I’m one to bury the lede, so I’ll start by presenting some All-NBA teams:
|1st Team||2nd Team||3rd Team|
They definitely differ from my peers’ lists at Premium Hoops, but this seems reasonable, right? I’m sure a healthy chunk of you are somewhere screaming about Towns. Perhaps more of you are angry about Gobert being above Embiid. Maybe besides that, can we all agree that this is a solid starting point for an All-NBA team? Okay, let’s try this set with some tweaks.
|1st Team||2nd Team||3rd Team|
Despite your reactions, you have to admit that between these two teams, most of the right names are there. Of course some great players will be left off, but that’s going to happen regardless. Now, here comes the fun part.
All of those teams were constructed solely with stats. That’s it. No eye-test. Turning away from narratives. Ignoring wins.
My goal today isn’t to convince you that we should bow down to our algorithm lords (we’ve already done that); it’s to make you question your own mental calculus when picking your end-of-season awards.
Whether you like it or not, every decision in life is based on a mental calculus. By this I mean that we constantly weigh certain options against counterfactuals, consider confounding variables, and adjust how much one variable impacts a certain outcome. We end up conceding power to algorithms because they simplify these decisions that are clouded by human emotions by relying only on cold, hard software.
While an uncritical usage of algorithms is dangerous, an under-discussed aspect of this “eye-test vs. analytics” debate is that you are using an internal mathematical formula when you make decisions. That mental calculus is just as flawed (if not more so) than “advanced” statistics. Let’s say that you value peak performance and team performance most when picking your All-NBA teams. Are you confident that you’re weighing each player’s peak value or his team’s performance the same? If you are, can you clearly and coherently go through each member of your teams and back each one up with the same, steady reasoning? My guess is no because you’re a human, and if anything, we’re defined by our ability to have a “team,” live with contradictions, and drown in emotions.
This is where a metric steps in: to consistently apply the same standard regardless of outside factors which is exactly how I formed my aforementioned All-NBA teams.
Introducing Total LEBRON Impact
When making a decision, it’s important to be honest about what you value as an input. For Total LEBRON Impact (TLI), I decided to only focus on a player’s total impact throughout a season because that’s what defines the truly best players. (For a larger scale discussion on this point, check out Ben Taylor’s post mortem on his Top 40 GOAT list).
Like I discussed in my previous defense article, Basketball Index’s LEBRON is my favorite impact metric, so I picked it to measure impact. However, I couldn’t just rely on it because it normalizes everyone’s impact on a per 100 possession basis. To fix this, I created TLI with the following inputs:
LEBRON*(Total Possessions/100)=Total LEBRON Impact
Since LEBRON measures impact on a per 100 possession basis, this metric determines the number of 100 possessions a player played and multiplies it by his per 100 LEBRON impact for, you guessed it, a look at his Total LEBRON Impact. For instance, if a player’s LEBRON is +1.0 and he played 3,000 possessions during the season, that would mean that he had 30 clusters of “100 possessions” of a +1.0 impact making his TLI +30.
What I like about this metric is that players who are immensely impactful without playing the full season can still yield league-leading impact. It of course rewards players that are both very impactful and consistently available. This is how I established the first All-NBA team at the beginning of this article, which I’ll post again here with their TLI scores.
|1st Team||2nd Team||3rd Team|
|Guard||Curry 242.45||Lillard 181.99||Paul 165.28|
|Doncic 185.23||Kawhi 174.56||Westbrook 147.58|
|Forward||Giannis 237.96||Tatum 191.183||James 165.87|
|Jokic 334.90||Butler 188.11||George 154.91|
|Center||Gobert 272.18||Embiid 187.89||Towns 124.79|
Only four players had a TLI above 200, and only Jokic had one above 300. His MVP should be unanimous.
Also, I want to address Gobert vs. Embiid. Is it really that wild to think that Gobert was more impactful across the entire season? Gobert played 20 more games than Embiid (more than 1,000 possessions), and he had maybe the most impactful defensive season in the modern NBA. Again, this is where I want you to check your mental calculus for inconsistencies: do you really think that Embiid’s impact was so much greater than Gobert’s that it surpassed the value of 20 games?
Introducing Game LEBRON Impact
Let’s now pretend that I had a change of heart and that I actually value a player’s peak impact more than anything. To accomplish this, I made a second metric called Game LEBRON Impact (GLI). Besides having a minimum floor of 1,000 minutes, the only thing it measures is how many points per game a player is worth meaning that I’m devaluing “the best ability is availability.” Here is the input for GLI:
LEBRON*(Possessions per game/100)=Game LEBRON Impact
Following this metric, you get the second group of All-NBA teams from the beginning of this article. Here they are again with their GLI. (Note: I also adjusted for games played, so a player could only make 1st team if he played in 80% of games, 2nd team if he played in 70%, and 3rd team was open to everyone).
|1st Team||2nd Team||3rd Team|
|Guard||Curry 3.85||George 2.87||Chris Paul 2.36|
|Kawhi 3.36||Lillard 2.72||VanVleet 2.54|
|Forward||Jokic 4.65||Butler 3.62||LeBron 3.69|
|Giannis 3.9||Doncic 2.81||Tatum 2.99|
|Center||Gobert 3.83||Embiid 3.68||Towns 2.49|
Even on a per game basis, Jokic is leaps and bounds ahead of everyone as he has the only GLI above 4. Now, let’s talk Gobert and Embiid again. You might think this whole exercise is beyond sketchy based solely on Gobert’s impact relative to his peers. Perhaps LEBRON is off about his impact, but as we’ve talked about multiple times on Sense and Scalability, Gobert is maybe the only player in the league who is a defense unto himself. The Jazz can build a roster of offensive-focused players because Gobert covers all of their mistakes. If you view him as a defensive James Harden, I don’t think his immense impact is that farfetched.
