Advanced stats are the joy of some fans and analysts and the bane of others. While the numbers reveal much, some of what they “reveal” is highly misleading. And when people use those misleading things as “evidence” in an argument, it turns the reasonable off.
The thing we need to bear in mind is that advanced stats aren’t designed to prove arguments. They simply show what things happen, not why things happen. That’s where analysis steps in.
Looking at which players “cheat” the stats and how they do it can give us fresh insight into both the numbers and the “basketball” of the conversation.
The important thing to remember is the “why” doesn’t replace the “what.” Statistics are not opinions. At their heart they are nothing more than recording events (e.g. a block, a rebound, a basket, a missed field goal, etc.). They are not distinct from what happens on the court because they are just annotations of what happened.
We can’t arbitrarily brush aside the things that happen because they don’t fit what we expect. We have to understand why they happened.
Here are three examples of players who cheat the metrics and how they do it. In each case, the numbers can bring us to a misleading conclusion, but when properly understood, they make a telling point.
Is Brandan Wright Really Better Than LeBron James?
Prior to being traded to Boston, Wright was on the Dallas Mavericks under coach Rick Carlisle, who utilizes players' abilities as well as any coach in the league.
Some coaches, like Tom Thibodeau of the Chicago Bulls or Gregg Popovich of the San Antonio Spurs, are masters at adapting players to their systems; Carlisle builds a system around every group of players he has on the court. Such is the argument Bobby Karalla makes for BBallBreakdown:
That’s the beauty of Carlisle’s flow offense. There’s no rule stating 'you must pass here, then run there, set a screen, run to the corner, and wait for the pass.' Everything that happens depends first on the defense, and next on the five-man group’s own tendencies. And, lucky for him, he has the depth and quality to execute such a system. As the numbers suggest, that decision-making process seems almost impossibly random across an individual game, too random to prepare for.
Wright does a few things really well. For example, he scores with incredible efficiency close to the rim. He’s 90-of-115 within five feet, per Basketball-Reference.com. He’s made 8-of-16 from six to 10 feet. He’s made two from outside 10 feet. Overall, 74.6 percent of his shots are assisted.
So while he’s efficient as all get out, it’s in extremely isolated situations.
Per Karalla, when Dirk Nowitzki sits, one of Carlisle’s favorite things to do is play a “Wright/[Tyson]Chandler frontcourt which almost exclusively runs 'Horns' sets and chucks threes in bulk—the club plays at a crawling 91.28 possessions per game when those two share the floor, per NBA.com.”
That plays into the strength of Chandler and Wright, who are great offensive rebounders and efficient on putbacks. Lots of threes means spread defenses, which gives way to more offensive rebounds and, in turn, more shots at the rim.
In fact, per NBAWowy.com, 24 of Wright’s 100 field goals came directly off snaring missed Mavs shots.
Carlisle asks his players to do what they’re good at and nothing else. His genius is that he’s able to fit pieces together in such a way that they assemble a whole. Wright thrives in that kind of environment because he’s never asked to do anything more than what he can.
PER is a prorated stat. But in Wright’s case you can’t extrapolate what he does onto a larger sample because that would mean playing in more situations that are less suited to his talents. His production would (and will) inevitably fall.
Wright’s PER is more a testament to how well Carlisle coaches than how well Wright plays. So the correct conclusion to draw from Wright’s stellar PER is that Carlisle is a genius.
Last year, James Harden of the Houston Rockets got the worst kind of “fame” the Internet can bring you thanks to a video lowlighting some of the worst defense (if you can call it that) you’ll ever see. Harden became synonymous with lack of effort.
This year, shockingly, he’s tied with Draymond Green of the Golden State Warriors for the league lead in defensive win shares at 2.0. Does that mean Harden has become an elite defensive player? In the words of Lee Corso, “Not so fast, my friend.”
But let’s not throw out The Beard with the bathwater here. There’s a little bit of area between being one of the worst and best defenders in the league.
What’s helpful here is realizing how defensive win shares are determined. The details are almost obscenely complicated. But the nutshell version is easy to explain: They utilize individual and team defensive ratings to formulate an estimate of how many wins a player adds to his team. If you want the specifics, you can get them here and here.
Most crucially, there is one huge qualifier to all of it, and it’s noted at Basketball-Reference:
Out of necessity (owing to a lack of defensive data in the basic box score), individual Defensive Ratings are heavily influenced by the team's defensive efficiency. They assume that all teammates are equally good (per minute) at forcing non-steal turnovers and non-block misses, as well as assuming that all teammates face the same number of total possessions per minute.
