The Theory of Everything (in football)

There was a point in time where people didn't write books on story structure. Nobody knew what an 'act' was, or what pathos was, or that a punchline goes at the end of a list of three. They just told stories, and a lot of those stories will have been pretty terrible. The stuff that has survived to the present day, about mythical figures like Zeus and Hades and Margaret Thatcher, are just a synthesis of the best yarns.

But people started to work out that there were things that were pretty universal to good stories, certain themes and rhythms that audiences just loved. And that helped make the chance of you hearing a decent story much better.

There are times when I think that football and football analytics are still in that campfire phase of storytelling. We've hit on some great stuff but there's still a lot of groping around in hope. You can work out a lot by putting data through a linear regression model, but I want to try and come up with some theory to hang it on.


Part I: Attack vs Defence

Football is a simple sport. You want to score more goals than your opponent.

We can only score goals when we have the ball (barring freak own goals) and we can only concede when our opponents have the ball. This means we tend to think of things as 'attack' and 'defence', but another way of giving these things names could be 'goal scoring' and 'goal preventing'.

On the surface, both of these things are easy to measure: scoring goals means you're doing well at goal scoring; not conceding goals means you're doing well at goal preventing.

But we know that goal scoring is not just done by the players doing the shooting. People like Kevin de Bruyne and Fran Kirby are very important in their teams scoring goals, but they don't often do the scoring themselves. They're called chance creators. And sometimes they're involved in what's usually called build-up.[1] The 'goal scoring' part of football has several fairly distinct concepts.

This is the part where storytelling theory would probably say I should tell a joke to keep you engaged through this exposition-heavy first act.

Why did the chicken cross the road in 68% of simulations? I don't know, you'd have to ask the video analysts.

This is the point at which me spending time telling you things you already know becomes worthwhile. Because these divisions, these ways of splitting up 'goal scoring': there's no such division that people make for the 'goal preventing' side, not in the regular public discourse at least.[2] Isn't that a bit weird?

Defending is just 'defending'. We talk about pressing a bit, but that's a method rather than a theoretical unit, more of an equivalent of through-balls rather than the concept of chance-creating being a distinct part of 'goal scoring'.

So, let's split 'goal preventing' up. There may be better ways of doing it what I'm about to propose, but I'm going to put forward 'block preventing' and 'direct preventing'.

'Block preventing' is a method of goal preventing where teams try to keep their opponents at bay. It's partly done through not allowing good passing options, but also partly through instilling fear and caution. Teams don't pass in U-shapes because they think it's the best way of scoring, or even because they're under direct threat of losing the ball -- they do it because they fear that they might lose the ball if they pass more directly. This is what good 'block preventing' does.

Meanwhile, 'direct preventing' is the kind of 'goal preventing' that is, well, more direct. If everything goes to plan, direct preventing is where you win the ball back; block preventing is where the final whistle does.

Dividing goal prevention in this way subtly changes our idea of how to quantify defenders. Just like how a player can be a primarily a chance-creator instead of a shot-striker, maybe a defender isn't a direct preventer, but could be a good block preventer. Ultimately, both things are part of the same aim -- not conceding goals -- but they are roles tied to distinct aspects of the game.

Public discourse has kind of started to do this already, albeit without a theoretical framework behind it. It's now fairly common to note that a defender's high number of tackles and interceptions means they're 'busy', rather than saying it makes them 'good'. But maybe they are good. Strikers who take lots of shots are 'busy' too, but many of them are good.

The thing that we recognise by differentiating 'busy' from 'good' is that there are different facets to goal preventing. But maybe defining and theorising about these different facets will help lead us in useful directions...


Part II: Hey, it's not my fault

The thing about assessing forwards statistically is that it's really easy. They do, like, one thing. They shoot, and sometimes they score. Expected goals models help work through some of the noise of coming up against good goalkeepers or bad variance for a stretch of time. But ultimately they kick the ball and someone makes a note of that and everything is done and dusted.

Except sometimes, a striker makes a run and their teammate on the ball doesn't see it and doesn't play the pass. Or the cross is slightly overhit. Or some other bizarre event that would be a good punchline to a list of three.

We don't tend to worry about these instances too much because it seems that teammates spotting runs happens frequently enough to give a signal, but the principle is always in the background. Forwards -- or shot-strikers to hark back to the theory in Part I -- can (usually) only strike shots if the chance-creators find them. And chance-creators can often only chance-create if reasonable capacity has been built up around them.

Similarly, a lot of 'direct goal preventing' only happens if defending elsewhere on the pitch has failed. This might either be 'block preventing' failing, where the team's structure has been punctured by the opponent, or it might be that a central midfielder made a mess of a tackle attempt, and now a centre-back has to step out.

This is one thing that makes statistical analysis difficult. Is a defender making a very small amount of tackles because they're a poor direct preventer, or because the rest of their team is so good? Is a striker taking a lot of shots really fantastic, or are they just a slightly-above-average cog in a fantastic attacking machine?

