Friday, January 27, 2017

Introducing Net Differential Part I: A New Tool for NBA Player Evaluation

By Buddy Grizzard

If you’ve followed my writing, you know I’m one of the most ardent proponents of plus-minus among NBA observers. There’s a very simple reason for this. In the game of basketball, if your team scores more points than the other team, you win. I’m constantly challenged on observations based on plus-minus, but for me the essence is this: Did a player contribute to winning basketball? What follows is an attempt to answer that question in a new way.

When I first joined ESPN TrueHoop Network as a contributor in 2012, stats such as net rating (plus-minus regularized to a per 100 possessions basis instead of the traditional box score, per game basis) were not in common use. On former TrueHoop Hawks blog Hoopinion -- and here on my personal blog -- I compiled plus-minus numbers by hand. When I later became aware of net rating on I quickly realized that it gave my observations more force.

For example, in this piece about Larry Drew’s absurd decision to sit Kirk Hinrich in the fourth quarter of Game 3 of the Hawks’ 2012 first round series against the Celtics, I noted that Jannero Pargo was -20 and Willie Green -18 through the series’ first three games while Hinrich was +18. This link shows net rating for those three games. Through Game 3 of the Boston series, the Hawks outscored the Celtics by 3.7 points per 100 possessions with Hinrich on court, but were outscored by 21.7 per 100 with Green on court and 41.7 with Pargo.

But what does that even mean? Here’s a thought exercise to simplify net rating. NBA teams score about 100 points per game and use about 100 possessions per team to score those points. It’s not exact, although there was an NBA season in which teams averaged exactly 100 points. But teams don’t play at the same pace. Some teams play faster and use more possessions. This is exactly why net rating is useful. By regularizing plus-minus data to a per 100 possessions standard, you can make more of an apples-to-apples comparison.

So, the thought exercise: Imagine that every NBA game is supposed to end in a 100-100 tie at the end of regulation. Imagine that every NBA team uses exactly 100 possessions to score those 100 points. Net rating tells you how close to meeting that standard a team was with a given player on or off the court. Thus, if you look at Pargo’s offensive rating (59.4) and defensive rating (101.1) in the first three games of the 2012 Celtics series, this tells you the Hawks were on pace to score about 59 of those expected 100 points while giving up 101 points to the Celtics during the segments of the game when Pargo was on court.

Obviously not even Larry Drew is insane enough to play Jannero Pargo for 48 minutes of a playoff game. But Pargo’s net rating of -41.7 tells us that, while Pargo was on court in those three games, the Hawks were on pace to get blown out by nearly 42 points in our theoretical 100 point, 100 possession game. Hoopinion editor Bret LaGree explained it perfectly in his recap of Atlanta’s 101-79 loss in Game 4 of that series.

[Pargo] was on the court when the game got out of hand. Surprising no one other than Larry Drew. It's tough to score 8 points in 8 minutes and have your team go from 8 points down to 19 points down. Pargo managed the feat.

It’s easy to assume that Drew sat Hinrich in the fourth quarter of Game 3 because he hadn’t scored a basket the entire game. Drew was likely hoping for an offensive spark from Pargo. But his disastrous -7 stint in the fourth quarter allowed the Celtics to rally and tie the game at the end of regulation before prevailing in overtime. This cost the Hawks an opportunity to retake home court advantage with a road win after splitting the first two games at home.

In the box score, Pargo finished -11 while Hinrich had a plus-minus of zero. This means the Hawks were on pace for exactly that theoretical 100-100 game with Hinrich on court. What this tells me is that, while Hinrich may not have been pushing the Hawks toward a win with scoring, he also wasn’t hurting the team’s chances. And this is what I mean about contributing to winning basketball. If one of your non-star players helps you keep the score close when he’s on court, it gives your star players a better chance of closing out a win. Hinrich didn’t hurt the Hawks in his 26 minutes, so he helped give his team a chance to win.

Former Atlanta Journal-Constitution Hawks beat writer Michael Cunningham did a possession-by-possession breakdown of that game and determined that, while guarded by Hinrich, Rajon Rondo did not score a single basket or record a single assist. With Pargo on court and Hinrich on the bench, Rondo slashed to the basket, scored and assisted seemingly at will. Thus why stats such as net rating provide a more nuanced way to look at basketball beyond the traditional fixation with a player’s points, assists and rebounds. Did a team play winning basketball or losing basketball while a player was on court? In Games 1-3 of the 2012 Celtics series, the Hawks played winning basketball with Hinrich on court and were on pace to get blown out by over 40 with Pargo on court.

