Are Manchester United and Arsenal Getting Lucky?
There are many ways to look at the effectiveness and reliability of a team but when you find a metric related to scoring goals that places them on the edges of historical parameters or beyond it pays to take notice. Both Man Utd and Arsenal sit perilously close to the positive edge of measures that surround the levels of good chances they have been creating and allowing as we shall soon see.
Due to the subjective element in clarification, Big Chances aren't my favourite measurement but much like PDO or error rates, they do a reasonable job of acting as a proxy and generalising, in this case to combine chances of similar quality into a catch all term. Ben Woolcock had a good go at measuring their value at the start of the year and we know they incorporate one on ones, penalties and any situation in which the attackers has no pressure and only the keeper to beat. People have incorporated them into expected goal models for the simple reason that they allow a differentiation to be made between these opportunities and others in lieu of measures reflecting levels of defensive pressure. The downside is the same as any measurement that isn't grounded in a simple, irrefutable definition, and it's a problem we encounter regularly in the design of metrics. Football's dynamism lends itself to a wide variety of potential analysis, but also a wide scope for disagreement as to the specific quantification of certain events.
For example, a shot or a save is pretty well defined, a big chance less so. Expected goal models, while long part of the parlance of analysts both private and public, still suffer conceptually from a lack of standardisation and multiple differently realised methodologies. Just this week we saw Arsene Wenger making an interesting move to mention them in conversation in reference to Aaron Ramsey's effectiveness in central midfield. This open mention of the metric was seen as a watershed by some but beyond an understandable reluctance to bring "secrets" to a wider audience, it would be surprising if clubs hadn't yet engaged somebody to monitor internal and league performance levels through such metrics.
With added complexity and judgement required to build more advanced metrics, it becomes important to avoid pitfalls on either end of the spectrum. Either elements of subjectivity creep into the process and potentially distort findings or the opposite, a lack of understanding leads to an unwavering belief in outputs, regardless of the actionable reality they need to exist within— "A little knowledge is a dangerous thing." Like any great recipe, each ingredient needs to be able to be identified and to have a specific and vital role; reason and understanding cannot be forgotten. Showing your working does not necessarily degrade an edge, for the quality of the analysis built from the tools at hand is the prevailing strength. Personally, I prefer a more multifaceted approach to team or player analysis, too much reliance can be put on integrating multiple inputs when a list of complementary but differing factors can give a broader series of checkpoints.
The purpose of this column is to lift the lid and give the reader a different slant than the wider narrative about what is going on in this league, so we press on.
These "Big Chances", exactly how big are they?
They are all those occasions in which fans leap out of their chairs to celebrate a forthcoming goal, except around 60% of the time there is no goal and the fan slumps back before turning to whoever is nearby and making severe allegations centering around the identity of the errant player's father.
Like any kind of metric involving shot conversions, rates can bounce around all over the place; just ask Roberto Soldado, who, outside of his fair record for penalties, managed to convert on only three of nineteen opportunities during his time at Tottenham. Even Sergio Aguero, who is about as good as a striker gets during recent years in the Premier League has only converted 35% of his non-penalty Big Chances.
The chances are BIG, the rate of conversion is comparatively high, but certainties they are not.
Looking at the rates at team level is nonetheless informative and gives us a couple of clues about teams so far this year. Similar to when I've looked at metrics before, when you have an ever-growing sample of seasons, it's easy to find outliers. The degree in which they sit at the edge of extremities or beyond in itself may not be enough to declare definite imminent reversion, but it acts as a significant clue to areas in which a team may be over or underperforming in comparison to longer term "normal" rates that have gone before. And that creates a valuable checkpoint for teams and their performance levels. Sample sizes are necessarily on the small side but have we any teams that sit beyond the horizon this year? We sure do.
(All data prior to 11th Dec, 120 team seasons from 2010-11 onwards)
- Their Big Chance For rate is higher than any full season in the sample (3.40/game, 2010-11 Arsenal: 3.29)
- The difference between their rate and the opposition's is higher than any full season (+2.33/game, 2012-13 Man City: +2.21)
We can posit two theories here, either a) this Arsenal team is one of the decade's great teams or b) they are a good team but there is room for their numbers to drop back. I'm keener to endorse the second theory for although this team has excellent volumes here both for and against, that they exist outside historical season long levels makes me inclined to estimate that it is more likely than not that they will revert over time. If we compare their generally very good but not significantly superior overall shot numbers, or indeed any of the variety of expected goals measures, this analysis is supported.
