Nylon Calculus Week 14 in Review: All-star starters and recency bias

SACRAMENTO, CA - NOVEMBER 20: Nikola Jokic No. 15 of the Denver Nuggets looks on during the game against the Sacramento Kings on November 20, 2017 at Golden 1 Center in Sacramento, California. NOTE TO USER: User expressly acknowledges and agrees that, by downloading and or using this photograph, User is consenting to the terms and conditions of the Getty Images Agreement. Mandatory Copyright Notice: Copyright 2017 NBAE (Photo by Rocky Widner/NBAE via Getty Images)
SACRAMENTO, CA - NOVEMBER 20: Nikola Jokic No. 15 of the Denver Nuggets looks on during the game against the Sacramento Kings on November 20, 2017 at Golden 1 Center in Sacramento, California. NOTE TO USER: User expressly acknowledges and agrees that, by downloading and or using this photograph, User is consenting to the terms and conditions of the Getty Images Agreement. Mandatory Copyright Notice: Copyright 2017 NBAE (Photo by Rocky Widner/NBAE via Getty Images) /
facebooktwitterreddit

This is the time in the NBA season for All-Star arguments, and after that the MVP arguments will begin. So I want to preface this article with something that’s been gnawing at me for a long time, and it’s something I’d mentioned before: why are we so concerned with good players on bad teams, and why are we still so flummoxed by players with a long string of missed games? This is the era where games are available everywhere, and there are free videos online of every relevant player. We have an incredible wealth of stats available as well, from full metrics to play type data and SportVU outputs. Why can’t we figure out if there are All-Stars on poor teams?

Surely stars aren’t worth 40 to 50 wins on their own — that’s nonsensical — so there have to be a few worthy guys on mediocre teams. Let’s not limit ourselves. Likewise, people have this tendency to view missed games as a binary decision: either you play full-time or you miss enough not to quality for, say, the All-Star game. This is strange. It’s just about balancing value and playing time. Just look at metrics, like BPM’s VORP for how it deals with, say, Chris Paul versus Bradley Beal. Or my metric HBox, which was trained on actual MVP votes for its MVP Index, for how it handles Stephen Curry’s low minutes totals. I’m not saying take these metrics for gospel — just understand the trade-off. If you’re very, very good, you can provide more value, or wins, in 30 games than someone else can in 40. This should be obvious, and it’s the best way to deal with missed games. And with that, let’s look back at the week in basketball.

The secret tunnel Swiss diversion

A week ago, the competitive Clippers-Rockets game, the dreaded rematch between Chris Paul and his former team, ended with some questionable calls. Naturally, the losing team, the Rockets were upset at how things ended, and they responded with the totally reasonable secret tunnel infiltration plan with the Swiss front door diversion. I know we all witnessed this absurd little miracle, but I can’t “review” a week and not mention it. Basketball Twitter lost its collective mind over this incident, as each “Woj bomb” descended Twitter further into chaos and laughter.

But sadly some things are too good to be true. Chris Paul did not lead them through the secret tunnel, and he and James Harden were being peacekeepers, which is why only Trevor Ariza and Gerald Green were suspended. Tarick Black, not Clint Capela, was at the door, but only briefly when he heard some commotion. The incident was a showcase in how a group of people can create their own reality by filling in caps and misconstruing the language that was shared, a Rashomon effect for the court. People are notoriously bad witnesses, and once there’s another level of abstraction — commenting over the internet — it’s even worse. But it was fun while it happened.

All-Star starters

The starters for the all-star game were announced for the upcoming game, and the initial takeaway is that we didn’t screw anything up too badly. This isn’t B.J. Armstrong in 1994 — all those guys selected below would have been invited eventually anyway. There’s no Zaza Pachulia monkey wrench. The game itself is near meaningless at this point, but we still care about the accolades associated with it, unfortunately. So while we may brush aside arguments about DeMacus Cousins being voted onto the starting lineup, for instance, or Joel Embiid over Al Horford, it still ultimately matters. Those arguments are okay and necessary, and I hope the voting doesn’t regress.