Introducing Net Differential

Having explained why net rating has become so fundamental to how I evaluate the NBA game, I’d like to introduce a new tool for player evaluation: net differential. This stat is obtained by subtracting a player’s off court net rating from the player’s on court net rating. The counter argument I always get is that plus-minus is impacted by the other four teammates and the five opponents when a player is on court. This is certainly true, but if one starter for a team finishes +10 points per 100 possessions for an entire 82-game season and another starter for the same team finishes -2, it should tell us something about the relative performance of those players.

Keep in mind that for every minute those two players are on court together and on the bench together, their net rating is identical. Any discrepancy is the result of minutes spent with one player on court and the other on the bench. To further shed light on such discrepancies, I decided to compare each player’s on court net against their off court net. While on court net is influenced by lineups, off court net is 100% isolated to the individual player. It’s a measure of how many points per 100 a team scores and surrenders to the opposition when an individual player is on the bench. Thus, for example, the Raptors are +9.2 points per 100 with Kyle Lowry on court and -3.4 with Lowry on the bench. This gives us a net differential of +12.6, meaning the Raptors are 12.6 points per 100 possessions better with Lowry on court compared with how the team performs with Lowry out of the game.

I don’t intend for net differential to take the place of catch-all stats such as Player Efficiency Rating and Win Shares. I look at this as simply another tool to influence the way we think about the NBA game. Net differential takes a lineup-influenced stat (on court net) and anchors it to an individual stat (off court net). There are obvious weaknesses. For example, Lowry has played 1680 minutes this season and has only spent 538 minutes out of the game. This means his on court net has a plentiful 1680-minute sample but his off court net has a much smaller sample.

By contrast, end-of-bench players will have a huge off court sample and a limited on court sample. This tells us that net differential for a player who has spent at least 1000 minutes on court and 1000 minutes on the bench will have a much more useful net differential than a player who has only played a few hundred minutes. For that reason, I have compiled net differential only for NBA players who have played at least 300 minutes this season. A tiny 300-minute sample may be somewhat useful, but I feel anything smaller won’t tell us anything.

Evaluating all 30 NBA Teams via Net Differential

What follows is the first part of a series of observations about all 30 NBA teams, in alphabetical order, influenced by what net differential tells us about individual players. I’ve included Kyle Korver’s numbers with the Hawks and with Cleveland, even though he hasn’t yet played 300 minutes for the Cavs. When Korver was traded, it immediately spawned a storm of think pieces from NBA writers who were convinced that it signaled the start of a rebuild for the Hawks. I found this point of view rather silly since Korver was easily Atlanta’s worst performer in net differential. I’m convinced that the Hawks crushed the Korver trade by getting a first round pick for the team’s least productive player AND Mike Dunleavy, who has outperformed Korver in net differential both as a Cavalier and a Hawk. These observations include thoughts on how each team should approach the NBA trade deadline and this summer’s upcoming free agency.


Atlanta Hawks

It should surprise no one that Paul Millsap leads the Hawks with a net differential of +12.5. What will be surprising to some are the next two players on that list. Tim Hardaway is second at +10.3 while backup point guard Malcolm Delaney is an astounding +8.4.

Throughout last season and into the recent preseason, I constantly talked up how superior Dennis Schroder was to Jeff Teague in net rating last season. The frequent counter argument, which I fully acknowledged, was that Schroder, as a reserve, played against second units while Teague played opposing starters.

Schroder now finds the script completely flipped this season. His -5 net differential is second-worst among Hawks with at least 300 minutes after the team jettisoned Korver. Meanwhile his backup, Delaney, is out-performing him in net rating just as he out-performed Teague last season.

It's tempting to conclude that this is all due to the difference between playing primarily against starters vs. reserves. But that's an oversimplification. I wrote a recent piece for BBallBreakdown in which I discussed the ways Al Horford's arrival has made Isaiah Thomas better. The Hawks led the league in fourth quarter net rating last season while the Celtics were average. Now, with Horford switching teams, suddenly the Hawks are closer to average while Isaiah Thomas is the NBA's leading fourth quarter scorer.

