When someone in the basketball world says “Trust the Process,” I have a visceral reaction. The phrase has become shorthand for the quantitative revolution that has swallowed basketball over the past decade. I have little interest in relitigating how we got here. I’m much more interested in what happens now that the revolution has won: how nuanced ideas became dogma, how models became commandments, and how intellectually lazy people flanderized an entire way of thinking into something that has, in very real ways, hurt The Game. My background in sports is coming up through the Analytics space. I’ve presented research at multiple conferences including the pre-eminent MIT SSAC. But I do not have a degree in statistics that makes me hellbent on spending the rest of my adult life iterating on my YAPM and making the increasingly marginal returns mountains of basketball enlightenment when they are just mole hills. Or maybe a better way of putting it is being incapable of thinking past that mindset or growing from it.
One thing I’ve noticed over the past decade working in sports is how often analytics-oriented people confuse a statistic with an argument. Player X has a better BPM than Player Y, therefore Player X is truly better. But BPM isn’t the argument; it’s evidence. The argument should be: I think Player X is better, and his BPM is one of the pieces of evidence that helps me reach that conclusion.
For whatever reason, I’ve never really done that. I look at all the same stuff, but no matter how technically capable the person I am having a basketball discussion with is, I think the only time I really point to that stuff is after I’ve made my point and we’re now both discussing the merits of my point.
To me, the distinction is whether the model informs your thinking or replaces it. Most people would probably reject the idea that they’re lazily leaning on model outputs, but I think they underestimate how often their opinions are functionally just filtered model results: pull up the relevant players, teams, or games, read what the model says, and repeat it.
Maybe you built the model yourself, and that’s where the sweat equity is. But I can read your model outputs myself. What I need from another analyst is everything that happens after that: interpretation, skepticism, context, competing explanations, and ultimately a judgment. If your contribution to the conversation is simply telling me what your model says, there isn’t much reason for us to pretend we’re having a real dialogue.
That might come off as a little harsh, but is it really wrong? I think what tends to obscure the problem is that statistical models are exceptionally good at some things humans are bad at, while humans are exceptionally good at some things models are bad at. Maybe it’s more useful to think of intelligence as existing across several different dimensions. I plan on writing an article about that next so please follow along so you don’t miss it. The mistake is assuming superiority along one dimension implies superiority across all of them. This isn't chess. The dimensionality is significantly greater.
In general, people who build statistical models to evaluate basketball almost universally acknowledge that those models have limitations. The more interesting problem, in my view, is that people systematically underestimate the importance of those limitations for social, professional, and political reasons. Nobody creates an all-in-one player metric and successfully argues, “There’s nothing wrong with my model.”¹ You’d get laughed out of the room. So everyone acknowledges limitations in principle. The problem is how much weight they actually give those limitations when the model is being used to make a real decision or defend a real opinion.
So why bring this up now? Because this past season, I watched a lot of people hide behind good process in the absence of good results. We had two coaches publicly pull the “well, we won the shot quality battle” card after losing a playoff series³ . And I can tell you for a fact: they weren’t the only teams dealing with that mindset.
Doing this is exactly the lazy, just-look-up-the-model-output approach I described a few paragraphs ago. If a coach genuinely believes that “we won the shot quality battle” is sufficient analysis of why his team should have won a playoff series, then he’s building a pretty compelling case for having his job automated. I don’t need a coach to tell me what the model says. I can look that up myself.²
But I don’t think most coaches who say this actually believe it’s that simple. The reality is that they’re in an incredibly difficult position, being asked to explain an unlikely outcome immediately after it happened. Shot quality gives them a clean and defensible explanation. I understand why they reach for it. What bothers me is when everyone else pretends explanations like that are deeper than they actually are.
We saw another version of this happen with the Jaylen Brown situation. In the aftermath of his departure from Boston, a surprising amount of the public discourse tried to turn the trade into an on-court analytics story⁴ . His on-off numbers weren’t great, because he’s a floor raising high volume guy playing on a team not designed around him. Maybe the Celtics had finally figured out something the rest of us hadn’t?