George Hill: Trade target

A trade rumor sprang up a few days ago that helpfully reminded people that George Hill is trapped in Sacramento on a long-term deal. The Cavaliers apparently inquired about the point guard, most likely in response to Isaiah Thomas’ poor play and their crumbling defense. George Hill was one of the most underrated guards in the league for a while, with the impact of a borderline All-Star and the reputation as a mere stopgap starter. His size — he has a wingspan of 6-foot-9 — is a key factor here; he’s been one of the best defensive point guards over the last five years. On offense, he can act as a “3-and-D” type, but he can also be pressed to do more for his team, like in 2015 and 2017, respectively.

The question is, is he beyond repair? He hasn’t been successful with the Kings, he’s already 31-years-old, and he missed half of last season and even had to sit out a few playoff games because of nagging toe issues. One might posit his result this season are due to an inevitable decline, but he’s a large point guard whose skills should age well and the metrics aren’t picking up everything accurately. His individual stats are actually near his career averages, as you can see below, but metrics like BPM and RPM show disastrous numbers because of how much worse the Kings are, and that’s not entirely his fault. (Those two metrics have team adjustments, essentially, so often a player at the same skill level will be sunk down into the abyss with the rest of the team.) And, of course, his per game averages are just reflecting his low minutes totals. He’s actually shooting really well.

Table: George Hill 2018 season in comparison

STL%BLK/100ORB%DRB%AST%3PA/FGAFTA/FGARim FGA/FGATS%USG%
20181.80.62.210.615.30.3920.2870.17158.617.1
Career1.70.52.210.218.60.360.2990.22956.718.4

For Cleveland and other organizations, the good news is that the last season of Hill’s contract, in 2020, is guaranteed for only one million, so the gamble isn’t too risky. I don’t think he’s playing as poorly as most think, and I’d guess that motivation is a factor too because the team is going nowhere. He was sold a delusional line, as the team said that by bringing in veterans Zach Randolph and Vince Carter, along with Hill, they’d be able to compete for a title — any halfway aware NBA fan knows that’s bogus. For a player who’d fit so well on a contender — most elite teams already have ball-handlers and scorers, and the skills that stack most easily are shooting and defense — I think he’s worth serious consideration. There’s a chance we’ve already missed his prime, but if you get the best of George Hill, you get it at a cheap bargaining price and a player who could plug in seamlessly.

Doubting Isaiah Thomas in Cleveland

As the trade deadline nears, we’re seeing Cleveland spring up frequently in trade rumors, and many of them have to do with importing a new point guard to replace Isaiah Thomas, from George Hill to Kemba Walker, recently made available, to Lou Williams. But Thomas has only been active since the beginning of the new year after a long absence — since the 2017 series against the Cavaliers with Boston, funnily enough — and he’s been recovering from an impingement with a tear in lieu of surgery for a few months. Is it fair to evaluate him already and dump him for another point guard, or should the Cavaliers move on already?

Thomas’ first few games have been rough, and it’s almost all because of shooting efficiency. He’s been getting to the rim less often, he’s been converting few of his shots once there, he’s getting to the foul line less too, and he’s hitting fewer of his jump shots too. His shooting numbers have been abysmal, and he’s looked the part too sometimes on offense, looking slower and hesitant. Thankfully, most of what’s changed about his game is contained to shooting, although he has been getting fewer assists, which makes sense with LeBron James’s role on the team. I’d be concerned that his free throw rate, steal rate, and his rim numbers are all down, key athletic indicators, but again, he could just be recovering.

You can see the highlights from a recent game he had against the Magic. That was one of his best, and you can see how he can help the Cavaliers: his ability to hit outside shots, with and without handling the ball first, is hugely important for LeBron James, and his creative passing ability is obviously an asset too. If he can resurrect more of his slashing and finishing powers, a historic outlier for someone his size, he’ll be the offensive firebug the team needs. Defensively, alas, he can’t fix their major, major problems, but he’s not the cause either, as both Derrick Rose and Dwyane Wade have been liabilities. I’d say put a pin in the Isaiah Thomas case and wait patiently, but Cleveland can’t do that and they may act quickly to avoid catastrophe.

Who’s going to be the requisite All-Star snub?

