Our lexicon for defending in football isn’t good enough

Problem: Why is making more interceptions correlated less well with conceding fewer shots (as pointed to here and here) than making more tackles?

Hypothesis: One can quite easily ‘passively intercept’, by essentially having the ball passed straight at you, through no skill of one’s own; one is much less likely to ‘passively tackle’.

Experiment: Go through all of the Opta collected interceptions in a selection of games and categorise them into ‘active’ and ‘passive’.


This was my starting point. I would look up online (I used WhoScored) the minutes in which interceptions were made and players who made them, and work out what action during the game this was. (This was generally quite easy, though infrequently required going back and re-watching the minute).

I was initially going to watch around 20 games (with a secondary goal of 50), all from the Premier League and aiming to feature every team at least once (and in the secondary goal of 50 at least 3 times). However.

As I was watching and categorising, it became apparent (immediately) that my initial system was flawed. Interceptions are not just ‘Koscielny stepping out of the line and taking three steps to intercept a pass that you didn’t even know was coming’ on one side, and ‘Steven Taylor gets hit in the face as Memphis Depay attempts a pass to Martial’ on the other.

This, from Morgan, is an interception.

As is this (also Morgan).

And this, from Stones.

And this from Dier.


Those are all very different types of ‘interception’, and is the reason why I ended up adding two categories to my short study so that I ended up with:

‘Tackle’ – Interceptions I would classify as a tackle myself (like Dier’s).

‘Passive’ – The initial starting point, interceptions where the ball is essentially passed straight to the interceptor, without them having made any deliberate actions to be in the path of the ball.

‘Anticipatory’ – The classic interception, ones where the player has actively made a deliberate movement into the path of the ball (or made a deliberate action to halt their run to be in the path of the ball).

‘Reactive’ – Kind of in between Anticipatory and Passive. I classed the first Morgan interception I showed above as Reactive, because it seemed more a case of reacting to a tackle that Kanté made. There were also a couple of times where Team X passed the ball badly, yards behind the intended target for example, and a player from Team Y stepped up to take the more or less loose ball.

My categories were certainly not perfect, and even having divided them it was sometimes difficult to decide where an interception should go. I’d also started the study because I thought that not all interceptions were equal (the active and the passive), and that separating the two could help gain some more useful defensive statistics. However, I quickly learned (as I probably should have guessed anyway) that not all Anticipatory interceptions are equal too.


Anyway, it became obvious that this was diverging quite a bit from what I had initially wanted to do, so I stopped after 10 games, in which 367 interceptions were made (although 13 of these I judged to be off camera). I got the following ‘results’ (although I would shift a few arbitrary percentage points from passive to anticipatory to account for some confirmation bias I may have had, trying to prove my initial hypothesis).

Type Count Perc%
Anticipatory 247 69.77%
Passive 56 15.82%
Reactive 18 5.08%
Tackle 33 9.32%
Total 354 100.00%

The Opta definitions, by the way, for tackle and interception are here:

Tackle: “A tackle is defined as where a player connects with the ball in ground challenge where he successfully takes the ball away from the man in possession.”

Interception: “This is where a player intentionally intercepts a pass by moving into the line of the intended ball.”

This isn’t strictly meant as a criticism of Opta. Watching the games and knowing, roughly, what was coming, it was at times still very difficult to work out whether an action should count as a tackle, interception, or something else. And this got me thinking.

Watch this interception by Morgan again.

What is this? I wouldn’t really think of that as an interception. It’s not really a tackle either. He deflects the ball, out for a corner, but he doesn’t exactly take it out of Giroud’s possession.   Dier’s interception, which I classed as a tackle, isn’t what you’d think of as a pure tackle either – really, it’s somewhere in between the two.


Obviously, watching players by eye, these blurred boundaries of classifications aren’t much of an issue – you look for anticipation, awareness, and technical capability, regardless of whether it’s a tackle or interception that a player is making.

For statistics though, it matters, and the stats are based off our general understanding of the game. The football stats used in the mainstream – pop ups in TV broadcasts or at half-time, or in newspapers – are those that we know, or instinctively feel like, matter (shots, goals, possession, corners, cards). As analytics has progressed, and certain areas have been shown to be less important than others, these mainstream ‘broadcast stats’ have changed and will change again in future.

Baseball’s box score was created by Henry Chadwick, who based it off cricketing score cards. This (according to the book Moneyball) later hampered the progression of baseball statistical analysis.

Defensive statistics are still such a mystery in part because judging defenders by eye is still such a mystery. Gabriel Marcotti’s transfer analysis from last summer is a nice summary, and suggests that even big clubs don’t really know what they’re doing in this department. You just have to look at the volatility and uncertainty with which central defenders in particular are regarded. Attacking players are often form-based flavours of the month too, but at least there it’s generally easier to discern who is at what level.


I don’t particularly have an answer though. This is just a comment, which is probably not very helpful. With regards to interceptions, even just through watching ~350 of them, it’s obvious that some are ‘better’ than others, but it would be difficult to separate them subjectively, let alone objectively.

Some, like Stones’ earlier in this article, are ones where the player has clearly anticipated an action and intercepted it fairly far in advance. Some are much more close range, where a player has reacted to the movement of their opposing player going to make a pass, and have stuck out the leg to stop it.

Some players see a sniff of the ball and go in to try and nick it before their opponent can control it much more than others do. Otamendi, for example, seems to have his default setting set to ‘intercept at all opportunities’. While this might help disrupt an opponent’s play, it may also put him at more risk of being caught out of position if he doesn’t judge when to go and when not to go well.

I imagine that there would be similar issues with tackles as I’ve found with interceptions. I mean, you only have to look at the Morgan interception that I’ve shown twice before – I might have been inclined to call that a tackle, if I’d been forced (or it was my job) to call it anything.

At the very least, I hope this prompts some discussion.



One thought on “Our lexicon for defending in football isn’t good enough

  1. Pingback: Top Defensive Action Football Data Analysis - Wednesday 26th April 2017

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s