Except Brad Stevens eventually stood in front of the media and told us, repeatedly, what the actual issue was: optionality. In other words, financial and roster constraints. Brown was going to make nearly $60 million this season, and Boston was staring at a world in which roughly 70 percent of its salary cap would be tied up in Brown and Jayson Tatum. Whatever you think of how the Celtics handled the trade, and I think they badly mismanaged the PR surrounding it, that financial reality is orders of magnitude more important to understanding what happened than Jaylen Brown having a bad on-off⁵.
There were obviously interpersonal dynamics involved too. Reporting after the trade suggested that Brown and Tatum's personal relationship had become virtually nonexistent, and Brown himself clearly wasn't thrilled with how the organization handled the process. But that's precisely the point: these decisions are complicated. They involve money, relationships, organizational hierarchy, roster construction, personalities, and basketball. Reducing all of that to “well, his impact metrics weren't very good” isn't sophisticated analysis, let alone an initial reason to go down this path. It's looking at the model output and working backward toward a story.
I just find it extraordinarily difficult to believe that Boston began this process by opening an impact metric, noticing that Jaylen Brown rated poorly, and deciding it needed to act. Their actions over the previous several years suggest almost exactly the opposite. They believed Brown was an extremely good basketball player. They gave him a supermax contract. What they eventually seem to have concluded was that as good as Brown was, committing roughly 70% of the cap to him and Tatum made constructing the rest of a championship roster extraordinarily difficult under the current CBA.
And maybe there's an uncomfortable interpersonal component to that financial problem as well. Jalen Brunson and Victor Wembanyama both recently accepted less money than they could have pursued in order to preserve flexibility for their teams. Would Brown or Tatum ever have done something similar in Boston? I don't think so, but who knows. But that question is infinitely more interesting to me than pretending Jaylen Brown got traded because somebody discovered his on-off wasn't very good. Would you say the vibes these guys put off are competitive-win-at-all-cost MFers, if they weren't willing to take paycuts?
So that’s where the analytics stuff comes in. It supports their thinking elsewhere, even if it's flawed. And maybe in absence of better PR skills that’s what Stevens’ camp let float out into the world⁶ . I don’t know. But the conversation became so loud that even JB felt compelled to talk about it and correctly point out the success the team has had with him in Tatum’s absence, and the accolades he won even in Tatum’s presence. I worry this might ultimately become a fork-in-the-road moment between nuanced decision-making and algorithmic decision-making.
Again, that’s where everything points to PR mismanagement here. This isn’t my wheelhouse but I imagine the first level of better PR management here would have been working with JB to make him feel more valued internally. I assume that had to have happened, and ultimately failed for whatever reason though.
In absence of that, Boston probably should have realized the moment they signed him to a 5-year, $285 million extension it meant that more likely than not, they were going to have to trade him at some point and all it did was buy them one or two more years with this iteration of the team⁷ . They won the title 11 months later. As Brian Windhorst likes to say I don't think they need to apologize about anything here. They just needed to be more self-aware. Front offices are very insulated, for better or worse, so maybe they lied to themselves and thought Somehow It Would Workout in having both JB and Tatum on max extensions? But how much can you lie to yourself about shrewd drafting and player development to make up for that? It's a pretty flawed premise, but humans love having flaws.
I don’t know for sure. All I have to go on is their public conduct and what they say to the press. That certainly leads me to believe they didn’t see this situation arising, though. I think the general NBA media has dunked on Boston and Brad Stevens enough so I won’t do it again here.
My larger point in all of this is how weaponized “analytics” and “trusting the process” have become. There’s a paradox here: the more widely these ideas are adopted, the smaller the competitive advantage they provide. Eventually everyone is doing roughly the same things. And yet, even as the edge disappears, the language surrounding those ideas becomes more and more pervasive and more authoritative.
