#NBAMathMail: Diving into Twitter Questions for Mailbag No. 1

by Adam Fromal in NBA Math Mail

Welcome to the first NBA Math mailbag, where we’re ruminating on NBA statistics and players to help provide the answers you’ve always wanted to know.

Want to participate in future editions? Send us your questions about anything related to basketball by using the #NBAMathMail hashtag. And when we say anything, we mean anything. Need to know about advanced metrics? Have a question about your favorite player? Hit us up with them, and feel free to get even more creative with what you ask. The best questions will be featured in each mailbag.

But for now, we’re diving into the first batch of inquiries.

Best Overall Metrics

I wish I had a definitive answer to this question. But if one existed, it would be the unanimous choice of every basketball analyst on the planet. We’d all be using that exclusively to come up with rankings, and NBA general managers would hand out contracts and trade proposals based on it. Sadly (or happily, depending on how you look at it), we’re instead left scrambling to make sense of a large collection of lesser numbers.

Ultimately, no all-in-one measure is perfect. That’s what makes this so much fun.

Our TPA metric is based solely on box-score statistics, and plenty gets left out of its scope. How do you quantify a player’s gravitational pull when he’s standing in the corner and dragging players out of the lane with the mere threat of a shot that never comes? That’s not going to show up in a box score. So many different facets of the game are omitted, even if the stat serves as a great ballpark estimate of a player’s value based on both per-possession efficiency and volume. Defense is also a huge weakness, and the defensive points saved portion of the metric should never be treated as gospel.

Understanding the flaws of metrics is always crucial to using them properly. You have to know that PER rewards players who shoot frequently (and almost entirely overlooks defense), since the cut-off for field-goal percentage is set a less-than-ideal level. Win shares are rudimentary metrics, but they can still be useful if looked at within the context of how many victories that player’s team accumulated. RPM—which is not a set of rankings—is still subject to on/off measures and is inferior to multi-year models. On/off numbers, for that matter, are entirely devoid of context and can be heavily influenced by the players accompanying the man in question.

A lot of good stuff is scattered throughout the interwebs (more on that later). But ultimately, it’s all useless unless you’re properly accounting for the flaws and applying contextual evidence. That’s why analytics and the eye test are in no way opposite ends of a spectrum, but rather two useful tools that should exist harmoniously.

Best Defensive Metrics

Honestly? Game film, if you can call it a metric.

If the all-in-one statistics are rife with issues, the defensive ones take that to the proverbial next level. Please, don’t use individual defensive rating or defended field-goal percentages. Nylon Calculus’ Krisha Narsu talked about the latter near the beginning of 2017:

Perimeter defenders do affect whether the shooter makes the shot with proximity. On average, players shoot worse as the distance between them and a defender decreases. The problem is that that Defended FG% captures that and about seven other variables. We can’t simply use Defended FG% to determine if a player is a good defender or not. Rather what we should be looking for is which players are consistently good at deterring shots, forcing turnovers and disrupting the offense.

But even the more “advanced” metrics have significant flaws. Our own defensive points saved, for example, factors in defensive rebounding and is intended to serve as nothing more than a baseline, since it’s calculated not by looking at defensive work, but rather subtracting offensive value away from total value and assuming the difference can be assigned to defense.

Play-type data seems like it should be more valuable than it is. But even if someone is stifling in isolation, their numbers won’t show how often they forced their foes to pass the rock away. Deterring shots is arguably the most valuable skill of all, but that’s not contained within the scope of the play-type analysis.

And that brings us back to the eye test. Without it, you might as well avoid analyzing defense. Avery Bradley and Klay Thompson tend to fare poorly in many defensive metrics, but watching tape makes you realize how erroneous those numbers can be. Context is king here, since it helps identify whether players are covering up for porous teammates, taking on tough responsibilities, impacting the game in off-ball scenarios that can’t truly be quantified and so much more.