Additionally, the Hawks were at the top of the league in generating wide open shots (no defender within six feet) in every year of the Horford-Mike Budenholzer partnership. This culminated in 22.7 wide open shots per game last year for Atlanta, the highest number ever recorded. But with Horford gone to the Celtics, the Hawks suddenly slipped to 17.9 this season, the lowest under Budenholzer. Meanwhile the Celtics jumped from 15th in the NBA in generating wide open shots last year to third in Horford's first season in green.

Thus, part of Schroder's net rating decline is certainly due to facing more minutes against the NBA's elite point guards. But a big part of it is also the absence of Horford, whose ability to stretch the floor complemented Schroder, whose greatest strength is driving to the basket. While Thomas is enjoying more paths to the basket than he ever saw before Horford arrived, Schroder has had to learn how to play with Dwight Howard, who typically camps out around the lane and closes off driving angles that were available to Teague and Schroder last season.

Kent Bazemore's net differential of -4.7 is obviously disappointing for a player who signed in the offseason for four years at about $17.5 million per season. Budenholzer should move Hardaway into the starting lineup and use Bazemore as Atlanta's sixth man.

A major problem for Schroder in particular and Atlanta in general this season has been turnovers. Hardaway has a much tighter handle than Bazemore and the ability to create his own shot. Replacing Bazemore in the starting lineup would reduce some of the ball handling and shot creation burden for Schroder. Meanwhile, Bazemore might benefit from an increased role as a go-to scorer and ball handler against opposing second units.

Additionally, with Mike Muscala injured and Budenholzer forced to increase minutes for Kris Humphries, the latter has emerged as clearly the better option as Atlanta's third big man. With Korver gone, Muscala's -6.2 is the worst remaining net differential on the team.

And lastly, the Hawks clearly won the Korver trade. Dunleavy has played only 133 minutes as a Hawk but he's a respectable +0.4. Korver, predictably, has been Cleveland's worst net differential player, just as he was for Atlanta. I've included all players with at least 200 minutes for the Cavaliers to show how Korver is performing in his first 202 minutes for Cleveland. Korver's -16.1 net differential for the Cavs is even worse than the disastrous -12.5 he posted in 894 minutes as a Hawk.

Last season Korver was a workhorse for Atlanta, posting a +6.3 on court net rating in 2401 minutes that was second only to Schroder among Hawks with at least 600 minutes. His plunge from a +5.6 net differential last season to -12.5 as a Hawk this season was a clear sign that Father Time is here to hand out yet another L.

Thus, obtaining a net neutral player in Dunleavy in addition to a first round pick for Korver is yet another stunning win for the front office tandem of Budenholzer and Wes Wilcox. The pair pulled an upset by winning the Hardaway trade and turned a disgruntled, expiring Teague into a lottery pick. Danny Ferry stepped down as Atlanta GM on June 22nd, 2015 just days before the June 25th draft. Although Budenholzer's promotion to president of basketball operations and Wilcox' to general manager weren't announced until June 30th, they presumably ran that draft and pulled the trigger on the Hardaway trade.

In light of Hardaway's emergence as an elite performer in net differential, the report that Atlanta considered trading him for second round draft picks seems delusional. Budenholzer would be wise to increase Hardaway's role and see what he really has ahead of 2017 free agency. If Hardaway's on/off numbers aren't strained too much in a starting role, he's clearly a player the Hawks should extend or retain in restricted free agency.


Boston Celtics

Net differential tells us that Jae Crowder, not Isaiah Thomas, is Boston's most indispensable player. The Celtics are 11.1 points per 100 possessions better with Crowder on court compared to when he's on the bench. Meanwhile Boston is only +0.6 with Thomas on court compared to when he's out of the game.

Horford, Amir Johnson and Marcus Smart are all positive contributors among Celtics who have played heavy minutes, while Jaylen Brown has predictably struggled in nearly 600 minutes as a rookie. What's surprising is how much Avery Bradley (-5.5) has struggled. Injuries have marred his season and Boston is surprisingly almost five points per 100 better defensively when Bradley goes to the bench.

