Expected Goals and Unexpected Goals

Home/Football Analytics, On the Pitch, Recent Posts/Expected Goals and Unexpected Goals

Expected Goals and Unexpected Goals

Now the 2016 season has come to an end, and before we get into the next campaign  I thought it would be fun to look at something a bit less serious than securing a Champion’s League place or avoiding relegation.  So let’s look at great goals.  Of course, what makes a great goal is entirely subjective, and often the perceived quality of a goal will be coloured by the context in which it occurs; I suspect that many of the goals we remember as outstanding were actually important in some way or other, perhaps scored in the dying moments of a match, or sealing a vital result.  But I want to be a little more objective and focus on the characteristics of the chance and its creation; do goals judged as great differ systematically from ordinary goals, and if so in what way?

Goals we judge as great are often subjectively experienced as surprising, astonishing, coming out of nowhere – in a word unexpected; statistically speaking they have a low probability of being scored.  If this is so, great goals should have lower xG (expected goal) values than ordinary goals. In this post, I’m going to test this idea.

My list of great goals comes from the BBC’s Goal of the Month competition which covers the English Premier League.  Each month, Match of the Day pundits select between six and eight candidate goals which are voted on by the general public; for fairly obvious reasons I call these the GoM goals. The winner is the GoM goal with the most votes. The Premiership also has its own competition (branded the Carling Goal of the Month competition); the candidate goals in this case are chosen by a panel of experts including such luminaries as Alan Shearer, Rio Ferdinand, Sir Alex Ferguson.  An eyeball check reveals considerable overlap between the two lists of goals.

In the 2016-2017 season there were nine monthly votes in the BBC’s competition, and 65 GoM (Goal of the Month) goals.  They represent about 7% of the non-penalty/non-own goals scored in the season, so perhaps not the greatest goals in the history of the game, but hopefully a reasonably non-controversial list of eye-catching strikes.   In the analytical database I actually have ten rather than nine winning goals because one month the vote percentages (which are reported to the nearest whole number) resulted in a tie for first place.

The chart below shows the xy locations of the GoM goals,  and a random sample of 100 ordinary goals (excluding penalties) for the season.  I’ve split the GoM goals  into winners and ‘contenders’ (in the Marlon Brando/On The Waterfront sense). It is clear that many, but by no means all of the GoM goals are scored from further out than the typical ordinary goal.

Fig.1 Shot locations for ordinary goals and great goals

The next chart shows the final balls for the GoM goals –  often long and laser-like precision passes.

Fig. 2 Final balls for great goals

The table below shows some of the main differences between ordinary and great goals, splitting the GoM goals into contenders and winners.  (The column labelled sig is the statistical significance of the difference between the ordinary (N=881) and GoM (N= 65) goals.  It is the probability of finding such a  difference by chance, so the smaller the value the more significant the difference is.)

Table 1. Descriptives for ordinary and great goals

MetricOrdinary (N=881)Contender (N=55)Winner (N=10)sig.
Distance to goal (m.)11.019.519.7.000
Driving distance (m.)3.88.06.1.007
Defensive pressure1.81.51.5.020
Defensive advantage0.81.31.3.001
xG.24.10.13.000

The distance to goal metric is simply the distance between the shot http://www.mindanews.com/buy-inderal/ location and the centre of the goal line. We can see that great goals are typically scored from further out than ordinary goals, confirming what we saw in Figure 1.

The next figure in the table is the driving distance. This is the distance between the location where the shooter received the final ball, and the location from which he shot.  The difference is quite remarkable; scorers of great goals cover twice as much ground before shooting as the scorers of ordinary goals. But again that accords with our intuition; creating a shooting opportunity is very much part of a great goal – and incidentally something that the xG metric doesn’t reflect. (Tech. note: Driving distances here are only calculated for chances preceded by a final ball – after all this is only a bit of fun, so I’m not going to the effort of calculating it for all shots.)

There is a small but significant difference in defensive pressure (indicating that somewhat less pressure is applied to the shot-taker scoring GoM goals), but the defensive advantage (excess number of defenders) is significantly higher.   This suggests that shots that penetrate a heavily defended area than average are more likely to be viewed as outstanding.

The last line in the table shows a large difference in xG values. Here, xG (or expected goal value) represents the probability of a goal being scored depending on such factors as the shot location on the pitch and game context.  We can see that xG for GoM goals is about half that for ordinary goals.

The chart below elaborates on the difference:

Fig. 3 xG for Goals of the Month and Ordinary Goals

As we can see in Figure 3 a few of the contender goals do in fact have a relatively high xG.  For instance, Giroud’s goal against Crystal Palace (xG=.47) was scored from about 12 m. out, following a cross from Sanchez on the edge of the box; it was Giroud’s amazing scorpion kick that elevated this goal to greatness, rather than the intrinsic difficulty of scoring on the break from that position.  Wijnaldum’s goal for Liverpool against Arsenal (xG = .45) was scored from around the penalty spot with only the goalkeeper to beat, a relatively easy chance under no pressure. What made this a great goal was the build-up play. Llalana started the move with some clever play in his own half, pinpointing a long pass to Origi running into space on the right, who crossed the ball to feet as Wijnaldum drove into the box. A devastating counter-attacking move which lasted no more than 20 seconds (Wijnaldums’s goal can be seen here).  Interestingly there are no quite so high xG goals amongst the public’s favourites, although the sample is really to small to draw any conclusions from. The highest is Mkhitaryan’s  scorpion effort for Manchester United against Sunderland (xG = .32).

Did any of the factors in Table 1 influence the public’s perception of ‘greatness’ as expressed in their voting? Not as far as I could see.  There was no correlation with the vote percentages at all.

Nevertheless, by and large, great goals are “unexpected” in xG terms.  A great goal is only half as expected – and  therefore  twice as surprising – as an ordinary goal.

This article was written with the aid of StrataData, which is property of Stratagem Technologies. StrataData powers the StrataBet Sports Trading Platform, in addition to StrataBet Premium Recommendations

 

2017-07-12T06:57:56+00:00 June 15th, 2017|Football Analytics, On the Pitch, Recent Posts|0 Comments

Leave A Comment