On a slightly different angle, they have also slightly ridden a positive conversion rate here, the opposition have only managed to convert 4 from 16 of their Big Chances, so to go with a low rate of concession is a low rate of conversion. They are a good team but their opposition has failed to finish good chances at a reasonable rate. This is also borne out by a look at their overall save percentage, which is still riding high at 80%. Although that has been the general rate throughout 2015, even a year and the volume of events within is still a sample too small to presume that there is something methodical behind it. They are clear title contenders but the only positive move left for them within this aspect of their game is an increase in their own conversion, all other aspects are more likely than not to come down.
2. Man Utd
The mystery of van Gaal and his shot-shy methods continues to provoke debate but here we find a different view: Man Utd have been ludicrously fortunate in their conversion rates from Big Chances:
- They have converted 63% of their own Big Chances, a full 9% higher than the highest rate in the sample (Chelsea and Man City 2014-15 both at 54%, only seven teams exceed a season long rate of 50%)
- Their opposition have converted at 24% a lower rate than any team in the sample (Everton 2013-14 were broadly similar at ~24%)
The result of both topping out their own numbers and bottoming out their opposition – and remember as these are conversions there is no factor for team quality or control here – is that the difference between these two rates is actually higher than the Big Chance conversion rate for the entire sample. (As an aside, this rate appears to be slowly increasing year on year)
As the For rate for Man Utd is so high, even a positive reflection that envisages them residing at the outer positive margins of this metric would mean that they would revert over time. Combine this with the truly dismal shooting numbers and you have an explanation as to how a team that shoots at a relegation rate can be hanging onto the coattail of the title challengers. When they have had good chances, they've been going in, when the opposition have had good chances, they haven't. That isn't sustainable. With low shot and goal event football, an edge gained, even if via fortune has made the difference between draws and wins; this skew is worth about eight goals (five for, three against) and for a team that has found 11 of 16 results finish either a draw or with a one goal margin, that's a huge impact.
With van Gaal's tenure hitting patches of ice this last week thanks to a dismal departure from the Champions League and another tepid performance in losing at Bournemouth, the potential for this wobble to turn into a full blown skid is very real. He must be hoping that similar to last season, improved performance levels will be forthcoming as the new year arrives, but for them to continue to compete, even for the top four, it is essential that the volume of their shots and creation of chances increases. They cannot rely on a historically significant positive skew in conversion to last forever.
This season, Bournemouth have been a fascinating combination of good, bad, unlucky and well, unlucky. We have seen them post reasonably solid shooting numbers, especially on the defensive end and shown solid territorial and passing work yet their save percentage has been at an all time historical low, something that has been unavoidable due to them serving up a raft of good chances from beneficial locations. All this is borne out in the Big Chance stats. Worryingly for them, they have an extremely low For and high Against rates (1.00/game and 2.47/game) and both are going in at a high rate (47% and 59%). That volume difference resembles that of doomed teams of yore like Reading 2012-13, Bolton 2011-12 or Blackpool 2010-11, but these were teams who had bad shots against numbers overall, not under 11 per game.
They remain vulnerable but are hard to truly understand. Occasionally you find teams that legitimately stand alone and for now Bournemouth fit the bill.
- Only Hull and Villa last year created Big Chances at a rate lower than this year's Chelsea (0.92/game vs 0.93/game)
- Norwich rate against is bad: 2.73 against looks like relegation level
- Liverpool's rate was high, 2.4+/game 2011-12 until it topped out in 2013-14, the season and a half since have seen decent overall shot volumes but Big Chance rates of under 1.5/game.
- Ditto post Ferguson Utd teams
- Leicester have conceded 13/19- a huge 68%... are they better than they even seem?
We'll leave the last word on this topic to Paul Riley, who lest we forget has provided a multitude of Premier League stat toys available on the site, from Chance Creation Maps here to an Expected Goals dashboard here
Thanks for reading!