Nylon Calculus Week 16 in Review: Trades and 3-point defense revisited

Jan 21, 2017; Denver, CO, USA; Denver Nuggets center Jusuf Nurkic (23) in the fourth quarter against the Los Angeles Clippers at the Pepsi Center. The Nuggets won 123-98. Mandatory Credit: Isaiah J. Downing-USA TODAY Sports
Jan 21, 2017; Denver, CO, USA; Denver Nuggets center Jusuf Nurkic (23) in the fourth quarter against the Los Angeles Clippers at the Pepsi Center. The Nuggets won 123-98. Mandatory Credit: Isaiah J. Downing-USA TODAY Sports /
facebooktwitterreddit

Nylon Calculus is about delivering basketball numbers to the people, and at no point has that seemed like an important part of the Democratic process until now. This may seem like a distraction or useless stat-wonkery, but the entire philosophy of embracing and applying facts and numbers is an essential component of a functioning modern society.

If people are swayed by falsehoods and don’t believe in the latest evidence, then we are lost. And don’t think this activism will vanish in the NBA either; it’s part of the league. How we perceive events and the numbers that describe them will always be an issue unless we’re willing to confront the hard data and take an objective look. It’s something we should practice in all areas of our lives, even sports. And with that, let’s take a look at the last week in basketball.

Denver-Portland center swap

Ahead of the trade deadline, two teams battling for the last seed in the Western Conference race decided to trade centers: Mason Plumlee for Jusuf Nurkic. Portland will also send a second-round draft pick for 2018 to Denver, while the Blazers will receive Memphis’ protected 2017 first-round pick in return. Since the pick’s protections are only from No. 1-5, that means Portland will probably get a decent draft pick in the mid-teens. You can see all sorts of trade breakdowns on the internet, but I think most fail to logically corral the goals of the trade: what will Denver do with Plumlee, and has Portland changed their long-term outlook with this new center?

Read More: In Nurkic-Plumlee swap, Trail Blazers and Nuggets both emerge as winners

Mason Plumlee has two functions in Denver: he can be the power forward next to Nikola Jokic, their clear star of the future who was surprisingly close to Darko Milicic, or he can be just a backup center. I, however, doubt that a Mason-Jokic pairing would work, mostly on defense — which guy do you want guarding stretch 4s? Plumlee, for now, is an upgrade over Nurkic as a backup center, but he solves none of their issues defensively and will be a restricted free agent this summer. Was this upgrade necessary and worth the price of a Memphis first round pick? This is a league where the center position is deep; it’s a buyer’s market

For Portland, they’re probably worse in the short-term, but look for Nurkic to fill some of their rebounding deficiencies. Denver is one of the best rebounding teams, and Portland one of the poorest; with how rebounds encounter diminishing returns, Nurkic should grab more boards with the Blazers. But the team will miss Mason’s passing, which has been an underrated subplot — few centers assist this much, at least since the 1980’s. Jusuf is, however, younger, cheaper, and has one more year on his rookie deal — coupled with the flipping of a second round pick to a first round pick, and that’s a worthy gamble. The trade is a bit fairer when you consider the cash sent to Denver, but I still like Portland’s move.

There’s still an unresolved question as to who has the higher value in a vacuum, and much of that depends on how you perceived Nurkic’s rookie season, where he looked outstanding for his age. Oddly enough, the Bosnian Beast had a higher overall Real Plus-Minus his rookie season than Plumlee has currently and just a hair under Plumlee’s best season, 2015. Remember that fit can affect individual numbers, even advanced ones: it’s possible Nurkic’s value has been depressed this season and could look better as the featured center in Portland lineups with plenty of shooting.

Table: Mason Plumlee

SeasonAST100ORB%DRB%Rim DFG%ORPMDRPM
20152.211.321.655.6-0.32-0.29
20165.410.422.252.50.322.11
201778.723.248.01.280.83

Table: Jusuf Nurkic

SeasonAST100ORB%DRB%Rim DFG%ORPMDRPM
20152.311.826.148.4-1.694.02
20163.712.922.758.6-1.291.08
20173.512.323.756.2-3.311.96

The Louisville Muskrats?

