After posting my adjusted defensive impact metrics yesterday, I said this on Twitter:
I was referencing the bars at the bottom of each player’s chart. Before I continue, let me explain how to read these. Here is Al Horford:
This means that when Horford is on the court as a defender, the probability that any given shot will be a close shot (which I define as <8 feet from the basket) is increased by 2%. The probability that any given shot will be a midrange shot is increased by 4%, and the probability that any given shot will be a 3 is decreased by 6%. These should sum to 0, although occasionally they may not due to rounding.
Now take another look at the adjusted defensive impact charts and compare, for example, Roy Hibbert to Paul George or, as @gswhoops mentions, Andrew Bogut to Andre Iguodala. The results show pretty consistently that big men tend to alter where an offense takes its shots. Wing defenders don’t. @WhrOffnsHppns asked if this might be a consequence of the method and as I sat at my work computer, with no access to my code, I realized that I couldn’t even really remember what the method was. I implemented this weeks ago and then just kinda forgot about it. So below the cut I outline the method in detail and look at some descriptives to buttress the big man findings. The short version though is that I do not think these results are just an artifact of the method and I want to double down on what I think is a true finding: “Interior defenders affect opposing shot location. Wing defenders don’t.”
I have finally completed work on my visual NBA stat: Adjusted Defensive Impact by Court Location. I first explained how this stat works here, but in a nutshell this is a way to visualize how a player defends shots in the NBA adjusting for the other defenders on the court with him, the expected probability of the shot being made, and the other (non-shooting) offensive players on the court. This new model also adds the possibility of a team-wide effect that you might attribute to coaching (this is not visualized in any way just yet). I had many requests to also include something about how players affect the location of a shot. You can now see this at the bottom of each player’s chart. This is a simple regression that controls for defensive players only and shows you how a given player affects the volume of shots that are close (<8 feet), midrange, or 3 pointers. I have lots more to say below the jump but here’s the widget, have fun poking around! Warm colors mean that opponents are more likely to hit a shot when the player is defending, cool colors mean opponents are less likely to hit their shots.
2013-2014 Adjusted Defensive Impact by Court Location
Select a team from the drop down below to see that team’s defense visualized for the 2013-2014 season. There’s nothing fancy going on with these visualizations–just raw, unadjusted comparison to the average. Blue squares indicate that the team defended that location better than league average, while red squares indicate that the team defended that location worse than league average. The numbers give FG% from that location for the opposing team. I talked about these visualizations previously here and here, and explained how I make them here.
I will have adjusted defensive impact for all players in the NBA soon-ish. It’s actually ready to go but I got gun-shy and decided I needed to test alternate model specifications. It might be a couple weeks before I get them all up, but it will be a more reliable product.
2013-2014 NBA team defense compared to average
This post was prompted by this twitter exchange:
I want to clarify this issue for non-stats people. Adjusted plus-minus and similar analytics, including my own adjusted defensive impact analytic that I debuted a couple days ago, use multivariate regression to ‘adjust’ the raw numbers. The basic idea is that a lot of players who are not very good get to share the floor with great players sometimes. Udonis Haslem gets a fair bit of floor time with LeBron James. When that happens, Haslem’s raw plus-minus is going to look great! But that’s mostly because James was also in the game with him. So the regression controls for all the players on the court, and tries to extract the contribution made by each.
Thanks everyone for the really great response to my previous blog on twitter. Some people asked good questions and I want to clarify a few things and talk about some improvements that could (maybe should) be made to this. This post will be kinda methods-y but may include some non-technical tidbits.
I’m finally ready to share a big project I’ve been working on for months. Try selecting a player from the drop down menu below. The visualization that appears shows the defensive impact a player has adjusting for all other defensive players and the offensive player who is actually shooting. Blue squares indicate that when the selected player is on the court, the probability that an offensive player will make his shot declines (so blue is good defense). Red squares indicate that when the selected player is on the court, the probability of any offensive player making his shot rises. Go ahead and take a second to play with the drop down menu. I’m not ready to rollout every player yet, but the Pacers, Bulls, and Grizzlies are all on there. After the jump, I explain the full methodology, and answer some questions that you’ll probably have after looking at the charts.
Adjusted Defensive Impact by Court Location
Everyone knows by now that mid-range shots are inefficient. The Rockets in particular have taken this idea to heart. Just 15% of their shots come from mid-range (I’m defining mid-range here as more than 8 feet from the basket but inside the 3pt line). For comparison, the Blazers shoot nearly 36% of their shots from mid-range. Great defenses like Indiana’s try to force mid-range shots.
But here comes the backlash. Zach Lowe, Seth Partnow, and Benjamin Morris all suggest that the mid-range game has a place. As I understand it, These are their primary arguments:
- Sometimes, a 2pt shot is all you have, and when that’s the case, you want to be good at making them. All things equal, an open 2pt shot is better than a 3pt shot with Tony Allen all over you.
- Defenses will adjust.
- Not everyone has the personnel necessary to take only 3s and layups.
I don’t have enough data to crack this question and I suspect no one does–there aren’t enough teams employing a Rockets’ style offense to make sweeping judgements about them. This post is more of a first cut at examining the data and making some tentative observations.
Offensive Rating vs. Midrange FG%
ESPN managed to gin up a little 4-point line talk recently with rumors that the league had talked about adding such a line. The reports turned out to be way overblown but it got me thinking about the issue. What this issue calls for, clearly, is a completely absurd approach via analytics!
I set out to completely redraw the NBA court with only one rule: PPS (points per shot) should be the same from anywhere on the court up to say, 30 feet away from the hoop. I used all shots taken in the current season to determine what NBA average FG% is 1 foot away from the hoop, 2 feet away from the hoop, 3 feet away and so on. Then, I picked point values for these distances such that PPS was equalized at all distances. It’s no good to start with 1 or 2 point shots, because doubling the point value of a shot means you need a huge decrease in FG% for PPS to stay the same.
After fiddling with the number for a bit, I decided that the lowest point value shot would have to be 3 points. Which means that on the high end we’re going to need, well, a 7-point line. Here’s the court:
What surprised me the most in looking at how FG% degrades as you move away from the basket was that once you’re 3 feet away, FG% doesn’t change for another 22-ish feet. Another 3 feet after that it takes another hit, although this is probably because shots taken from 27 feet out are almost always rushed or pressured. If there was a reason to take shots that far out, players would shoot a better % from there, and we would have to move my 7-point line out. Anyways, expected PPS from every part of this court is about 2!
Lazy Sunday, time to post some more NBA defenses. The first game today is OKC v. LAC, two very good defenses and an opportunity to see how good defenses actually operate very differently. First up is the Clippers, who are not a particularly good defensive team in the paint but are currently leading the league in 3pt defense, where they absolutely shut down the opposition. As a refresher, the numbers are PPS from the square they are on. The Clippers defend the arc so well that PPS from behind the arc and from midrange is almost a wash.
Opposing FG% compared to league average, Los Angeles Clippers
This is just a catch-all post about methods that I will reference in the future when I post a graph or a regression or whatever. My plan is to update this every time I add something new that I think requires further explanation. So without further ado…
Adjusted defensive impact by court location
From here on out, the way I do the shot-chart visualizations should be fairly stable. There are really only a couple of things that need explanation here. Data is usually current as of the date of the blog post but does not update automatically, so backdate appropriately. All shots taken against a team or by a player or whatever it is are grouped into 1ft x 1ft squares that cover the court. 2 and 3 point shots are not mixed in this process. Basically if a square’s center is inside the arc, it should contain only 2pt shots, and if the square’s center is outside the arc it should contain only 3pt shots.