Wednesday, November 30, 2011

The Premier League's Best Defenses of the Last Decade

As a second installment in the "best of the decade" series, here are the decade's best defensive teams in the English Premier League as measured by goals conceded per match. By far and away, Chelsea's 04/05 squad conceded the fewest goals per match, at a rate of less than .4. In contrast, most of the decade's best defensive teams conceded around .6 goals per match and about half of what the average team concedes - still not bad, and good enough to land on the Top 20 list.


Noticeably, the list is entirely composed of squads fielded by Arsenal, Chelsea, Man United, and Liverpool. But something else pops out: given that we are looking at 20 teams from 10 seasons, does the list include each season's champions? As a handy way to assess this, I color coded the above chart to indicate the champions in solid orange, the runners-up in light orange, and teams that finished elsewhere in the table in blue. Turns out, the answer is not exactly, but close. 7 of the 10 championship teams during the decade also had the league's best defensive performance (as measured here). And 6 other teams in the Top 20 were runners-up.

The reason we have some teams in the Top 20 that were neither champions or runners-up has to do with the fact that there are some seasons when the top teams were particularly stingy when it came to goals conceded. So, for example and very unusually, all four clubs made the Top 20 list for the 07/08 season. And interestingly, the 06/07 Chelsea team had a better defensive record than eventual champions Manchester United

Settle any arguments yet? Probably not, so stay tuned for further explorations of the decade's Top 20.

Tuesday, November 29, 2011

The Best Premier League Offenses Of the Last Decade

Each Premier League season 20 clubs field teams to contest the league. So over the course of a decade, 200 teams (20 clubs * 10 seasons) have competed for the championship in England's top tier of football. One things fans love to debate is who's been "best"? And what does it mean for a team to be "best"? Did the Chelsea squad that won the league two years ago outperform last year's or the previous year's Manchester United team?

In a first installment geared toward settling some of these debates (and perhaps generating new ones), here are the decade's 20 best offensive teams, as measured by goals per match.


Offensively speaking, the 09/10 Chelsea squad was a juggernaut, and an order of magnitude more productive than the rest of the 20 best teams during the last decade at about 2.7 goals per match. They clearly were outliers, as was the entire season when it came to goal production.

The rest of the Top 20 produced goals at a rate of between (roughly) 1.9 and 2.3 per match. But it's also notable how Man United and Arsenal had squads that managed to produce goals year after year. The only non-Arsenal, Chelsea, and United squads in the decade's Top 20 were the squads representing Liverpool in 08/09, Newcastle in 01/02, and Man City in 09/10. Talk about offensive dominance by the Gunners, Blues, and Red Devils.

Friday, November 25, 2011

Moneyball Special On BBC Radio 5 Live


Last night, BBC Radio 5 live broadcast an interesting discussion about playing "Moneyball" in football and other popular UK sports. The discussion provides a nice overview of the issue, along with interviews of Billy Beane, Damien Comolli, Steve McLaren, and others. Here's the synopsis from 5 live:

"On the eve of the UK release of the film Moneyball, based on the influential book, 5 live Sport takes a special look at the growing influence of statistics in sport. In the United States, The Oakland As' general manager Billy Beane, through the unsentimental use of statistics, exploited inefficiencies in the market for baseball talent and built a low-budget team that triumphed over lavishly-funded opposition. Colin Murray is joined by reporter Jonathan Legard, Phil Clarke director at the Sports Office, Gavin Fleig performance analyst at Manchester City and Professor Bill Gerard, performance consultant at Saracens to discuss whether the methods of Beane can be as effective within UK sport."

You can directly access and download a podcast of the program here.

Monday, November 21, 2011

Roberto Martinez, Insurgent Leader, Or: Why Wigan Played Guerrilla Football Last Year


Roberto Martinez plays guerrilla football. Sure, he seems a decent guy – mild-mannered and well-behaved on the sidelines – but underneath beats the heart of Che Guevera. The Latics have been in trouble every season they have played in the Premier League. Yet, once more, Martinez’s Latics narrowly escaped relegation last year, returning to the DW Stadium to play another day.

