asprilla

There are many aspects of this Premier League season that have been hard to understand and few, if any, will leave having had their prediction skills validated.  I am going to have to live with the fact that I stated that the rise of the middle classes was a usual blip that would quickly revert, there has been no pay off for my continual writing off of the chances of Leicester City (there is a tweet in existence in which I comically suggest Ranieri will be gone by November) and after their blistering start, I laughed at ludicrous predictions that Manchester City would come fifth. Ahem. This being football, a ton of things can happen to make you look foolish, others can make you look smart, and that’s why in the main, peeling back the numbers and taking a more detailed look at matters can reveal the truth.

Simple stuff can leap off the page.  Chelsea allowing Swansea to create ten shots on target in their very first match was a huge red flag and with subsequent matches improving this starkly poor rate only slightly, it was possible to see even in a handful of games that Chelsea were in trouble. The rot set in and changes came later, but there was evidence from a long way out that although Chelsea had won the title, they had not been performing at league leading standards for a good deal of late 2014-15. It’s important to temper this though, nobody could have predicted how far they eventually fell before Mourinho’s departure, or that so many players would concurrently underperform, or that they would also find themselves on the bad end of a negative skew. But all that came to pass and probably weeks late, he departed.

Elsewhere Norwich’s problems this season were visible from a good way out. On the surface it looked like they were making a reasonable fist of their return to the league but after an early flurry, their shooting numbers were significantly poor, they have continued to struggle and are now close to the trap door.

Tottenham’s slow start points-wise disguised extremely solid shooting numbers and quickly identified problems solved in defence. While the traditional media happily worried abut the flash-in-the-pan nature of Harry Kane, it was straightforward enough to rebut this even though his goalscoring had temporarily vanished. All other aspects of his shooting game remained, it was clearly a matter of time, and so it proved.

Informative analysis can be simple and understanding your team’s performance metrics goes a long way towards describing performances in the past and projecting likely events in the future. There are lots of tools for this around. The big sites carry loads of statistics, expected goals models are there to inform and everyone can get smarter for betting, their office fantasy football team or even just to win the pub argument. You can even write about what you find out and there are plenty of people who will find it interesting and latch on.

Here’s a few other stories that have unfolded more recently.

Arsenal and Manchester City

After twenty weeks of this season, we hosted a post on this site in which a variety of modellers gave their predictions for teams to win the title. Every single entrant bar one had the likelihood of anyone other than Manchester City or Arsenal winning the title at under 20% (I believe Constantinos Chappas was the outlier). Over in the Twittersphere, another prominent predictor was confident enough in Arsenal’s chances to declare in mid January that they would have been “over the line” model-wise with a victory (they drew). None of this is to suggest that everyone was “wrong”, merely to note that the small probability outcomes are looking as if they will come to pass: the twenty percent looks likely to be heading towards the winning line. So what has happened to the two early season giants?

The late winner against Leicester was the spark that was supposed to ignite Arsenal’s season, yet they’ve had to wait til this weekend to find another win, the far more routine Arsenal-style “get two ahead and sit” kind that they last pulled off at Bournemouth. This has left them teetering precariously on the very edge of the title race and looking far more likely to end up in, lo and behold, third or fourth. Plus, this week, we’ve had the annual “should we get rid of Wenger” discussion which is interesting, because from an analytics perspective we have a slight divergence; expected goals loves this team and will fully expect them to cruise home the rest of this year and possibly kick upwards next. Even within this analysis they are only projecting as the best team in the league this season, not in any season, for this one is compressed.  Simple shot analysis is more content to christen this team ” a typical Arsenal team”, good but not tearing the roof off the league and it is worth looking at how they’ve trended more recently.

Since the halfway point, there has been a negative swing of four shots and two shots on target in their figures; the latter a negative swing larger than any in the league. This goes alongside entirely flat conversion rates across both periods, league average on the attack end and advantageous on the defensive end (save percentage continues to be super high). What this means is that without any increase in the fluctuating skews that affect any team, Arsenal have distinctly declined since the turn of the year. They have no room to dig up extra luck, they have simply been worse and their results have suffered because of this. It would be pertinent to note the injuries they suffered in particular during the early part of this period but–despite the continued absence of Cazorla, who as each week passes appears more fundamentally tied to their effectiveness–more recently they have had a larger squad to pick from, yet haven’t really picked up. Then Ramsey goes down, Iwobi appears and we’re none the wiser as to who or what Arsenal really are.

Sacking the manager would be a bad idea. One year, maybe they stay fit and they do challenge late and long and… well, we’ll see. Pep Guardiola may have steamrollered the league after a sixteen player transfer splurge at City by then.

