By Freddie Wilson – @thewonderofmu
The Placement of Penalties
The award of a penalty brings the expectation among fans of a goal, but what does the data tell us about the specific probabilities of success?
On average, penalties have a 76.6% conversion rate. In terms of expected goals, this means that they have an xG value of 0.766.
Using a dataset of over 9000 matches taken between 2011-2017 from the Premier League, La Liga, Bundesliga, Serie A and Ligue 1, we can investigate where the most successful penalties are placed. The table below shows the penalty placements and their success rates.
This means that, on average, if a penalty is simply on target then it has at least a 76.7% chance of being successful, with the worst place to put it being the centre of the goal (76.7% conversion rate) and the best place being either of the top corners (96.4% conversion rate for the top left corner and 95.7% for the top right corner).By definition all off target penalties are unsuccessful and I think that it would be best to exclude the “top centre of the goal” category as there were only four penalties placed there. For anyone curious, these four penalties are Johan Cavalli (AC AJACCIO v Sochaux 02/05/2012), Roberto Trashorras (RAYO VALLECANO v Barcelona 21/09/2013), Isma Lopez (SPORTING GIJON v Athletic Bilbao 21/08/2016), Adam Ljajic (TORINO v AC Milan 16/01/2017).
It does not seem like there are many attempts at the top corners of the goal, these are the least common on-target placements. However, they reap the highest reward but the lack of frequency may be explained by the risk of the strike being off target for these goal locations.
Variety of Placement
Following Romelu Lukaku’s saved penalty against Leicester on Saturday, 26th August, his tally became 10 out of 15 penalties scored in domestic football (dating back to his Anderlecht days). Similar to his previous penalties, Lukaku placed that penalty to his left and so Schmeichel was able to guess correctly. We therefore investigate whether varying the placement of penalties offers more success than keeping to familiar placement.
Let’s look at one of Lukaku’s teammates at United: the superb Swede Zlatan Ibrahimovic. He has been a prolific penalty scorer during his career; of his 37 penalties taken in the dataset, he has scored 34. (Note: the data is taken from games between 2011 and 2017; hence we are missing some of his penalties. His full record stands at 69/79 scored)
His placement record for all penalties taken looks like this:
|Bottom left corner||Bottom right corner||Centre of the goal||Top left corner||Top right corner|
So we see that Ibrahimovic definitely favours the bottom left corner when it comes to penalties, but how often does he vary his placement? The answer is 58% of the time. That is to say, 58% of the time, the penalty he takes will be placed differently to his previous penalty.
On the face of it, that seems pretty varied. But looked at another way, it means that 42% of his penalties will be placed in the same location as the previous one.
This plot measures the rate of change in penalty placement (compared to the previous penalty by that player) against the success rate of that penalty taker. Note that these figures are for on-target penalties only and for takers with at least ten penalties to their name in the dataset. The first condition is because we are investigating penalties saved and so we need the penalty to be on-target to begin with. The second condition is to ensure each player has a sufficient number of penalties to his name.
We can see that there is little correlation between change in placement and success. For those readers who are statistically-orientated, the correlation coefficient is 0.015, indicating very little correlation. According to this, varying the placement of penalties does not necessarily bring success.
We find examples of players with a low percentage of change and a high success rate (Paulo Dybala of Juventus) and also players with a high percentage of change and a high success rate such as Francesco Lodi.
So perhaps there are other factors that come into play. Firstly, the measure of variety that we are using only considers one penalty to the next, rather than all of the player’s penalties. Secondly, as we saw in Part I, there are shot placements that have such a high success rate that it may not matter whether the previous penalty was also placed there, such as the top corners. A more sophisticated classification of shot placement would also be useful here, rather than relying on broad categories of placement.
So from this we have no conclusive evidence that Lukaku’s repetitive placement was the reason for his penalty being saved, it could simply have been that it was a poor penalty in general.
If you fancy having a play around with the data yourself, then it can be found at kaggle.com.