This is nothing short of miraculous, say some. The Swiss Ramble, a blog on football finances, put Wigan’s and Martinez’s challenges and achievements like this: “Given their substantial financial disadvantages, Wigan’s ability to survive in the Premier League is a minor modern miracle.”

Divine intervention is difficult to quantify, and so it is hard to know what role it plays in Wigan’s outcomes. A look at how Martinez’s club scored goals last season is easier, and reveals a particular – and some might say peculiar – style of play.

How peculiar? Consider this. For the Premier League as a whole, on average, teams scored 1.4 goals per match last year. More than 70% of these came from open play. That means teams scored roughly 1 goal per team per match this way. In contrast, a scant 6.3% goals came from free kicks. Here’s another way to think about that free kick number – teams needed to play 28 matches before once finding the net this way. If you’re counting on free kicks, good luck.

But Wigan under Martinez isn’t your average club. Data from Opta Sports show that fewer than half of Wigan’s goals (45%) last season came from open play. Statistically, that is far less than one goal per game. In fact, Wigan relied so little on open play goals, they went an astonishing 21 matches without scoring a single goal from anything resembling a patient build-up last year. Wigan supporters, instead, were treated to an exceptional display of fast breaks and free kicks. Compared to your average team, Wigan scored twice as many goals from fast break opportunities. And here’s the kicker: they also managed almost four times as many goals from free kicks as other teams (3.7 is the actual multiple).

These numbers are too odd to be coincidental. It seems Robert Martinez’s men had a simple strategy, but one that resembled no one else’s: they laid in wait for their opponents to lose the ball one too many times and then rallied the troops to punish them on the counter attack.

And they must have been practicing those accurate free kicks every chance they had, while corners did not seem to be part of the practice regimen (or they just didn’t bother to try too hard and move upfield on corners because they didn’t want to be caught on the counter themselves).

All sounds like a bit like insurgent warfare, doesn’t it? Lie in wait, and use your sharpshooters when given the chance.

Aggressively courted by Aston Villa to take on their vacant manager position during this off-season, Martinez declined, apparently out of loyalty to his current club: “Over the last two years the chairman has been very supportive to me and loyal, and now I feel I need to be loyal and supportive back to him. I haven't finished my job at Wigan Athletic; there is much work still to be done.” Sounds like the words of a man who has a cause, and is on the march. I am curious whether the Latics' luck will finally run out this year - the good news is that they have a leader who doesn't mind fighting the good fight with the weapons he has.

Tuesday, November 15, 2011

An Interview With ESPN's Soccer Analysts Paul Carr and Albert Larcada

With apologies to Gil Scott-Heron, the data revolution that is currently happening in football will not be televised. Instead, it is taking place in the cramped back offices of clubs where scouts are now as likely to be hunched over a computer to check a player's performance indicators as they are to be swapping stories of watching a young Algerian striker on a rainy night in Copenhagen. Instead, the data revolution is taking place online - on blogs and fan websites, but also on the screens of the most powerful and most watched sports network in the history of humankind, the Entertainment and Sports Programming Network, aka ESPN. With a few clicks, fans of any sport that matters can find all the data they want to become literate in stats or confirm their hunches that their favorite team really is as good as they think (hopefully).

I am currently in London on my way to Manchester, tracking down and talking to some of the people interested in football's evolving relationship with numbers from one of hostility to one of friendly co-existence and perhaps mutual admiration. But before I left to probe English football's affinity with evidence-based decision making, I had a chance to interview a couple of fellow data revolutionaries back in the U.S., Paul Carr and Albert Larcada.

Paul and Albert are the two guys responsible for what ESPN does in the areas of soccer analytics. They work for and with ESPN's Stats & Information Group, which consists of four departments: Production Research, Statistics & Analysis, BottomLine and Analytics. Together they provide accurate and up-to-
date information for all aspects of ESPN, including studio shows, game broadcasts and digital media.