First, let’s get this out of the way: expanding the All-Star roster will not resolve the snub issue. The snub is just someone just on the outside of being picked — talent is relative — and those guys will always exist. In fact, due to the talent distribution of the league resembling a sort of bell curve, expanding the roster would actually increase the number of snubs because you’d be digging into the meat of the fat part of the curve. Also, it’s tough enough finding playing time for everyone.

As it stands, I think the Eastern Conference is fairly easy to predict, aided by the numerous articles out there already on the subject: Kyle Lowry (Toronto is doing well, don’t forget); Victor Oladipo, since everyone loves his story anyway; Al Horford; and Kristaps Porzingis seem obvious. Then there’s Andre Drummond, John Wall, and Ben Simmons — I see them listed frequently. Kemba Walker could be the odd man out, and if he’s selected I’d guess Ben Simmons was the “snub” since some coaches don’t think a rookie has yet to earn the distinction of playing in an exhibition game. But he has definitely played worse the last few weeks, as his shooting percentages have tumbled. The “snub” factor in the East is pretty weak, save for Kemba Walker, Kevin Love (who’s been dogged by poor defense anyway), or fans of Bradley Beal.

The Western Conference is tough. Let’s throw out the obvious guys: Jimmy Butler has been showing MVP-level impact; Draymond Green, the defensive lynch-pin of the best team in the league; LaMarcus Aldridge because people pick players by winning teams and need a Spur to represent; and … uh probably Russell Westbrook, who was MVP last year and still does a lot on the court. But it’s murky after that. Guys either have huge question marks on defense or their teams aren’t doing well.

If I had to guess, I’d say the reserves end up being Russell Westbrook, Jimmy Butler, LaMarcus Aldridge, Draymond Green, Karl-Anthony Towns, Paul George, and Damian Lillard. Klay Thompson is the most likely candidate to swap out for any of the last two guys. In that case, you know who I’m going to say the snub is: Nikola Jokic. He kills it with almost every metric, including my own HBox, which has him at 13th. Due to injuries and the decline of the Grizzlies, the snub factor isn’t too high in the west this year, but it would still sting seeing the likes of Klay Thompson, Nikola Jokic, and Chris Paul staying home.

So I guess people are serious about the Kevin Durant DPOTY talks

I wanted to ignore this, but the discussion hasn’t died and I think some voters are serious. With few standout performers on defense, Kevin Durant is one of the leading candidates for Defensive Player of the Year. I suppose this is because the top defensive teams have no one obvious or with flashy stats, as Boston has Al “understated stats” Horford, San Antonio hasn’t even had Kawhi Leonard, the Thunder may split votes, and the 76ers have Joel Embiid and his low minutes total. But Durant, despite averaging over two blocks a game, isn’t the answer. Blocks are only a small portion of the game, and the Warriors actually defend a lot better without him on the court — it’s tough for a true Defensive Player of the Year to accomplish that while starting. You can see some of the splits here too; they even defend well at the rim without him. He’s definitely not below average, but those aren’t great signs and I wouldn’t even call him top three on his own team. Let’s cool down on the Durant-Defensive Player of the Year talk please.

The LaMarcus Aldridge default

At this point, we’re all just resigned to the fact that Aldridge will be an All-sSar because, the thinking goes, the Spurs have played well, they were without an NBA superstar, and they “deserve’ at least one selection. I’ve pulled this old quote before, but this is the faulty parliamentary representation thinking of the All-Star game. We don’t owe it to the game to represent by how well the teams are doing — this isn’t like selecting the appropriate number of congressmen for California based on their population. The game is supposed to have the best players in the NBA. That’s all. I suppose one could argue we should represent either many or the best teams, but that’s not stated anywhere.

Every time Aldridge’s case is mentioned, someone will bring up how the Spurs are (or were) third in the West. Well, sure, that’s true but it’s misleading. Their point differential suggests a weaker team, and the conference is dominated by the two teams at the top — a ranking is definitely not the best measure of a team. We can do better. Based on how well Kawhi Leonard played last season and how often he played, this is actually about as good as you’d expect them to play without their star. They’re just a really good deep team. We shouldn’t be surprised by that fact with Gregg Popovich-led teams anymore.

