Creating new data by stealing from basketball

An origin tale

A couple of weeks ago, I read an article on FiveThirtyEight. It was about a bit of basketball tracking kit, the Noahlytics Data Program, which gives you data on the angle of shots and where they go through the rim. (I only partly mention this because the name is a Biblical pun: Noah's arc -- although it doesn't insist players take things two by two).

The nifty thing about the program, according to the article, is that players can get good, concise feedback on why their shots aren't going in, and quickly too. There's an anecdote in the piece, which is by Ben Dowsett, where then-Toronto Raptors player Kyle Lowry was going through a cold shooting patch. The Noahlytics data showed that the arc of his shots was at 41 degrees, whereas when he was shooting well it was around 46 or 47.

From there, I assume basketball coaching takes over, whatever basketball coaching looks like, but I was intrigued by the thought of how this type of thing could be transferred to football.

'Player motion data' -- a term I will coin here -- is not exactly new. The video cameras that capture positional tracking data (think Football Manager dots for every player on the pitch) can also be used to capture players' body orientation. The software needs to be more advanced, and the work to use the data is more difficult, but it's been applied to the NBA for a number of years.

Body pose is damn sophisticated though, and the relative simplicity of Noahlytics struck me as something that might be a more easily applicable concept to transfer. You don't need to look at a computer-generated ragdoll or a bunch of ankle-knee-hip angle readouts; just a few numbers about a repeatable skill.

But, the question is, what is a worthwhile repeatable skill in football to apply this to?

For regular questions like this, subscribe to the newsletter

Oh football, you bringer of despair

The problem with football as a sport (well, one of them) is that shots don't happen very much. On average, any given team in any given match will take about 11-12 shots, give or take a couple. In basketball, the average is almost ten times that. Teams take around 30-40 three-point attempts and 45-60 two-point attempts (let's not count free throws for now).

The flip side of this is that each shot represents a tiny amount of the team's eventual scoring plays, but the fact remains that if you try improving a player's shooting, you'll see results pretty quickly. This is even moreso the case when you consider that shots in basketball generally look pretty similar compared to the mechanical variety in football shooting. For a start, NBA players will never be taking a shot with their head.

According to FBref, teams make around 40-60 passes in the final third per game, a comparable number to three-point attempts, which is nice. These, I think, are completed pass numbers, so the amount of attempts will be a little higher. We have an action here that has comparable frequency to basketball field goals and a not-insignificant impact on the eventual total of a team's scoring. We might even be able to expand this to passes in the opposition half if we wanted a larger sample.

But what would you measure? Unlike in basketball, there's no fixed target or repeatable goal. You can't consider the intended recipient themselves to be a replacement for 'the hoop' in basketball, because what about through-balls? What about when you want to set the teammate up to receive on the turn?

Unlike for shots, where hitting the ball cleanly does seem to be a problem, I can't think of many 'mishit' final third passes. There are also a lot of different techniques involved depending on the type of pass: long balls, chipped passes, crosses, more regular passes too.

An actionable idea?

Let's think about this a little differently. What about if, instead of pass technique we were to look at first touch instead? The reference points could be the player's own body and the nearest defender (within a given radius). Could you identify when a player's first touch puts the ball too close to the centre of their body, not on either foot? Or when it puts the ball practically underneath them, or too far away from them, making it tough to make a next action?

The information about the nearest defender would be necessary to add context of course. A player's first touch might put the ball overly close to them, but that might be because a defender was right on their back. However, maybe in those circumstances you'd be able to say "hey, we know you're receiving under pressure, but if you put your first touch a little further away from you and onto your stronger foot a little more then you'd be able to offload it quicker".

This would also potentially fill a gap in (most of the) existing data provision in football. There's a wealth of data on shots, a just-about-managing of data for passes, but there's almost nothing on how players receive the ball. But what does every player do before they make a shot or pass (barring first-time efforts I suppose)? Take a touch!

Oh football, you bringer of despair (reprise)

This is, as much as anything, a lesson in the difficulty in trying to 'take inspiration' from other sports. Although basketball and football both involve two teams, a ball, the concepts of shots and passes (with no restrictions on where passes can go), they're still different. The way that the games are played means that, while concepts might be similar on the surface, they way they play out in practice can alter things hugely.

Actually, hold on.

The way things 'are' played?

Let's take a step back to where I -- stuck in my sense of the present day -- said that football teams have about ten times fewer shots per game than basketball teams. There are a number of reasons for this: the presence of a goalkeeper; the different dynamics behind getting within shooting range; the fact that basketball hoops are above players, meaning shots have a free path, once they're in the air, to their target.

But another is simply that basketball players use their hands and footballers use their feet (or heads). The difference between humans and other animals is (among other things) our opposable thumbs, not really anything particular about our feet. As far as ball sports go, basketball, rugby, NFL -- they're playing on easy mode. Football -- where players, as Brian Phillips wrote, are forced to 'compete in a state of artificial clumsiness' -- is where hard mode gets turned on.

Considering they are literally paid millions to do this, elite footballers miskick the ball a surprising amount, especially when shooting. Sometimes they plain just fall over in the motion of failing to hit it.

But imagine if, perhaps using something like the Noahlytics program, they were trained in striking the ball cleanly. I'm not expecting perfection, but imagine if footballers, across the board, got tangibly, markedly better at getting a good shooting motion. Would teams still be taking 12 shots a game?

Players getting better at shooting threes in basketball probably isn't the main driver of why three-point attempts continue to rise, but it surely is a driver (at least compared to when they were first introduced into the NBA in 1979). If footballers got 30% better at shots from further than the penalty spot, would they make more of them? I bet they would.


So, maybe you can take an idea from basketball and apply it to football, but needing to change the area you're applying it to. Or maybe you can apply it to the same area, and you might change the sport as a result.