Crowder has possibly the best contract in basketball. He’s locked up through next season at a piddling $6.2 million this season and $6.7 next. Danny Ainge should hang up the phone if anyone calls about Crowder if they aren't offering something close to an All-Star.

Otherwise, this is a pretty balanced roster. Horford is paying obvious dividends and while Brown isn't a breakout star, he hasn't embarrassed himself. He looks like a solid rotation player for the foreseeable future in Boston. The Celtics should bide their time, carefully consider any offers and continue to build through the draft via the lopsided trade that sent Kevin Garnett and Paul Pierce to the Nets for what has turned out to be multiple high lottery picks.

Unless an established star becomes available via trade, Ainge should continue to look for a star in the draft while also hoping for a reunion between coach Brad Stevens and his former player Gordon Hayward. As with Horford last summer, Hayward can be an unrestricted free agent this summer and the Celtics could potentially sign him without surrendering significant assets. A new Celtics dynasty becomes a possibility depending on if Ainge gets his draft picks right or packages one or more for the right impact player.


Brooklyn Nets

With what will probably be high lottery picks from three consecutive drafts surrendered to the Celtics in the most lopsided trade since the Hawks gave Bill Russell to Boston, the Nets must build through free agency, internal development and wise selections later in the draft. I cringed when Brooklyn traded starting power forward Thaddeus Young to the Pacers for the pick that was used to draft Caris LeVert last summer, but it's looking like a defensible move.

LeVert leads the Nets in net differential (+8.9) and 2016 second round pick Isaiah Whitehead is next at +7.6. Brook Lopez, the only obvious NBA starter on the roster, is a respectable +5.3. Jeremy Lin has been limited to 300 minutes due to injury, but he's respectably in the positive for those minutes. Rondae Hollis-Jefferson hasn't made the same defensive impact he did as a rookie, but he's also been a positive contributor.

The bad news for Brooklyn has been the play of Bojan Bogdanovic and Trevor Booker. The latter signed as a free agent last summer for two years and $18 million. It was hoped that he would fill some of the void left by the trade that sent Young to the Pacers, but Young has been a much stronger player for Indiana than Booker has been for the Nets. It was hoped that Bogdanovic would emerge as the shooting guard of the future for Brooklyn but the Nets are 4.2 points per 100 better when he goes to the bench.

Although it will be painful for Nets fans to watch more picks and potential stars go to the Celtics after the disastrous overreach by former GM Billy King, those fans should be encouraged by the talent evaluation that has brought LeVert, Whitehead and Hollis-Jefferson into the fold. These may not be emerging stars, but they appear to be long term building blocks that have this Brooklyn rebuild off to a better start than the worst case scenario.

Lin's lost season obviously hurts as the Nets were relying on him to energize the fans in the absence of any real hope of competing. But injuries are part of it. Brooklyn must continue what it is already doing and find more hidden gems in the draft and more impact free agents.


Charlotte Hornets

Kemba Walker's All-Star nod is well deserved but Cody Zeller has been a revelation for Charlotte. The Hornets have also been bolstered by the return to health of Michael Kidd-Gilcrhist, who has soaked up over 1300 minutes so far. MKG combines with Nicolas Batum and Marvin Williams to give Charlotte one of the deepest small forward rotations in the league.

Although Williams has primarily moved to power forward, he's still more than capable of spot minutes at small forward, the position where he played most of his career. Second year power forward Frank Kaminsky is slightly in the negative but has established himself as a rotation player with over 1000 minutes played. Roy Hibbert has also performed slightly south of zero and has only played 628 minutes, but he adds depth to a Charlotte front line that doesn't scare anyone with its athleticism.

Marco Belinelli stretches the floor with his outside shooting, but he's a net negative due to the drop-off Charlotte's defense suffers whenever he's on the floor. The biggest problem for the Hornets has been the disappointing performance of Ramon Sessions as Walker's backup. Spencer Hawes has been a disaster but Charlotte hasn't needed much out of him with the solid-if-unspectacular play of the power rotation players ahead of him.

Altogether this is a solid and balanced Hornets roster that is a star player away from being contenders. It will be tough to find that player in the draft since they don't look like a team headed for the lottery. What Charlotte could really use via trade, free agency or the draft is an upgrade at backup point guard and a rim protector who is more mobile than Hibbert.