These questions about opportunity aren't unique to defending, and may not even necessarily be more significant for goal preventing than goal scoring. It's just that we haven't found good stats yet, so we think about these problems more.


Part III: I contain multitudes

As I write this, a certain type of Marvel Cinematic Universe fan is annoyed that Spider-Man: No Way Home isn't in Academy Award contention. The merits, or lack thereof, of this Spidey-for-Oscars campaign aside, it's a topical reminder that there are different ways for films, or stories, to be 'good'.

Not only are there different ways for stories to be good, there are a ton of different methods that they go about achieving this. Now, in football what is 'good' is a little less subjective, but there are still some different methods of going about being it.

A player who has the quality of 'good at scoring goals' has a whole set of component skills that they possess, and each of those players has a different set of them. So will each 'good chance creator' player. Technique, perception, anticipation, mentality, fitness, agility -- no two forwards will be exactly the same as each other on all of them. And that's before we break down these skills down into sub-skills, like different techniques for different types of shots or passes.

Similarly, a good 'direct goal preventer' will have different skills. A good 'block goal preventer' will have different skills. They too will need to have good technique, perception, anticipation, mentality, fitness, agility. They too will have to be a good manipulator, or manager, of space.

When we talk about defending statistics, we're far more like to go to the specific skills. Like in Part II, where I talked about the issues of opportunity, the lack of a decent overarching metric has forced people to drill further down. But these questions exist for attackers too.

I say this because there's a temptation, when trying to statistically analyse defenders, to go down a rabbit hole that focuses in on individual skills. Because we (currently) have no overarching stat that flashes a big light saying 'good defender!', the temptation is to get more specific.

What I mean is, a tracking data model of how defenders manage space wouldn't be the equivalent of xG, it would be the equivalent of a tracking data-based model of how attackers exploit space. Either of those tracking data models would be incredibly useful, but also incredibly difficult to produce. Expected goals came quite easily, because of a quirk of both the sport and the data that was already being collected.

Splitting up 'defending' from a theoretical standpoint might help. There are different ways that a player can be good at 'direct preventing', but we probably have stats that fit those jobs (e.g. tackles, interceptions).

'Block preventing' is the harder part, and where we're more likely to try and find the different skills that makes a player a good block preventer. Anticipation, perception, agility, space-management, decision-making. If we had stats that seemed useful for assessing 'block preventing', very few people would be considering these.


Part IV: A New Hope

Where does this all leave us?

I've proposed thinking of football in two modes, 'goal scoring' and 'goal preventing'. This is a bit simplistic, and I'm aware of that, but as a starting point it's easy to understand. I proposed splitting 'goal scoring' into shot-striking, chance-creating, and capacity-building; and 'goal preventing' into block preventing and direct preventing.

I've talked about how shot-striking often involves chance-creating, and direct-preventing often involves a breakdown in block-preventing. Things are less linear than that, but it's worth keeping in mind that these types of dependencies happen on both sides of the ball.

And I've talked about how each of these theoretical units, like chance-creating or direct-preventing, can be achieved through different methods, and each of these methods can be split into individual skills.

I think that what I would do next with this is to think about what block-preventing and direct-preventing entail. I'd also look at existing statistics and consider what extent they are useful for assessing block-preventing and direct-preventing. I would also try to think about which players are doing which type of goal preventing while watching some matches. And, of course, wonder about whether this theoretical framework is correct or useful at all.

I may do this in a future post, but this one has run its course.


A man went to see his doctor.

"Doctor," he said, "I'm having terrible trouble. Life is cruel and confusing. I'm losing all hope. Football, a sport I loved, seems both so primitive to be without meaning and yet so impossibly intricate I'll never understand it. What am I to do?"

"Ah, treatment is simple," the doctor says, "There's this great newsletter I know, Get Goalside, that'll help you figure it out."

"But doctor," the man said, bursting into tears, "I just finished reading one!"


[1] "[...]what's usually called build up." || I'm not a huge fan of that term, I think it implies that if you do enough building you progress a level to chance creating zone, which I don't think is really true. A more accurate way of describing it is often something like 'we have the ball and we want to score but we also very much don't want to concede the ball here'. Friend of the newsletter Tiotal Football has talked about this part of football as building 'capacity' for an attack, which is a phrase I like too.

[2] "[...]there's no such division that people make for the 'goal preventing' side[...]" || I'm aware that coaches often talk about 'defence' and 'transition defence', as opposites to 'attack' and 'transition attack' -- that's useful for coaching but I don't think it's a hugely useful framework for building a theort about what football is, mainly because it leaves the non-transition aspects too large. This is also a good point to mention why I haven't talked about football as "in possession" vs "out of possession". Again, I think that's useful for coaching but not for football theory - your goal is not to have the ball or not, it's to score or prevent scoring; what you do in and out of possession is just means to those ends.