Paul Carr
Paul Carr is the lead soccer researcher for ESPN’s Stats & Information group. He covered the 2010 FIFA World Cup from Johannesburg and now runs the Five Aside blog on ESPN.com. You can also follow him on Twitter at @PCarrESPN.

Albert Larcada
Albert Larcada is an Analytics Specialist in ESPN's Stats & Information group. He conducts advanced statistical research with the intention of better modeling and forecasting sporting events and leagues. You can see all of his written ESPN work here, and you can also follow him on Twitter at @adlarcada_ESPN.



Friday, November 11, 2011

Where Do Shots Go To Die? The Premier League in 2010/11

Blocked shots don't typically make the highlight reel. Of course, they don't, you might say - they're a non-event. Well, they are, and they aren't. They're an event that wasn't of much consequence in the scheme of things. But they are still "events" in the way that that the good folks at companies like Opta think about them.

In any case, I've been wondering about blocked (and other kinds of) shots lately. Lots of times, statistically minded fans and commentators will talk about shots as a measure of offensive or defensive team performance. Creating shots or not conceding them is generally taken to be a good thing. Hard to argue with, really, and that's not the point of today's post. Instead, I was wondering to what extent shots on or off - which is typically what we see in box scores - reflect the full panoply of shooting that happens in a match. Maybe, the little statistics nag in the back of my brain might say, there's a selection effect. Are we looking at a particularly interesting subset of shots, rather than all of them? If so, we may be under-estimating the real odds of shots creating chances in a match.

Thanks to the excellent Guardian Opta chalkboard feature and data I collected from last season, we can take a look at that to see the full profile of shots in the Premier League. So to start, below is the overall distribution (in percentages) of all shots taken in the league last season, categorized by goals (shots on target that went in), shots on target, shots off target, and importantly, shots that were blocked. I find this last category particularly interesting, because they are the shots that never really were.


As the graph shows, the most common category of shots are those that go off target; over 40% of all shot attempts fall into that category. It's not quite 1 in 2, but its not all that far off. The second most common category is blocked shots, it turns out, as about a quarter. So combined, shots off target and blocked shots comprise 70% of all shots. Who knew?! I certainly didn't. In contrast and predictably, goals make up slightly less than 10% of all shots, and about 1 in 5 shots are actually on target (21.3%). 

So far so good. One question these numbers don't answer is whether some teams were particularly good at avoiding blocked or off-target shots. The next graph provides the answer. 

Wednesday, November 9, 2011

A Short Interview With ESPN's Five Aside


A few days ago, I had the pleasure to answer a few questions from ESPN's awesome Five Aside blog, a product of ESPN's Stats & Information Group. In case you're curious, you can find the interview here. In case you're in the mood, feel free to provide feedback on this blog.

Monday, November 7, 2011

Trends in Efficiency: Are the Germans Getting Better, Are the Italians Getting Worse?

As I've said a time or two, one way to tell how good a team is is to look at how much of something it does on the pitch. More goals = better. Seems logical, it's not very hard to figure out, and you don't need to know very much about numbers to understand it. Hence, it shouldn't be surprising that counting things up has been the primary way of evaluating quality in football.

But another and perhaps more intriguing way to gauge performance is by looking at how well someone (or several someones) perform a particular task. That's where we get into the business of measuring and evaluating things like individual or team efficiency. Efficiency is a measure of quality, not just quantity.

Assuming most people wouldn't disagree with the idea that players who score more goals on fewer tries are better, I was curious if there are any discernible trends in efficiency across the big leagues of European football. Are leagues getting better (here defined as "more efficient")? Do we see improvements in football efficiency, as we do in many other areas of economic activity, from logistics to how we use energy?

So just for kicks, here's a quick look at goalscoring efficiency across the big leagues measured by the goals to overall shots ratio (measured at the team/match level) in the Bundesliga, the Premiership, La Liga, and Serie A. Below you will find graphic representations of efficiency ratios (goals to shots) averaged across all teams and matches for each season since 2005-06 (the period for which I was able to get data for all four leagues). Guess what?!