Regarding Manchester City, only Watford’s goal volumes have decreased by a greater amount, and that’s the top line “where have the results gone? indicator but the real issue surrounds their accuracy. While their shot volumes have broadly held up (minus 0.6 shots for per game), their on target rates have fallen by a huge amount from over 6.5 per game to a frankly average four. So from being a team that had shot volume and accuracy, and projected very well because of it, they are now finding it hard to break through.  They’re still taking two shots for every one from their opposition but the on target rate is more like 4:3 (literally), they are converting their shots at 8% and the opposition 13%. The conversion part should shake out in time but it’s all a lot nearer par than you might expect from a team of such clear talent.  Like Arsenal they have been hampered by repeated injuries and the physical cost of being unable to rest an ageing squad but the 26 shot, three on target effort against Manchester United felt all too familiar and one wonders how motivated the troops are, knowing that a tough drill instructor has been recruited for their future.

Sunderland and Newcastle

There are clear similarities in the shooting numbers of the North East relegation rivals. Both produced abysmal totals throughout the first half of the season and both have improved since then. However, the nuance from there differs.

Things have actually started going right for Sunderland but from a very low base.  They have produced a swing of 3.6 shots and 1.7 shots on target per game, which is enough to have raised their numbers to just sub-par levels. All their shot conversions have moved with them too, Sam Allardyce has found some efficiency, at least for now and they are turning a high 39% of their on target shots into goals since week twenty. This creates the discussion of whether Allardyce’s methods are adept at squeezing the maximum return from typically minimal inputs and the timescale is too short to have any confidence there, but he has realistically improved the output of the team. Whether it will be enough to save them is another story, they still aren’t good and the seats on the Premier League 2016-17 bus have filled quickly. And of course, there are long held problems surrounding squad building and team quality that still need resolving, something that can only be realistically done if they stay in the league: no easy task.

Newcastle have improved by a far greater margin than Sunderland. Since the turn they have enjoyed a plus 8.6 shot swing and plus 3.6 shots on target. Nice job Steve McClaren! Post Christmas, his team has been performing with a shot on target ratio ahead of Manchester City and Chelsea, way ahead of Arsenal and just a shade behind Leicester and Liverpool. Isn’t it great when things work out for someone?

sot weeks 20 onOh hey, look at Tottenham…

Anyway, the volumes are the good side of Newcastle and over time, you would presume that they might turn into goals and points but sadly for Steve and his P45, it has come all too late. During this period of reasonably solid and improved underlying shooting numbers, the conversions have been thoroughly and entirely in the bin. Newcastle have been converting 7% of their shots (low) and conceding 15% of the opposition’s (high). Of the on target efforts, they have been converting 19% (very low) and have a save percentage of 49% (utterly miserable).  All that has contributed to them scoring ten, conceding 21 and winning two from eleven games, despite promising and much improved shooting rates. They too have suffered with injury, though mainly in the autumn and if this improvement continues and Rafa Benitez grabs a bounce they could find themselves fixed and quickly. Either that or they are going to find themselves as unlucky to go down, much as Hull City were last year, as they ran out of time.

Quick bits

  • Everything Watford shaped has taken a big hit in the second half. They had enough on board to cushion them but both in results and metrics they have struggled through 2016 and may need to do more than take stock in the summer.
  • Swansea haven’t improved but they have got a few more breaks
  • Leicester’s metrics are insane. As their attack dried up around the halfway point, their defence has gone supernova. While they are effectively limiting on target efforts, which is to be applauded, the opposition is currently converting all of its shots at under 4%: a rate far below any realistic long term expectation. There is no strategy in the world that can cause the opposition to score at such a low rate long term, so we are left with “they are lucky” or “they have solved football”. I lean towards the former. In season they are now a sub-50% shots team, which would normally as you might expect, peg them as a par team for the league. Again shot on target or expected goals paint a rosier picture, one of the fringes of European qualification, but nope, they are five points clear and seemingly striding without trouble towards the title. It is quite remarkable.
  • The Palace collapse has two factors, firstly a large negative swing in their ability to get and prevent shots on target and then a total collapse in the conversion rates to almost Newcastle levels. These latter factors are largely out of a team’s control and given that they skewed very positively in the first half of the season, this opposite effect appears to be creating a reflection of their true ability: not so great.
  • Liverpool games have become more high event. Goals and shots are up on both ends and I joked recently about them seemingly being plugged into a random results generator. The Southampton game, and to lose in such a manner despite strong shot volume, endorses this silly idea.
  • Bournemouth and Chelsea have similar profiles. Neither has particularly improved their shooting numbers, but each has started to enjoy the benefit of a positive rate of shot conversion, the exact opposite of what came before. Bournemouth’s pre Christmas goal concession was so severely bad as to be clearly beyond a level that would sustain, and so it proved. They have produced very solid defensive volume numbers almost throughout.

So, there’s a few aspects that have driven results since the turn of the year. Plenty of other stories exist within the league and plenty of tactical and personnel inputs will drive aspects of play too, but in taking a look at the numbers we can understand what aspects of a team are performing well or badly and which aspects are likely to cool off or heat up given time.  Even in a year as surprising as this these truths remain.

______________

Thanks for reading!

Long pieces on Tottenham and Manchester United by me.

@MoeSquare on Marseille

@Fla_Futbol on Sampdoria and Genoa

Improve Performance and Productivity in Your Club:
State-of-the-art Football Analytics