Chicago Bulls

What better time to evaluate Chicago's roster than now, after the organizational meltdown triggered by Wednesday's brutal loss to the Hawks? Jimmy Butler and Dwyane Wade lit into the younger players on the team afterward, but a look at net differential for the season casts doubt on Wade's fitness to author such criticism.

Nikola Mirotic obviously had a horrible night shooting the ball, going 3-for-11 from the field and 1-for-5 from three-point range. Ironically, his +3 in the box score led the team. Meanwhile Butler scored 40 points but was -3 while Wade scored 33 and was only +1.

Looking at net differential for the season, Butler is very obviously the team's standout performer at +12.1. But guess who's second on the team? The Bulls are 5.9 points per 100 better with Mirotic on court than on the bench, which trails only Butler. Next up is Doug McDermott with a net differential of +2.5, followed by Jerian Grant's +2.

The young members of the Chicago roster are well represented at the top of the net differential list. So where's Wade? The Bulls have been 4.6 points better per 100 possessions with Wade on the bench than they have been in his 1284 minutes on the floor. Rajon Rondo has been a respectable -1 in 1037 minutes and is clearly a better option than Michael Carter-Williams (-4.1). However, Jerian Grant in 540 minutes has made a strong case to move to the head of the point guard rotation with a +2 net differential.

The Bulls seem like a good candidate to make a panic trade and it seems no one in the organization has any idea what they have in Mirotic. The Hawks have shown the league the value of big men who can stretch the floor from Horford to the less-heralded Pero Antic, who played a big role in the first half of the 60-win season before injuries prevented Antic and Thabo Sefolosha from closing as strong as they started. Even though Antic shot below league average from three, there was value in pulling an opposing big man out of the middle of the lane. If the Bulls are selling, Mirotic should be high on any shopping list.

Keeping Wade company at the bottom of the net differential list is fellow veteran Robin Lopez. With only two seasons remaining on Butler's contract before he can opt out in 2019, this Bulls roster is a ticking time bomb. Wade has already gone into predictable and irreversible decline. Rondo, in sticking up for the younger players, has highlighted the lack of a unified voice among the team's veterans.

What once seemed like a franchise on the rise, ready to ride the crest of Butler's stardom, now looks like a dumpster fire. From the failures of leadership to the front office's counter-intuitive refusal to embrace the rise of pace and space, the Bulls have become a model for how to wreck a promising franchise.


Cleveland Cavaliers

Speaking of teams going through a mid-season existential crisis, whew lawd, the Cavs. Korver should have been better as a Cav than he was as a Hawk. Dennis Schroder has yet to prove that he can competently protect the ball and set up teammates. Korver goes from a team with a point guard experiencing major growing pains to an established All-Star in Kyrie Irving.

Additionally, the Hawks have fallen off a cliff as a three-point shooting team, ranking 25th in the league. As such, opposing teams were able to key on Korver even more than in past seasons since the Hawks boasted fewer than a handful of players shooting above league average from deep.

But in 202 minutes as a Cavalier, it's become even clearer that this is the end of the road for Korver. Opposing teams are attacking his defense practically every trip down. LeBron James and the Cavs shouldn't be surprised by this since they employed the same tactics against Korver as a Hawk.

What is surprising is just how bad it's gotten in Cleveland. The Cavaliers are -12.5 points per 100 with Korver on court and +3.6 with Korver on the bench. Not helping matters is that Channing Frye, another sharpshooter the team added for spacing, has a net differential of -12.7. JR Smith at +6.4 remains in the top four in net differential. Korver hasn't replaced him and neither has Iman Shumpert (-6.1).

Further, a look at the top of Cleveland's net differential list highlights why Carmelo Anthony won't likely be joining the team any time soon. Kevin Love's +14.5 net differential exceeds even the King's +12.1. Anthony's -1.8 puts him squarely in the bottom half for the Knicks. A couple seasons back, Doc Rivers scoffed at a rumored Blake Griffin for Carmelo Anthony trade rumor. There's a reason those trades won't happen. Love and Griffin are much better players than Anthony.

Up next: Dallas through Indiana.

No comments:

Post a Comment