Arsenal, Score Effects And A Season Of Two Halves
Arsenal have been one of the more intriguing teams in the Premier League this season. Sequences of results appear at odds with their underlying numbers, which for the most part have appeared impressive, and we enter March with them in a familiar position- not good enough to make a title challenge but heading the chasing pack and seeming a likely bet to maintain top 4. The method in which they have acquired this status couldn't be more different to that which powered their 2013-14 run. My interest in this notion was sparked by the significant drop off in their recent shot totals and a desire to find out “why?”. Let's see what we can find out.
Over the course of a number of years, and at least since the advent of the super-rich clubs, Arsenal have been a solid, top 4 standard dominant shots team. Last season, by their own standards was a weird outlier and their shooting numbers were in amongst the chasers, the 5th to 8th brigade, with the eventual top three clear. This demarcation is well represented by looking at the “+/- shots on target” stat:
That they came out of this with a comfortable 4th place was testament to the remarkable start they had to the season and the method here will inform our analysis of 2014-15. At the start of 2013-14. Arsenal won 8 of their first 10 games (and 11 of the first 15) but most interesting was how they managed to do it. In those first 15 games, they took the lead in 13. That taking the lead would have a strong correlation to winning a match is no revelation but what's important here is the time of the first goal: ten of the 15 matches featured an Arsenal goal within the first half hour. As such, Arsenal spent a significant proportion of their game-time in the early season ahead. The knock on effect from this (as has been shown by Benjamin Pugsley, Sander @11Tegen11 and more recently Garry Gelade) is fundamentally intuitive if you think about the sole purpose of a football match: to win. Once a potentially winning state has been achieved, the onus of each team changes; the leading team is less inclined to attack and therefore shoots less often and the converse is true for the trailing team. In isolation this scenario can have an impact on the nature of results, other things can too. Things such as conversion rates:
Not only were Arsenal taking the lead quickly and regularly, the opposition were finishing at a progressively woeful rate. So within 15 games we had a huge skew that informed the rest of the season. All of this masked a deficiency to compete at the very top of the table, a gap that became vividly clear in subsequent heavy defeats at the hands of the Chelsea, Liverpool and Man City. At that point Arsenal were nominally in the hunt for the title, as the season wore on , and in no small part hindered by injuries, they regressed to what was probably their true level: the back-end of the top 4 race. Indeed in the remaining 23 fixtures, they only took the lead within half an hour in 4 matches and team conversion levels fluctuated sharply in both directions. Overall the seasonal stats were clear: this Arsenal team had significantly overachieved against expectations derived from their shot statistics. They had also spent very little time under any threat from teams beneath them. Only a strong late run from Everton gave cause for concern and eventually they petered out. The battle for the top 4 never came to fruition and Champions League qualification become a formality, with a series of late season wins.
So what's this got to do with this season?
For much of the early season and through into autumn, Arsenal's results to underlying stats were the opposite to last season. They shot like contenders and repressed the opposition at an impressive rate. But the kicker was in the mixed results. In recent weeks, however, the opposite has been true. In particular, since being thoroughly outperformed yet drawing against Liverpool, their shot numbers have become decidedly par, yet results have improved. Initially, I set about writing this as a warning to Arsenal's long term prospects; they seemed to be declining as the season went by:
So straight away we have a storyline: first 13 games, title contending shot numbers but only 1.54 points per game (PPG), a rate which isn't even going to qualify a team for the Europa League. Since then, negative total shot numbers but 2.15 PPG. That's a top 2 points rate and bottom half shot numbers. As a visual aid, I can show the rolling 5 game numbers and it looks decidedly and progressively worse:
Not good, right? Moving in the wrong direction. Warning signs? Well: yes and no. Obviously it is preferable to have a title challenging shot profile but when you're winning matches and other aspects of play are coming to the fore, is it less important? Having spent a deal of time pulling out shot totals at given game states for Arsenal's season, I'm in a position to observe what has been going on beyond the raw shot totals. And it's revealing.
At a tied game state, Arsenal post excellent shot totals, basically a 2:1 ratio. Going further in we can learn a few more things:
At 0-0 in particular Arsenal are extremely solid, they restrict the opposition and take a lot of shots themselves. However, in looking at the two halves of the season, there is a substantial reduction in the time of the first goal (39 minutes to 27). As we can see in the first part of the season, Arsenal only scored the first goal in 6 out of 13 games, so, not only did the first goal of the game take longer to arrive, they were also regularly faced with a deficit. Each of these aspects seems to lend itself towards higher shot rates. Because they were now struggling to obtain leading positions, they were obligated to attack more, either to obtain a lead, or to recover from being behind. This is the opposite scenario to the start of their 2013-14 season.