The Bundesliga has seen consistent improvement over these six seasons, while shooters in Serie A have gone on a steady decline when it comes to efficiency. By the 2010/11 season, Serie A was clearly the least efficient of the four big leagues. In contrast, the Premier League and La Liga do not show clear trends and instead appear to fluctuate a fairly constant mean. One exception: the 2008/09 Premier League season, in which teams were almost as bad as Serie A has been for a few years now.

So what does this all mean? As a native German who is only too aware of cultural stereotypes (thanks to helpful family members), I did cringe a little bit when I saw the Bundesliga getting more efficient and the Italian league become more profligate. All stereotypes aside, we do have to give the "Most Improved Efficiency Award" to the Bundesliga and the  "Most Consistently Efficient Award" to the Premier League.

But more than anything, I was surprised to see Serie A go on such a noticeable decline over such a relatively short period. Of course, it could be that the window of time covered by the data isn't long enough and that the picture would look different with, say, 20 years of data. We don't, so that's speculation at this point. But when you couple these numbers with the leagues' financial health or attendance numbers, you do start to worry about Serie A and you feel good about the trajectory the Bundesliga is on.

Tuesday, November 1, 2011

Are Teams More Efficient At Home? Data From the Premier League

(c) 2010 Premier League
According to the BBC, when Chelsea played Birmingham City FC the afternoon of November 20, 2010 at St.Andrews, they took 25 shots on Ben Foster’s goal. Birmingham took only 1 – yes one – shot the entire match. The end result? Birmingham 1, Chelsea 0. One way to read these results is to say Birmingham got lucky. Another way would be to see that Birmingham were ridiculously efficient that day.

Efficiency is an interesting thing. Analyzing football in terms of efficiency gets us away from just counting up good or bad stuff that happens on the pitch. Instead, it allows us to compare how well or badly teams or players made use of the footballing resources they had available. It won't surprise anyone to hear that Chelsea produced more shots than Birmingham that day. But it will surprise people that Chelsea were incredibly profligate, while Birmingham managed to make the most of what they had.

Speaking more technically, efficiency means connecting the inputs and outputs teams produce on the pitch. Take shots (inputs) and goals (outputs), as in the Birmingham-Chelsea example above. Birmingham's efficiency rate was 1, while Chelsea's was .04. The Birmingham Blues were literally 25 times as efficient as the Blues from southwest London.

Though it's a great example, this particular match also happens to be an outlier - an unusual occurrence. Goals to shots ratios are typically much more similar across teams in the top leagues. Commonly, some of the best teams outperform the worst ones by a 2:1 margin on efficiency (rather than 25:1). And some of the very, very best teams like Bayern Munich or Barcelona managed to produce goal/shot efficiency rates almost 3 times as good as some of the very worst like Bari or Sampdoria.

Part of what made the Birmingham-Chelsea match so interesting is that the home team was clearly much more efficient than the away team. Is this a pattern we should expect to see elsewhere? More specifically, are teams more efficient when playing in front of their own supporters than playing away from home? We know that home teams typically perform better than the away side in an absolute sense. They tend to shoot, score, and win more, for example (though the jury is still out on why exactly that is). But that's not the question here. Instead, the question is whether home teams are also more efficient - that is, whether they perform better in terms of connecting inputs to outputs.

Honestly, it's not clear to me that they should be. It's one thing to imagine that home teams, pushed along by a friendly crowd, referee, or familiarity with the home field environment, will produce more chances in an absolute sense. But why, given a chance, would a striker be more likely to turn that same chance into a goal, or a goalkeeper be more likely to prevent one at home than on the road? Yet that is what we seem to see, as the next graph with data from the 2010/11 season for the Premier League reveals.


Now before you get all up in arms, do note that the difference is small. A yield ratio of .1135 for away teams is smaller than the .1276 ratio for home teams. But it's not a huge difference. The difference translates to 1 shot per match: while away team had to take 8.8 shots on average to find the net, home teams only needed 7.8 shots, give or take.

So is this is something we see consistently across most teams? The quick answer is, well, not really, as the next chart shows (except if you're a Newcastle).