Again, we’re in free preview October! If you want to sign up ahead of November, please do so:
And if you like what you read here, feel free to:
On using all the tools at your disposal
In honour of the return of the NBA this week, I thought I would focus on a nice interview with Seth Partnow on the prospects of the Toronto Raptors this season, to get at the kind of analysis it would be really nice to see a little more often in the soccersphere.
First, I should say in advance. I have been a Partnow guy for a while now. Partnow is most recently of the Milwaukee Bucks and, like a lot of the better basketball analytics people, he is uber-detailed without being pedantic, clear to the layman but authoritative. One of the Partnow pieces that has stuck with me most over the years is on Godhart’s Law: “When a measure becomes a target, it ceases to be a good measure.”
I don’t know if we yet have an example of this in football, a team that has so bought into the principles of xG that they have solely focused on only taking shots from high xG areas. But Partnow’s point is that you can’t, and shouldn’t, coach to the numbers.
But that’s not what I want to talk about today.
Rather, it’s Partnow’s style of analysis that I find the most helpful. This is in relation to a newsletter I wrote a little while back, on the fact that we’re in a moment right now where there is a lot of normative analysis in public football analytics (“Is this player/team, in fact, good?”) and less prescriptive analysis (“Team X should consider doing X if they want to avoid Y”).
Partnow spoke with Blake Murphy for the Athletic to preview the Raptors’ chances this season (they have both been hoovered up into the Athletic’s good sports journalist vacuum cleaner).
What strikes one first is the palate of tools available to analyze a particular player, and then sensible caveats that come with those tools, and then comparing two fairly predictive measures to come to a reasonable conclusion.
In the 33-year-old Kyle Lowry’s case, the first concern is how his age might affect his production. Here, Partnow first notes that the age curve is an average that smooshes the reality that no two individual players age alike. As Partnow says:
Individual players are much more subject to more step-wise movement: a player peaks, falls off a little bit at 29, stays relatively stable til 31, drops again to a lower level for his age 32 and 33 seasons, then his production falls off a cliff at 34 and he joins a studio show. '
From there, Partnow goes into a range of factors traditionally tied with age-related production issues, starting with offense and moving to defense, citing each measure as he goes—even naming their progenitors!
From there, Partnow goes into a succinct but information-rich discussion of the effects of Leonard Kawhi and Danny Green’s departure to the LA Clippers and LA Lakers, respectively.
Partnow: Aside from the names you mentioned, there aren’t really any obvious candidates for big usage bumps. Leonard was Toronto’s best shot creator in 2018-19, and replacing him is complicated by the fact that other than a few Jeremy Lin cameos, the roster didn’t feature a huge amount of additional creativity:
The chart above compares the proportion of each players’ attempts that were “self-created” after the player had possessed the ball for more than two seconds as well as their efficiency on those shots, compared to league averages. As I discussed when I explored Oklahoma City’s new emphasis on ball movement, this touch time cut off is a reasonably good proxy for dividing shots that come from one-on-one play as opposed to off-ball movement.
By this analysis, Norm Powell or Fred VanVleet might be able to make up some of the gap in shot creation while maintaining a degree of efficiency, but only to an extent. Other than that, the Raptors will have to find some ways to allow dependent scorers like Anunoby or Serge Ibaka to get more shots set up for them.
What I like most here is how effortlessly and effectively Partnow switches between player and team in diagnosing the potential problems.
And the best part of all this is that the conclusions are more or less actionable. It’s one thing to say that having two of your best players (well, maybe not Green) leave your team will have a negative effect; it’s another to try to measure in precise terms the vacuum left in production and what it might theoretically take to fill it.
“But basketball is more data-rich than soccer!”
I know.
But I’m not certain, and I’m not convinced, that soccer is less metric-rich than basketball. I think, right now, for various reasons, there is too much of the “these are my metrics and I’m going to use ‘em!” variety of soccer analytics writing, and less of the sole practitioner, “Let’s use all the available tools to discuss in plain terms what is wrong with a specific club or player” kind.
Maybe it’s because most soccer metrics, even conventional ones like xG and their myriad offspring, are self-generated according to many individual ‘secret spices.’
At the risk of repeating myself for the millionth time (I used to make this point a lot), if you’re an analyst looking for credibility and getting hired, make sure you more valuable than your metrics. A holistic approach, one that reveals plaintive gaps in team or player performance, that can switch from the granular (individual players) to the general (front office issues) in a seamless, intuitive way, would go a long way.
Where is football’s Partnow? Or its Zach Lowe? I think the likes of Michael Caley, Mike Goodman and even Knutson in his day are great examples, but I still think the we’re missing the big five tools guy who could do this sort of thing on, I dunno, the Athletic for a while, before they got hired out for good.