Monday, 5 May 2014

Defensive Errors in Football

Mistakes are a massive part of football. You only have to think back to Kompany's skewed clearance for Coutinho's goal or Gerrard's slip against Chelsea to realise just how important a single error can be to a game, or even final league positions. They can be a big spanner in the works for anybody looking at the stats too, a player might make 29 successful passes on the pitch but the 30th is misplaced straight to the opposing forward - oops... goal. In this post I'm going to go through a few things I've found looking at defensive errors. 

“Football is a game of mistakes. Whoever makes the fewest mistakes wins.”
Johan Cruyff

Error = Big Chance
Since the start of the 2011/12 season there have been 1265 defensive errors that then led to a shot on goal and 491 goals scored. This is a 39% conversion rate, an equivalent rate to a 'Big Chance' or 'Clear Cut Chance' in Opta parlance, that is "a situation where a player should reasonably be expected to score usually in a one-on-one scenario or from very close range".  (see Opta's definitions). In other words, you make a defensive error and let the opposition in on goal, they are going to get a good chance to score. 

The chart below is a plot of errors leading to shots vs. errors leading to goals.  It's not a great R2 value but you can see the trend is definitely there. Each point on the chart represents a individual season for a team. 

This Season
The next chart below is just for this season. Spurs have made a lot of errors and been punished severely! This is one of the biggest outliers in this data. Liverpool have made the most defensive errors that led to a shot on goal of any team and conceded around the league average number of goals from them. City have conceded a similar number of goals from defensive errors but from half the shots. Strong defensive teams such as Chelsea, Hull and Crystal Palace have built their season on rarely making costly mistakes.

Defensive errors don't look repeatable, for most teams. The chart below shows the plot of each team's 'errors leading to shots' from one season to the next for the clubs that have been in the Premier League for each of the past 3 seasons. Overall, there's no trend to speak of, however, there are some patterns that stick out. It's also worth noting that errors (for either shots or goals) do not correlate with 'regular' shots/goals conceded, i.e. teams who concede a lot of shots don't necessarily make a lot of errors.
Arsenal are shocking, amongst the most error prone of teams each season. They're the only team that has a consistently poor trend with regards to errors and their goals conceded each year  tracks this. You can also see how varied most teams are from one season to the next. Some teams have remarkable seasons. Swansea's 'sterile possession' and defensive strength under Rogers in 2011/12 stand outs out on the plus side.  Ditto WBA under Hodgson in the same year. Last season was the worst of the three here for errors leading to both shots and goals and Newcastle led the calamity. 

15% of all Goals from Errors
In the past season's there have been around 160 goals as a result of an error, which is about 15% of the total goals scored. This is a fair chunk, and not to be sniffed at. However, it's not the be all and end all of a team's defensive performance, far from it. Liverpool had an excellent defence and the most clean sheets last season (43 GA, 16 CS) despite conceding a lot of errors (leading to 10 goals). When and in what context these errors occur is clearly important. They are perhaps particularly telling for clubs battling relegation. Last season Wigan conceded 17 goals (the most ) from 36 shots from errors (the most). It was their undoing. They were pretty sound defensively otherwise.

Regarding Other Defensive Stats
As mentioned at the start, errors like these can really throw a spanner in the works of any analyst looking at a team's or player's defensive performance. One loss of possession or misplaced pass can quickly undo a whole game's worth of stoic defending. I spent almost a full month of evenings a few summers ago trying to 'unlock' the secret of clean sheets, looking at every defensive stat I could, including time in/out of possession, opponent pass accuracy, final third passes conceded, shots conceded, lots. 

Whilst all of these metrics pointed roughly in the right direction there was always a few teams facing completely the wrong way, similar to what StatsBomb found themselves this week looking at opponent pass accuracy. That errors are not generally repeatable means as an analyst we can ignore them as 'real' to some degree, but we'll have to accept that they are a permanent source of noise.

Further work on errors would involve reversing the data to see if certain teams force a lot of errors which then lead to shots or goals (e.g. gegenpressing). It also be good to take a closer look at some of the stand-out data points from this and recent seasons: Why do Arsenal conceded so many? What did Newcastle do wrong last year? What did Rogers do so well at Swansea? For now though, one more thought from Johan...

“Actually I never make a mistake, because it takes a huge effort for me to be wrong.”
Johan Cruyff


  1. Good stuff!

    Any thoughts on why the 2011/12 numbers seem so much less than the more recent years (Arsenal apart!)?
    It seems to have run at about 2/3 2013/14 and just over half 2012/13.
    Maybe changes in data collection methods?

    1. also... just seen you're blog. great stuff, will be sure to follow.

  2. No idea Big Ripple, I noticed that too. I'd have to say my 'instinct' is that it's real but we would need more than 3 sample years to go from to say either way.