Everett Rogers described something relevant to this in Diffusion of Innovations. Innovations don’t spread through a population all at once. They move from innovators and early adopters, through the early and late majorities, and eventually to laggards. Complexity matters too: the harder an idea is to understand, the more difficult it is to communicate and the slower it tends to spread. But I think something else happens along the way. Complex ideas don’t always diffuse intact. They get simplified. Nuance gets stripped away. What began as a complicated framework eventually becomes a heuristic, and then eventually a slogan⁸ .
That’s how I interpret a lot of the analytics discourse from this past season. We’re no longer arguing about whether analytics belong in basketball. That argument ended a long time ago. We’re watching simplified versions of analytical ideas get used as ready-made explanations for complicated events. Jaylen Brown gets traded and suddenly people are pulling up his on-off numbers, as though a complicated organizational decision involving nearly $60 million in salary, the second apron, roster construction, interpersonal dynamics, and organizational hierarchy can be reverse-engineered from a few impact metrics.
And maybe this is giving Boston too much credit, but part of me wonders whether that confusion has actually worked beautifully in its favor, at least for ownership. The Celtics have spent the last two offseasons aggressively creating financial flexibility. In 2025, they dismantled expensive pieces of a championship roster to escape the second apron, obviously due to Jayson Tatum’s injury. A year later, they traded a 29-year-old former Finals MVP who had just made second-team All-NBA. Those are, at minimum, cost-cutting decisions in the literal sense that they materially reduced Boston's long-term financial commitments.
Yet somehow the public argument has frequently become about whether Brad Stevens and Joe Mazzulla are too beholden to analytics rather than whether Celtics ownership simply decided there was a limit to what it was willing to spend. Compare that with someone like Tom Dundon, whose cost-cutting in Portland has become part of his public identity. Bill Chisholm largely hasn't received the same label.
Maybe that’s intentional. Maybe Stevens got left holding the bag. Maybe everyone involved is perfectly comfortable with the arrangement. I have no idea. But if you were trying to cut costs without developing a reputation as the cheap new owner who broke up a championship core, having everyone argue about Jaylen Brown’s on-off numbers instead would be a pretty convenient outcome.
To bring this full circle: analytics is no longer some emerging counterculture that needs bold innovators to pave the way while Ivy Leaguers fight to prove they belong running basketball teams instead of former players. The quants won. They took over. It’s time to dispense with the idea that they’re still the outsiders⁹. We are now on to the era where it's being weaponized politically.
Something I think about a lot is how former players used to denounce the nerds and their spreadsheets, skeptical of what any of it could possibly contribute to a front office. At the time, I viewed a lot of those arguments as political and I still think many of them probably were. But in 2026, when those same arguments are made, I have a much harder time seeing them as having significant merit in addition to be political. Analytics has already won its seat at the table. Dismissing it wholesale now is political in much the same way that coaches and GMs hiding behind the numbers when they come up short is political. The weapon changes depending on who has the institutional power.
All of this comes down to something simple: don't hide behind the numbers, and don't hide behind your model. I can’t tell you how many times I advocated for changing something because the results weren’t good enough, only to get pushback that the process was sound. Some of that may have been on me and how I communicated. But when you work for a basketball team, you are ultimately in the results business, and I was disappointed by how often I saw that reality discounted around the league¹¹ .
Good process is incredibly important. It protects you from panicking over variance and making stupid decisions because something went wrong three times in a row. But “the process was good” cannot always become an excuse for inaction either. When a good process repeatedly produces bad results, your response should be curiosity: What are we missing? What assumptions might be wrong? Is there something the model doesn’t see? Is there something we can change without abandoning everything we believe?
There is obviously a fine line here. Reacting recklessly to bad results produces exactly the kind of decisions that become podcast content for the next decade. Defaulting toward process is better than that. But if your goal is to operate in the top 1% of the 1% of competitive basketball, simply having a good process isn’t enough. You have to care deeply about whether it actually works for you.