Win shares also are helped by the average point differential a team has. So, a player getting a large number of minutes on a good defensive team is going to get defensive win shares, even if his individual defense is bad.
Additionally, the only true box-score numbers that influence defensive win shares are blocks, steals and defensive rebounds. Thus, players who accrue those can create the illusion of being good, even if it’s in the process of actually playing bad defense.
There may be no greater testament to this anomaly than Carlos Boozer’s fifth-place finish in 2011-12 with the Chicago Bulls’ top-rated defense. He collected 8.3 defensive rebounds on an elite defense, and that’s about all he offered to it.
Harden’s defense this year is actually decent, but it’s not great. Opponents are shooting 2.9 percent below their season averages against him. He’s averaging two steals and a block per game. He’s mentally engaged.
But he’s usually left to guard the weakest perimeter player on the other team. Trevor Ariza and Patrick Beverley account for the better two.
Still, what we can see is that Houston is able to build an elite defense (ranked No. 2) with him in it, and his win shares do prove that.
Is DeMarcus Cousins the Second-Best Player in the NBA?
Plus-minus numbers are simple enough in concept. If a player’s team outscores its opponents by three points per 48 minutes while he’s on the court and is outscored by four while he sits, the difference between those numbers is seven. Ergo, he’s a plus-seven. If the opposite happens, he’s a minus-seven.
Theoretically, the better the plus-minus, the better the player.
The nice thing behind the concept is that it attempts to measure things that don’t show up in box scores. How well a player sets screens might not show up in his stat line, but solid picks create better shots for his teammates. That means more points for them.
Or, a player might get two steals per game, but in doing so he shoots passing lanes, and that results in opponents driving by to score easy points. Thus, a negative (opponents’ points) looks like a positive (steals) in the box scores.
Plus-minus numbers are intended to convey the hidden impact of a player on the court. They ideally reveal those extra points scored on screens or surrendered by senseless gambling on defense.
The problem with plus-minus numbers is that they are noisy. They would be perfect if all other things were equal, but that’s not the case. When starters play, they tend to play with other starters. When benches play, they play with other reserves. And the competition is going to similarly vary their lineups.
Drawing on advanced statistical modeling techniques (and the analytical wizardry of RPM developer Jeremias Engelmann, formerly of the Phoenix Suns), the metric isolates the unique plus-minus impact of each NBA player by adjusting for the effects of each teammate and opposing player.
The RPM model sifts through more than 230,000 possessions each NBA season to tease apart the 'real' plus-minus effects attributable to each player, employing techniques similar to those used by scientific researchers when they need to model the effects of numerous variables at the same time.
To a large degree, that’s effective but not completely so. In extreme situations, anomalies still occur. That’s because, at its core, real plus-minus is going to convey the difference between starters and bench players, and there’s only so much that can be done about that.
There end up being extreme situations that can trick RPM. One of those is this case with DeMarcus Cousins of the Sacramento Kings, who is currently second in the stat at plus-6.78. While most would agree that he’s come a long way, few would suggest he’s the second-best player in the league.
Part of the reason Cousins’ RPM is so high is individual success, but it’s not the whole reason. The Kings’ starting lineup with Cousins, Darren Collison, Rudy Gay, Ben McLemore and Jason Thompson is the second best in the league with a plus-6.8, per NBA.com/Stats.
Cousins missed 10 games with viral meningitis. While he was out, the starters were outscored 388-374. The players who replaced him scored a total of 62 points in 182 minutes with the other four starters. Comparably, in 210 minutes with the starters, Cousins has 135 points. That accounts for almost 40 percent of his 580 total minutes played.
So what you have here is a confluence of events. Cousins looks really good while he’s with the starters but has missed significant time. Thus, he has a more significant portion of his minutes with them than they do with him. His replacements are vastly inferior, so the other starters’ plus-minus numbers take a hit by playing with them.
Those things combine to boost Cousins’ RPM, giving the illusion that he is exclusively responsible for the starting lineup’s success, when he’s really just a big part of it.
That’s not to say RPM is worthless. It tells us a lot if we look beneath the surface. By analyzing why Cousins’ RPM is so high, we discover the Kings have a quality starting lineup and a big need for an upgrade at the backup 5. RPM as a player-ranking metric has limited value, but in terms of telling us what happens within a team construct, it is invaluable.
In each of these cases, there’s a window into the larger picture of understanding how advanced stats work. They may not always lead to the kinds of easy conclusions we’d like them to make, but closer scrutiny indicates that there is still a lot to be gleaned from them—even in the anomalies. It all just comes down to discovering the why.