For anyone interested, here are the top 27 players by raw LEBRON charted with the GLI and TLI. Our MVP is the outlier.
Raging Against the Machine
I’m neither a prophet nor a mind-reader, so I can only hope that you’re shaking your head at my stat-selected All-NBA team for two reasons. First, I want you to be angry at a couple of my selections (Gobert over Embiid? KAT? VanVleet?). Second, I want you to realize that most of the All-NBA selections based on those two metrics are in the ballpark of critical consensus.
Regarding the former concern, I want to revisit my discussion of Gobert vs. Embiid (for the third time in this article). Here are my reasons based on these metrics:
- Gobert’s LEBRON is better than Embiid’s
- Gobert played nearly 1,200 more possessions than Embiid
- Gobert’s TLI is better than Embiid’s
- Gobert’s GLI is better than Embiid’s
The evidence might seem like some spreadsheet mumbo jumbo, but at least each one is backed by data that you can openly double-check by looking at the undergirding philosophies that I used to build them. If you don’t build a tangible algorithm, you cannot do the same with your mental calculus. The philosopher Wittgenstein spoke of a similar idea with his “Beetle in a Box” thought experiment.
Suppose everyone has a box that only they can see into. No one can see into anyone else’s box. Each describes what he or she sees in the box as a ‘beetle’. I know what a beetle is from my own examination of what is in my box, you from yours. Wittgenstein points out that in this situation while we all talk about our beetles, there might be different things in everyone’s boxes, or perhaps nothing at all in some of them. The thing in the box, could be changing all the time. Whatever it is, he maintains that it cannot have a part in the ‘language-game’.Wittgenstein and the Beetle in a Box, Virtual Philosopher, 2006
When analyzing NBA players, your mental calculus is the beetle in a box. We all can talk about impact, peak performance, winning, and the like, but none of us can directly communicate what we mean by these terms. Heck, none of us can communicate these terms to ourselves making it impossible for us to establishing a consistent and unbiased feedback loop. Building an algorithm allows you to tinker with these inputs depending on the ongoing data that you receive.
Growing Skepticism of One’s Mental Calculus
I’ve posed this question a couple of times, but I’ll ask it again here. How many games do you need to watch before you’re confident in your valuations? For the sake of argument, let’s go with 10 games (approximately 1/8th of the season). If you have League Pass, let’s say you can watch a game in 1.5 hours. Watching all 30 teams once would take you about 22.5 hours. To see them all 10 times, that’s 225 hours or the equivalent of 5.6 40-hour weeks.
Now that you’ve watched 225 hours of NBA basketball, you go through your copious notes since it’s impossible to remember everything you’ve seen. From those notes, you meticulously compare and rank each player until you establish the perfect All-NBA teams. After doing all that, I ask you one question. Why are you so confident that your All-NBA teams are more accurate than the ones constructed solely from TLI and GLI?
This isn’t meant to be a snarky mic drop (though it looks, smells, and tastes like one). Instead, I want it to be an invitation to check your mental calculus when evaluating players.
A Rejection of All-NBA Discourse
When blathering about numbers and All-NBA teams one day, my Sense and Scalability host, (or, dad?) Scott, simply rejected the discourse around All-NBA teams. In so many words, he said that he just wanted to watch basketball, judge players off vibes, and if he’s wrong, so what? He has no interest in participating in the messy All-NBA discussion.
To me, this was the perfect response to everything I’ve said so far. Taking and interpreting his words now, I see this as accepting the impossibility of comparing players who are playing different roles, in different schemes, on different teams, surrounded by different teammates. At first glance, this is a curmudgeonly answer to a perfectly fun exercise. I don’t hold that opinion anymore (yeah, I said it – anymore). This is, instead, a mature dodging of an impossible question that allows him to focus more of his attention on what he can analyze: the game in front of him.
As Scott says at the beginning of his incredible Tatum article, “In order to determine how Tatum can elevate the Celtics through this lens, we must look at what he does well, what he does not do so well, and which play types play to his strengths.” You cannot use TLI, GLI, LEBRON, EPM, RPM, BPM, or any other impact stat to answer the question that Scott set off to solve. You could reasonably construct Tatum’s game from player tracking stats, but doing so would miss the nuance that watching and analyzing the game brings.
The dirty secret of NBA awards is that they don’t take process into account. If you’re over the discourse about All-NBA teams, I suggest taking Scott’s route and grounding yourself in the actual game. If you still want to participate in the conversation, I invite you to interrogate your own conclusions and to borrow one important piece of advice from Scott: if you’re wrong, so what?
Cody Houdek is a writer and podcaster for Premium Hoops where he co-hosts Sense and Scalability. He also assists with videos for the Thinking Basketball YouTube channel. You can find all of his work (articles, videos, and podcasts) here.