There’s been some expansion talk flittering about the league with proposed cities like Seattle, Mexico City, Las Vegas, and, strangely, Louisville — no offense to Kentucky, but I don’t see a compelling reason to start a franchise there and the metro area is ranked 43rd in total population (US only.) Only New Orleans and Salt Lake City rank lower, and Oklahoma City and Memphis are just a hair ahead. But Seattle’s ranked 15th with about three times the population; the only non-NBA city with a larger metro area is Riverside, CA. Even Las Vegas has more people, and Vegas has a higher growth rate too. Nonetheless, besides bringing the NBA back to a city that unfairly lost a team, Mexico City is the most intriguing option because the NBA would have its feet planted in another country and could draw the interest of an entire large nation, not to mention all the social ramifications of the league working with a country disparaged by the US’s current president.

I’m most excited by the geographical implications — if we put two new teams in the west, do we move Memphis or Minnesota to the east? — but the first reaction people usually have to expansion discussion is about league quality: dilution. I think that’s bunk, however. Many viewers prefer college basketball, especially the March Madness tournament, and the league quality there is objectively much, much lower. We will always have below average players by definition, and if you cut out a bunch of the worst players in the league the NBA probably won’t see a surge in shooting percentages and points — defense and offense are competing factors. Plus, I sincerely doubt even experts would see a drop in league quality with the addition of a team or even two, besides seeing new names; it’s a small drop in the pond. If you oppose expansion, do so for more logical reasons like there’d be too many teams to keep track of or, like the Roman empire, you don’t want the NBA over-extending its grasp. League quality shouldn’t be an issue.

The future: Wouldn’t that be nice?

The NBA is continuing with its technological advances. Seen through the eyes of a fan from the 1980’s, some articles now read like science-fiction. Teams are now using virtual reality to train players, and it’s not always something directly related to basketball, like watching your own form during made free throws in immersive VR. Some players walked a virtual plank, an exercise where people find it surprisingly difficult to jump off into the abyss even though they know it’s not real. That seems far removed from the NBA court, but it’s all about brain training and it’s another tool teams are using in an increasingly advanced and crowded competitive field.

Block me, Bebe, one more time

I’ve long been a believer in Lucas “Bebe” Nogueira, so it’s validating to see him succeed in Toronto and turn heads with some spiffy advanced stats. As I’ve always argued, he’s a long and quick center who put up decent stats against tough international competition at a young age — those are great signs. After being initially drafted by the Hawks, he spent some time overseas and then was traded to the Raptors in a salary dump for Atlanta (John Salmons’ contract) along with Lou Williams. After playing scant minutes the past couple years, he’s finally become a part of a rotation. And when people see that he’s ranked in the top 25 on ESPN’s RPM, I imagine there are some confused faces out there. But why does he look so great in advanced stats this season? Is it legitimate at all?

Besides the afro, the first thing people probably notice about Bebe is that he’s quite skinny — he’s a mop with arms or legs and has the physique of Sideshow Bob. It’s no surprise to see him struggle on the boards then; he currently ranks 41st out of 47 centers, via Basketball-Reference, in defensive rebound percentage. But he’s a high-steal, high-block player — there have only been 33 seasons with at least 4 blocks and 2 steals per 100 possessions in comparable minutes, and 13 of those were from Olajuwon and David Robinson. He’s been an extremely efficient scorer this year as well, and he’s a frequent target for lobs. Most traditional views of big men focus on points and rebounds, but he lights up a few of the other stats well.

Although some metrics peg him as a young star, shockingly, he does have a few weaknesses, and even as a fan I understand the concerns. He gets bullied by bigger defenders pretty easily. For example, he was regularly pushed out of the way versus the Pistons — here’s a play where Aron Baynes just walks right through him on a missed free throw and tips the rebound in. But his quickness and length are definite assets, where he can guard the rim and the perimeter, as in that same game he blocked two 3-point shots and one layup. His quick hands are a nuisance too: while on the move he pokes the ball away from a guard and it leads to a Kyle Lowry 3-pointer.

On the offensive end, he has a tiny usage rate but he’s an active screener and a lob threat. The strength concerns aren’t as important as they appear. The league is more about perimeter action and movement than slow, low-post play. The Serge Ibaka trade will balance the lineups for Toronto, and Bebe will play out of position less often at power forward — but if they want to pick up some games in the standings, they shouldn’t cut his minutes any further.

