111) and say “that’s the value of taking the shot because we are measuring the end results of the action,” the “best shot” would be above the break threes. Another way to frame these results — my preferred interpretation — would be that the three-point shot has the greatest effect on the variation of ORAPM in that it has the largest estimate values of the explanatory variables (1.278 and -0.759). Because of these drastic values, three-point shots have the greatest risk between making and missing them. Don’t forget to look at the standard errors for those variables.

  3. If we use a similar interpretation as the point above, is there any reason to take a designed long midrange two-pointer?

  4. The difference between the estimate values of at rim shots made and short midrange shots made is 0.035.

  5. Despite having the worst t-statistic, the results for usage rate suggest that higher usage rates have a negative effect on ORAPM. 

Conclusions

These regression results only confirm the conventional wisdom of the current NBA. Getting to the rim and forcing teams into foul trouble are great ways to positively impact the game, three-point shots giveth and taketh, and why bother with 21-foot jumpers. 

But the main reason why this article even exists is because of the results for short midrange shots made and missed, specifically shots made. As we continue to watch the 2021–22 NBA Playoffs (I’m currently watching Game 2 of the Western Conference Finals between Dallas and Golden State and you’re reading all of this in an undefined time), we see the elite defense pick up the intensity and make those shots at the rim and behind the three-point line more difficult. Teams know one another so well. You gotta make difficult shots. And what difficult shot has a similar impact to those everso important shots at the rim? That’s right, those pesky short midrange shots. 

Having an effective floater, push shot, or pull-up jumper at the elbow is crucial to being an offensive threat. It is in no means a replacement for shots at the rim, rather the shot you need when teams are simply walling off the paint and there aren’t better options. Let’s use RJ Barrett and Immanuel Quickley as examples. It’s safe to say that both players can make three-point shots, right? I think so and I’m going to. That’s one level of scoring that both players share. RJ Barrett is very adept at getting to the rim, but that in-between game needs work in order for him to have three defined scoring levels. Whether that is a floater or a pull-up jumper (I prefer the later), Barrett needs a shot in that zone to make him a three-dimensional scorer versus a two-dimensional scorer because in the playoffs, teams can take away a dimension, leaving a player with only one dimension of scoring and in turn making him significantly easier to defend. Immanuel Quickley, on the other hand, has the in-between game with his floater (he also needs a pull-up midrange jumper but that’s for a different discussion), yet does not get to the rim as frequently as he should. Quickley did improve his at-rim frequency as the year progressed, but that needs to be sustained over the course of an entire season. 

So let’s wrap this up so you can review the luck-adjusted version of the regression to see how much three-point shooting varies. If there is any sort of takeaway I would like you to get from this article and analysis, it’s that a team’s primary offensive options truly need to be three-level scorers. They must be able to get to the rim, make threes, and be able to get to a spot in the midrange where they can make a shot on good efficiency. These shots are intertwined and effect one another in that if you take a Daryl Morey Houston Rockets interpretation of these results — shoot a bunch of three-pointers and get to the rim — you’re limiting your offensive diversity because a significant section of the midrange are also good shots. Obviously don’t take 22-foot two-pointers when you can simply can a corner three or move back two feet to take an above the arc three, but you can’t just not have a 14-foot jumper or an 8-foot floater. Those are actually good shots. More risky than at the rim, yes, but still a good shot. Take them.


Addendum: Luck-Adjusted Dependent Variable


Alternate article art

Previous
Previous

Knicks 121, Pistons 112: Just another team — in a good way

Next
Next

Knicks 120, Timberwolves 107: Let it rain