If you’re looking for a great example of high-quality defensive analysis, check out Matt Moore’s detailed breakdown for CBS Sports of Kawhi Leonard’s middling numbers and shocking on/off impact during the 2016-17 campaign.

Negative TPAs

First of all, I love this question. It allows me to address one of the fundamental misconceptions of TPA, because a negative score isn’t necessarily a bad thing.

With this metric, a value of zero doesn’t indicate replacement-level contributions, but rather league-average work. And because the best players tend to play the most minutes, that’s skewed in the positive direction in such a way that playing “average” ball for hefty minutes is still an impressive accomplishment. Plus, no team—not even the 2017-18 Golden State Warriors—is ever going to boast a rotation comprised entirely of above-average players; the depth of talent and salary-cap restrictions stifle that possibility.

So imagine that someone is slightly below average on a per-possession basis. They also suit up for all 82 games and play 20 minutes per contest, essentially functioning as a crucial depth piece. That hypothetical player is quite valuable, but his score will inevitably wind up looking quite negative. Again, context matters.

Additionally, players can fit into specific roles that allow for them to add value in certain areas. Offensive specialists may have negative overall scores because of their defensive woes, but they’re still crucial to their teams because of their knack for getting buckets. Take Jamal Crawford’s career as an example:

Crawford has only posted a positive TPA in five of his 17 professional seasons. His career score of minus-658.39 leaves him behind 3,069 players. Directly ahead is Norris Cole, whose lifelong figures look like this:

Is Cole really “better” than Crawford? Of course not, even though his scores are technically superior!

Almost every Crawford season appears in a useful quadrant, since he was above-average on offense for so many years. He also played quite a bit more basketball, which forces him to deviate more from the origin than a career role player—true both on a single-season and career scale.

Anything negative automatically has, well, negative connotations. But with TPA, that shouldn’t be the case in every situation.

Limited Scope of Assists

I wholeheartedly agree with the sentiment here.

Not all assists are equal. Those that lead to three-pointers should be considered more valuable than those generating two-point buckets. Passes that lead directly to free throws should also be rewarded. Fortunately, NBA.com’s fantastic stat databases already do this—and also show crucial information like secondary (or hockey) assists. Here are last year’s leaders in assist points created per game:

  1. James Harden: 27.1
  2. John Wall: 25.3
  3. Russell Westbrook: 23.8
  4. LeBron James: 22.7
  5. Chris Paul: 21.7

I don’t think you’ll ever see free throws incorporated into the raw assist stat because it would throw off far too much of NBA history, making it easier for players to record triple-doubles and shatter previous records. John Stockton’s career mark would suddenly be touchable. But that data is at least out there in publicly accessible fashion.

Nikola Jokic’s Trend Within Denver Nuggets Franchise History

If you’ve even been a casual follower of NBA Math, you know we’re all aboard the Nikola Jokic train. Need reasons why? First, check out this compilation of his assists from 2017-18:

Next, look at how his TPA scores stacked up with the rest of the league during each of his two professional seasons:

At this point, you’re perfectly justified to ask where he could rank in Denver Nuggets franchise history. He’s already within the top 10 in career TPA for the organization:

  1. Fat Lever: 1,735.64
  2. Marcus Camby: 997.51
  3. Dan Issel: 956.51
  4. Dikembe Mutombo: 867.33
  5. Nene: 865.74
  6. Bobby Jones: 712.77
  7. Michael Adams: 691.03
  8. T.R. Dunn: 608.5
  9. Nikola Jokic: 508.08
  10. Reggie Williams: 422.18

Yes, that leaves him ahead of notable figures such as Carmelo Anthony, Danilo Gallinari, Ty Lawson and Kenneth Faried, as well as plenty of short-tenured figures. But climbing all the way to the top of the franchise leaderboard and surpassing the tremendously underrated Fat Lever still makes for a tough task.