Dunk of the Week: Big Willie style

Willie Cauley-Stein had a sensational dunk a few days ago, and not only was it the dunk of the week it could be the dunk of the season. You can see a few of the candidates here and here. Outside of Larry Nance Jr., I don’t think anyone else’s dunk is as electric. Willie was drafted based on his size and athleticism, and you can see why in the clip below. Unfortunately, he’s a microcosm for the failures of the Kings. I’m not one who ever suggests drafting for fit, but Sacramento has drafted a curious amount of big men and added some free agents too. But they’ve yet to add another star or key supporting player for Cousins, as Cauley-Stein is yet another player who has been disappointing. Curiously, his BPM has dropped by 2.3 points, which is a huge change, yet his individual stats have hardly changed, except for his usage rate and that’s actually increased (This is a good exercise in how BPM is calculated. If you can figure out why this happened, I think you understand the metric pretty well — I think it’s about how his teammates accrued stats). Enjoy the dunk because as we’ve seen from the past, it’s smart to be pessimistic about a Sacramento player.

Morey Index

I’m way too optimistic in believing my ideas and invented stats will become widespread, so it’s disappointing to see the Ringer rank point guards by various components with a measure I’ve discussed frequently. The writer ranked point guards by “Moreyball” and was surprised to see James Harden coming up fourth. But it’s because he didn’t use my “Morey Index,” which is just (3PTA +FTA)/FGA. Harden’s is over 1 and he’d rank over every one of those guards. He did, however, use a stat Seth Partnow discussed a while ago: Moreyball (yes, the same title too.)  But I’d argue that since free throw rate is more indicative of the Morey philosophy, it’s more pertinent to use my stat, the Morey Index. Ignoring the free throw, in fact, skews the numbers and makes James Harden look like more of a mid-range shooter because he draws so many fouls at the rim. We can’t ignore those opportunities, and we should all agree that what makes Harden unique is his ability to create free throws.

Loveless

Kevin Love will be out for several weeks, as it looks like my dire warnings about the Raptors or Celtics passing them in the standings were prescient. I’m concerned that Carmelo Anthony and the Knicks will get an undeserved All-Star selection, but I assume more people are concerned with the team. This is not the time to panic though for the Cavaliers, who just need an okay seed and good health in the playoffs — their final record is immaterial. But it’s disappointing this happened to Kevin Love in his best Cavalier season, and it happened not too long after the incredible pass below. Love’s outlet passes remain one of our nation’s true art forms, and I hope we get a few more of them before LeBron James ages.

Cleveland will be tempted to either make trades to supplement the roster while Love is injured or just simply ride LeBron with heavy minutes. They don’t want to lose homecourt advantage in the East, but I think that’s a distraction. I am in no way connected to the organization, and I doubt they’d heed my advice anyway, but what they should care about is how well they can perform in the playoffs, not the regular season. LeBron should receive plenty of rest, and if there’s a trade that’ll help them in the playoffs too then go for it — otherwise it’s a waste of resources. Homecourt advantage isn’t valuable enough to risk their star’s health — it’s a change in venue of one game out of seven. Play the long game.

Team Defensive Rating: Adjusted by Opponent 3PT Luck

A couple weeks ago, I found that opponent 3-point percentage was essentially random and I subsequently created a model to better measure 3-point defense. To summarize, opponent 3-point percentage was found to be useless, and instead league adjusted averages, turnovers created, 3-point attempts allowed, and defensive free throw rate could be used together as a composite for a more predictive and stable proxy for “defensive” 3-point percentage. While that’s all well and good, a theory that’s not applied is a theory wasted, so let’s apply the model to real NBA teams and see what the results bring.

The first step is to adjust the model for possession type effects. If you state that in an idealized world an opponent “should have” missed more shots, then you have to account for both offensive rebounds and defensive rebounds (they generally lead to be a more efficient possession than a made shot.) Via my research when reinventing PER, I found that possessions after made shots were 0.96 times as efficient as the average possession. And, remember, this is all based on the change in the amount of made or missed shots as suggested by the model. The effects here are actually small. Let’s say a team allows 20 3-pointers a game, and the change in percentage causes a loss of one made 3-pointer — that’s actually one of the larger changes you’d see. That’s three points directly, but it’s only one possession changed, and most misses are rebounded by the defense anyway.

