In his debut blog for OptaPro, data scientist @WillTGM analyses central defender passing trends in the Premier Leag…… 17 Jan

VACANCY - we are recruiting a data scientist, who will work across @OptaPro & @OptaSuit. Opportunity.… 16 Jan

Fantastic to have @WillTGM join the team. We'll be publishing Will's first blog for OptaPro this week, which explor…… 14 Jan

Valuing passes – @GarryGelade considers how passing performance can be evaluated using a Pass Reward model. Vision.… 10 Jan

The OptaPro Blog

RSS Feed

Welcome to the OptaPro blog, featuring news and analysis from OptaPro's cutting-edge research team.

BLOG: Saints due to regress?

Nine games into the current Barclays Premier League season, Southampton have conceded just three goals. Currently leaking goals at a rate of one every three games, the Saints are currently on track to eclipse Chelsea's 2004-2005 Premier League record of conceding only 15 goals in an entire season.

Southampton have an exceptionally strong defence, but is it on a par with the best defence in the Premier League's history? Probably not. Unless we believe that Southampton could theoretically be making a legitimate bid in the Champions League (given the opportunity), we're forced to admit that their defensive record is currently operating at an unsustainable rate. While Southampton's defenders have been worth their weight in gold in Premier League Fantasy Football teams, selecting them has, in all likelihood, become a game of chicken. At some point, they must begin to regress toward the mean.

There are two ways to avoid conceding goals: prevent opportunities, or effectively mitigate opportunities when they inevitably arise.

Southampton have been good at preventing opportunities. They have conceded only 87 shots this season, tied with Tottenham. Only Manchester City has conceded fewer, with 85. This is exceptional, even when considering the Saints' comparatively easy schedule over their first quarter of the Premier League season.

(Click on the graph to enlarge)


*Note: This analysis does not include own goals and penalty kicks.

But, shot volume doesn't tell the whole story. Using OptaPro's shooting model (an updated version of this), which is based on a handful of different measurements including location and pattern of play, we can estimate the probability of a shot resulting in a goal. Southampton, while conceding very few opportunities, concede relatively dangerous ones. However, many of these dangerous opportunities have gone awry, conveniently preserving the narratives surrounding the Southampton defence.


Quantitatively determining how well a team mitigates a high-quality opportunity is a bit more hairy. Looking at the volume and quality of shots that Southampton has conceded, they have conceded 4.6 fewer goals than predicted by the model – currently the largest positive deviation in this year's Premier League (Conversely, it also provides a glimmer of hope for Sunderland, as their opponents' conversion rate seems similarly unsustainable). Deviation from this model can be attributed to two things: un-modelled bias and luck.


One thing this model assumes is that the player attempting each shot will convert his chance at a league-average proficiency. Of the elite teams that Southampton have faced in their first nine matches, it's only included a struggling Manchester United and a Suarez-less Liverpool. This strength-of-schedule imbalance undoubtedly adds some bias in favour of Southampton's defence.

No model is perfect, including this shooting model, so evaluating model residuals should only be done with an adequate understanding of the model's limitations. Sometimes luck outweighs the bias, but other times it may not. The distribution of the effect of un-modelled bias and luck on model variance is unlikely to be constant.

This analysis does not prove that Southampton have been exceptionally lucky, just like it fails to prove that Southampton's defence is as good as Chelsea's during their record breaking 2004-2005 campaign. But, unless Southampton are consistently repelling opposition chances in a way that's so revolutionary that it entirely eludes our shooting model, it's unlikely that Southampton has achieved this outlying defensive efficiency purely on their own accord, and the stage is set for them to start conceding goals at an increased rate.

Posted by Devin Pleuler at 13:54

Related Links

In my experience, there's no such thing as luck... Except in football

Neil Charles breaks down what constitutes 'luck' in football, and which teams are most affected.

Click here to view article

Goal expectation and efficiency

Constantinos Chappas and Colin Trainor take a look at Goal Expectation.

Click here to view article


Colin said...
When calculating expected goals, does this model take into account goalkeeper ability?
October 30, 2013 01:29
Devin Pleuler (author) said...
Colin -- It assumes average league goalkeeping across the model. Teams with exceptional goalkeeping will skew this measure. But, past research has suggested that there isn't a huge effect. There is only a small selection of shots that differentiate the strong shot stoppers from the weaker ones.
November 1, 2013 01:07
Tony said...
Great analysis, David. I am a Saints fan, so I tend to be a little biased about my team. I am also an avid baseball fan, and love the statistically based analysis of that game (I am a Sabrmetric follower). I did wonder just how good the Saints' defence really was, as several times a really good goal-scoring opportunity for opposing teams seemed to end up as either a miss, hit woodwork, or great save. Chance effects inevitably even out over a large sample size. I look forward to the rest of the season.
November 5, 2013 07:02

Post a comment