Latterly, they have taken the lead in 10/13 games and have scored a number of early goals. This is very reminiscent of early 2013-14 and their seemingly mediocre shot totals follow this blueprint. In those most recent 13 games, they have spent only around 120 minutes in a losing position in comparison to nearly 300 minutes prior. They have also spent over 600 minutes leading or rather in the position in which they will aim to end the game. This is over treble the total from the first 13 games. Where have Arsenal's shots gone? They simply have no need to take them at the rate they would if level or behind.
So what we can see here is that beyond the raw shot totals we have found nuance in the 'score effects'. The generally sub-par recent shot totals are only one window into the whole shape of the team's shooting performance. Is anything else informing the storyline?
To say Arsenal's opposition conversion rates were verging on the fortunate in the early months of the season would be something of an understatement:
Here we can see that throughout the season Arsenal have been fluctuating around league average with regard their own shot conversion. In contrast, and something that must be factored into their moderate early form, their opposition have spent the entire year converting at well above league average rate. That they have now managed to bring that rate down to around league average will have had a significant impact in powering their recent better form.
There are still problems with Arsenal's results in games with their rivals for the top 4 and Chelsea and City. Their pragmatic hit-and-run victory at the Etihad and a dominant win over Southampton are the only two wins in nine such fixtures and whilst in the first 5 of these fixtures they weren't out shot, the more recent big games have been characterised by poor shooting performances. Again the Liverpool game appears a watershed, being outshot 7 to 27 was both very good by Liverpool and pretty abject from Arsenal. Similar shot totals were recorded in the defeat at Tottenham (7 to 23). In Arsenal's defence, each match involved a long spell of leading but that they were ultimately unable to repel the opposition and dropped points suggests that there are still personnel and tactical issues that are required to be looked at before Arsenal can ever consider challenging for the top two places in the league.
The huge signings of first Ozil and then Sanchez were rightly envied by the rest of the league. Sanchez is already one of the best players in the league: his shot contribution of 6.2 per/90 and his goal contribution of 0.95 per 90 (0.60 Goals + 0.35 Assists) are extremely high and he's a worthy contender for Player of the Year. With quietly Ozil's goal contribution levels not too far behind (c. 0.80 per/90), and despite a lot of criticism for him, there are two bonafide world class talents scheming in Arsenal's attack. With able assistance all around the attacking midfield positions and Welbeck coming in to provide solidity, if not too many goals (c.0.4 goal contribution per/90), as an alternative to Giroud, the attack seems pretty well set.
Finally signing Gabriel, albeit 6 months late, should give some kind of respite to Mertesacker, who appears to have suffered a World Cup hangover and adds long required depth. As ever there are question marks around the lack of a dominant central midfielder with the talents of an aging Arteta, functional Flamini and now Coquelin not seeming to be long term solutions.
It is rarely a positive reflection of the goalkeeping situation when cloudiness appears around the first choice role either.
Having climbed to third, and now within sight of Man City's wing mirrors, Arsenal can feel relatively confident that they can repel challengers beneath them. Throughout this season, i've had them projecting to be the 3rd or 4th best team in the division and I can represent this here via Shots on Target:
In truth I could have used any one of four or five metrics to represent this. We're not looking at shot locations here but as an approximation of how teams have performed over the course of the season, it's an interesting tool. The inadequacies or otherwise of a few other teams can also be seen here. It also shows that where last season there were 3 superior teams and 5 chasers, this year we have probably 5 'better' teams with the top two clear of the others. Man Utd and Tottenham have exceeded expectations across underlying metrics and it would require a significant improvement, a further increase in already good fortune or maybe an act of God for them to secure a top 4 slot.
Personally, I suspect the top 4 of 2013-14 will claw their way back to the top. I posited some while ago that Brendan Rodgers had a talent for improving his team throughout a season, and whilst it has required an unorthodox formation for him to extract the best from his squad, I feel their rise through the league is unsurprising and likely to succeed. After that, a possible battle between Southampton, Man Utd and Tottenham to avoid qualifying for the Europa League should be fascinating.
As someone who regularly writes content based around shot metrics, it may seem counter productive to add in an extra layer of analysis that may or may not undermine simple conclusions that can be gained from shooting numbers. But the more I poke around the numbers generated by the league, the more I realise that the storylines that can be read from them are more nuanced than may initially appear. This season's Arsenal team are an ideal case to represent ideas around score effects in a hopefully accessible manner and show that simple biases are generated purely by how long a team exists within a specific game state and the scoring structure of a game.
Thanks for reading.
Apologies for the lack of a round up column this week, you get this instead!
A small consequence cup final has prevented my Premier League data from being complete. Let us hope order is restored soon.