The next step is a validation. I’ve already tested the model out of sample and saw significant, useful results, but there’s nothing wrong with testing more, especially when you incorporate a tweak. Thus, I used a team’s pace-adjusted SRS and the changes from the model, and compared them to the team’s pace-adjusted SRS for the next season. This side-steps the sampling issues from how the model was built and can show if the model’s 3-pointer adjustment is a repeatable skill. Even by changing the filters for teams with too many new players or the year range, the model adjustment was still statistically significant. I know that’s not the most robust test methodology, but I’m pretty confident this is an improvement over the alternative.

As for the actual implementation, the effects are surprisingly sizable. The Cavaliers, strangely, get the biggest boost; that’s the season after LeBron James left. It’s also one of the greatest single season collapses, so perhaps this adjustment makes sense, ironing out an extreme. There are some 2017 teams here, and they’re all massive disappointments, particularly Portland and Dallas. For the most part, the teams with the biggest improvements are poorer teams.

Table: biggest improvements

TeamSeason3PT% (model)DRtgAdjDRtgSRS*AdjSRS*
CLE201136.0111.8108.9-9.5-6.6
DEN201235.1106.2104.13.45.5
PHO201736.2112.5110.4-5.4-3.3
PHI201035.6110.3108.3-4.3-2.2
DAL201736.0108.6106.6-1.50.5
NYK201535.2110108.1-10.4-8.5
POR201735.7112110.2-1.60.1
CLE200235.4108.2106.5-3.9-2.1
MIL200035.5107.9106.2-0.11.7
SAC200937.1114.7113.0-9.1-7.4

*Pace adjusted (source: b-ref)

On the flip-side, you can see the teams with the biggest declines below. The headliner is this season’s Warriors team, and if you’re skeptical there let me mention that this data is through Sunday and on Monday night the Nuggets destroyed their 3-point defense with a 24-for-40 shellacking. However, there are a number of good defensive squads on the table, and that’s where I get nervous. The model itself still gives the best defensive teams the lowest predicted 3-point percentages, because that’s how it’s built, but the range is so low it regresses overall defensive ratings by significant margins.

Table: biggest declines

TeamSeason3PT% (model)DRtgAdjDRtgSRS*AdjSRS*Diff
GSW201735.5103.9106.712.910.1-2.8
SAC199535.9106.5108.9-0.8-3.2-2.4
BOS200835.698.9101.110.28.0-2.3
NOH201235.2105.1107.3-3.5-5.7-2.2
HOU201535.1103.4105.34.02.1-1.9
DET200434.395.497.25.73.9-1.8
BOS201234.398.2100.02.50.7-1.8
LAC201435.8104.8106.47.66.0-1.6
LAL200235.1101.7103.27.86.2-1.5
CLE200936.1102.4103.99.88.3-1.5

*Pace adjusted (source: b-ref)

There were some great teams that were helped by this method, like the 2005 Spurs or the 1997 Sonics and the 1990 Suns, but they were considerably smaller in quantity. It’s at this point where I became concerned. The best illustration I have is the graph below. It’s the change via the model 3-point percentage versus the team’s original defensive rating. It’s clearly and highly correlated with defense — elite defenses are often extremely penalized. An adjustment of up to two points is no joke; that can translate into five or six wins.

3pt-defensive-luck
3pt-defensive-luck /

I’m unsure of what the issue with the model is, frankly. Consider these facts: on out-of-sample data, it was better at predicting second-half opponent 3-point percent from stats taken from the first half of the season than just the team’s own opponent 3-point percent. It was then better at predicting defensive rating in the same manner. I looked at how well a team’s rating tracked from one season to the next, and, again, that model’s adjustment factor improved the prediction. However, some of the results are strange– Detroit has a better adjusted defensive rating than Golden State this season, for example — and when I booted up a binomial title predictor, the adjustment made the prediction significantly worse. What exactly is going on?

I understand that some results should appear strange. If they don’t, then the numbers aren’t providing any new, revelatory findings. Of course most of the “great” modern defenses will have the biggest adjustments — they’re extreme values, and to have the best team defensive ratings you do need some luck too. I’m just unsure of the balance I have here, and it’s going to take more work than I thought.

These exercises at the end of the week in review articles are supposed to be fast, experimental pieces. Maybe I create something useful — or maybe I fail. The point is to try. But this topic deserves another week (at least.) I already have a few theories and fixes, but I’m running out of space here.

Next: Can Isaiah Thomas be the centerpiece of a contender?

Until then, we can ask ourselves, what exactly is propelling Golden State’s excellent 3-point defensive stats? Can it be sustained? What we saw versus Denver might be a blip or it might be telling information.