Jokic earned 165.84 TPA as a rookie (the No. 30 season in franchise history) and then added another 342.24 (No. 6) in 2016-17. That latter score trails only Bobby Jones’ 1976-77 and four years from Lever. If he’s able to post identical numbers going forward, he’ll need just over 3.5 seasons to become the all-time Nuggets leader, pushing to the top slot sometime during the 2020-21 campaign.

Injuries and regression could throw this off, of course. Ditto for him signing with a different organization, though the Nuggets will obviously do everything in their power to prevent that nightmare from becoming reality. But even if he regresses all the way to his rookie figure, he’d need a bit more than seven seasons to surpass Lever. At that point, he’d still be in his 20s.

Unless something goes wrong, he should almost assuredly get there.

Meaning of Value Added

The value-added portion of play-type data is what differentiates NBA Math from other sources. Essentially, we’re showing how many more points a player added in that specific play type, as compared to a hypothetical league-average contributor using the exact same number of relevant possessions.

But this is most clear when using a specific example. So let’s take Otto Porter Jr.’s work as a spot-up shooter, since he led the league in value added for that specific play type.

The Washington Wizards small forward scored a whopping 397 points on 303 spot-up possessions, giving him a (rounded for the sake of textual convenience) 1.31 points per possessions (PPP) in spot-up situations during the 2016-17 campaign. The league as a whole scored 52,526 points on 52,600 possessions, which equates to an even 1.0 PPP. Rounding errors have led to a slightly different figure appearing in our databases, but the calculation is as follows.

Porter’s 1.31 PPP gives him 0.31 PPP more than an average spot-up player on his typical possession. That’s subsequently multiplied by the number of possessions he used (303) to get to 93.93 value added as a spot-up marksman.

And if you’re looking for exactly how we’d phrase that, here are a couple options:

  • Porter added 93.93 points above expectations as a spot-up shooter.
  • Porter added 93.93 more points than an average player would’ve with his spot-up possessions.

That last part is crucial, because he didn’t technically add 93.93 more points than an average player—if the sentence ends there. You have to specify that value added stands in contrast to someone using the same possession distribution, whether by comparing to expectations or the more explicit explanation.

Per-36-Minute Stats

Congratulations to Romeo Alcantara II for asking a truly impossible question!

Unfortunately, we have no way of answering this. If Player A is logging just 15 minutes per game, we can only speculate as to what he’d do when averaging 36 minutes. If he does so the next season, you can’t realistically compare those numbers to the previous season’s in a vacuum because so many confounding variables exist. His teammates likely aren’t the exact same, and the same is true of their minute distributions while he’s on the floor. The schemes are different. Ditto for the opponents he’s facing.

So if you want to know exactly what he’d average when playing 36 minutes per game, I can’t really help you. Given the tradeoff between volume and efficiency (both on a per-possession and per-minute basis), he’d likely come in shy of 20 and 10. But how shy? Your guess is as good as mine, and both require loads of contextual evidence.

Was Player A suiting up largely against backups? Because if that’s the case, he’ll likely see his numbers decrease more substantially when facing starters. Did Player A have an extreme usage rate in his 15 minutes per game? If that’s true, he’ll also see a larger decline because he’ll probably be spending time alongside stronger contributors with a role increase.

This might feel like a cop-out answer. But every situation is different, and no universal translation exists.

Other Places for NBA Numbers

I would start by going and reading this article by Ben Dowsett on Basketball Insiders. Next, I would bookmark that article. After that, I would read it again.

Seriously, that’s about all you need to do if you’re looking for a comprehensive breakdown of the various sites that host different NBA statistics. I’d also add that checking out Nylon Calculus is a great idea if you’re looking for not just data hosting, but also informative and novel breakdowns using numbers.

 

Adam Fromal is the founder and Editor in Chief of NBA Math. Follow him on Twitter @fromal09. 

Follow NBA Math on Twitter @NBA_Math and on Facebook.

Unless otherwise indicated, all stats are from NBA Math or NBA.com.