Aldridge himself is having his best season in San Antonio. It’s not just that he’s able to shoot more often with Kawhi sidelined; he’s been shooting more efficiently too and his peripheral stats are slightly better as well. But the net effect is not one of an obvious “lock.” He should absolutely need arguments to get included over the many quality players in the west. And does anyone really think another quality big man — for some reason Nikola Jokic comes to mind — wouldn’t flourish on the Spurs too? Even the hobbled and nearing 40-years-old Pau Gasol has been playing surprisingly well. Aldridge shouldn’t be an automatic pick; we should dwell on this one.

The recency effect and team ratings

Even with SportVU data flooding the league and all the other statistics we have access to, teams are still being measured by the tried-and-true measures of win percentage and point differential. There are an incredible number of ways to adjust ratings, but only a few are usually tried. One of those is by recency. The thinking is, how a team plays recently is more important than how it played, say, four months ago. You can understand that logic even if one adjusts for player minutes and trades. One major site, FiveThirtyEight, includes this in their team ratings and for their predictions, but is it useful? Does it have predictive power?

I wanted to attack this with a simple method, and it should be simple. It’s just about if the recency of the game is an additional and significant factor. To test this, I’m looking at every game in my database from 1984 to 2017 in the regular season. My base measure is net rating — something to adjust for pace, but nothing too fancy. For testing, the data is split into two portions: the first two-thirds of the season will be used to predict the net rating of the latter third. I quantified recency by days. The first game of the season was day 1, and the last day was usually around 170-something.

The respective team’s net rating for a single game is multiplied by the day. Thus, the 100th game was 100 times “more recent” than the first, and it’d be up to the regression models to figure out what the best balance is between that and the flat net rating. All told, even with that recency net rating, the results aren’t extreme because there are enough games to balance things out at both ends, and the recency ratings, even with those unadjusted weights, are still close to the “normal” net ratings.

Over several years of data with nearly 40,000 games, the results are summarized in the table below. Here’s the quick takeaway: the recency factor was on the borderline of being statistically significant. With slightly different methods you can push that into statistical significance (i.e. a p-value over 5 percent) or further into the other direction. It’s a slight gain in the model’s effectiveness, which you can also see from the change in the R-squared value. It’s not the secret X-factor that people are seeking for team ratings.

Table: model results for recency team ratings

ModelNet rating coef.Recency coef.P-valueR-squared
Recency0.6650.2222.9%0.6215
Base0.886N/AN/A0.6200

I could try other methods and further refine the numbers. But this is what’s really bothering me about the results: even with midseason trades and major injuries, recency is still not a dominating factor. That’s not a good sign. It seems prudent to invest more time in adjusting for who’s on the court than plain recency, which doesn’t accurately summarize all the micro-effects of injuries and signings and deals. If your metric doesn’t like Houston anymore, for example, but it doesn’t know that James Harden is coming back for the next game, can you really call that an intelligent team measure?

Next: Nylon Calculus -- Donovan Mitchell, Brandon Ingram, and lessons in the development of NBA players

I tried a few different variations in good faith too. I looked at other stretches of seasons. I even let the first chunk of the season predict the next two or three games instead of the rest of the season — same results even though the dependent variable was drastically different. Another problem is that once you have a few weeks of the NBA, teams are remarkably stable and they don’t have huge swings in performance as often as you’d think. I’m open to the idea that this effect is more pronounced, but I don’t see it in the data and if anyone does, please direct me to the nearest study with some reproducible results or at least a clear methodology.

Recent results exert a great influence over people, and it can color our perceptions. The Bulls are seen in a different light now because of how they’ve played lately, as many of us have all but discarded the first portion of the season for them. The recency bias would make us think this is how they’ll play from now on, but aside from adjusting for Nikola Mirotic’s presence, the wisest procedure is to look at their entire season and not be too focused on what’s happened lately. And if the Bulls continue to outperform expectations it’ll be an exception and not the rule — teams rarely radically and permanently change in-season. The recency effect has only a small significance in the NBA, despite what our brains tell us.