An analogy from outside basketball: if you're trying to be the best discretionary hedge fund manager in the world¹² , there is no way your entire job consists of reading statistical model outputs and simply doing whatever they tell you to do. You’re not managing an index fund. The models are inputs. Your job is to make judgment calls given those inputs, and over time the good managers make more of the right calls than the bad ones do. If you can’t explain where that additional value comes from, it’s probably worth spending some time thinking about that.
In both jobs, the person making the call can attribute good outcomes to skill and bad ones to variance, and nobody can prove otherwise. But there's a second incentive at work: consensus is safe. Departing from it creates the possibility of being spectacularly right, but it also creates the possibility of being conspicuously wrong. A lot of people prefer this. Instead of thinking harder about a difficult problem and accepting the risk that comes with reaching their own conclusion, they choose the professionally safer path of landing somewhere in the middle and relying on consensus thinking. The moral hazard Danny Leroux always talks about with GM's starts to apply here and has started seeping through whole organizations.
Statisticians are taught in college and early in their careers not to put their thumb on the scale, and in my experience they tend to take that lesson very seriously. That is good advice when you're building a model, which is an estimation of something: be ruthlessly committed to producing the most honest estimate you can. But being generally correct is not enough when you’re trying to beat 29 other NBA teams. Once the model has produced its answer, your job isn’t complete. Now you have to decide what to do with it. Context matters. Judgment matters. Basketball knowledge matters. Knowing when the model might be missing something matters.
So when I’m having a basketball conversation and someone begins their argument by telling me what a model says, I’ve learned to recalibrate the expected utility of the conversation accordingly. Tell me what the model says if I haven’t seen it. Then tell me what you think. That second part is why I’m talking to you instead of looking up the model myself.
If you made it this far, please take a moment to follow me and share this article with someone.
¹ Jeremias Englemann once said that to us at ESPN when he was talking to me and Royce Webb about continuing to run his real plus-minus model once he left for the Mavericks. We never really figured out how to run his R scripts that scraped bref and weren’t designed to really run more than once a month and starting only two months into the season. Maybe I’ll blog about the history of RPM and RPM 2.0 in the future. Let me know if you want that.
² To be clear this isn’t what they are doing. This is PR, and I think they maybe didn’t excel here in phrasing things in way that would go over well.
³ The other one I’m referring to is Joe Mazzulla, who has a little bit of a rep for being to process oriented in shot selection. I don’t think he had any super bad soundbites like Kenny, though.
⁴ This is like, Bobby Marks fault, and he seems remorseful about it, but I 100% believe him when he said he was just relaying to him what an NBA staffer observed. Like I said, the process maximization line of thinking is way more pervasive in the league than people seem to acknowledge. ⁵ And yeah I know, Paul George's deal isn't much better value proposition-wise, but it is somewhat better. I'm pointing out their line of thinking here and how it got misconstrued, not pointing out that they executed on that well.
⁶ Given the timing of these events and what happened afterwards, regardless of whatever Brad’s true intentions were, there’s no indications he executed on them perfectly.
⁷ This might be a future blog post idea of me. Let me know if you want to see it. But I remember when Larry Coon went over the new “supermax” in a Sloan panel about 8 years ago and he noted to some effect that these contracts had was they almost all got traded. Ever since then it’s made me mentally estimate how many more years a player will play for their team the moment they sign one of these deals.
⁸ This also makes me think of Goodhart’s Law and how earlier in the Basketball Analytics Revolution about a decade ago you saw teams just chucking threes for three’s sake. Specifically low major D1 teams in buy games against high majors.
⁹ I’m already mentally on to the Do We Need To Stop Talking About Talking About Analytics Won mindset, personally.
¹⁰ But they didn’t know it, per se. Like they consciously knew something was wrong, but it stayed in their subconscious that this was a political attack on the game.
¹¹ To be clear this sentiment extends beyond my time working in NBA front offices.
¹² The other job Ivy Leaguers love to have