Should direct Free Kicks be more indirect? | Opta Analysis
Wondered for a long time why more trick free kicks aren’t attempted?
When Boro are chasing a game and you’re wishing them to play a reverse pass around the wall as they stood over a free kick outside the box.
Surely a clever attempt such as that is more dangerous than a direct shot?
Football analysis is at a sufficiently advanced state that clubs should have worked out their best strategy when it comes to free kicks. shouldn’t they?
They must know if they should be shooting or passing from free kicks. Well, lets try and find out.
The first step is obvious; we need to assess how successful players are when shooting at goal from free kicks.
Using the data solely from a Premier League season and choosing a representative from each Premier League team to give an idea of the conversion rates of ‘specialist’ free kick takers.
Each player listed has taken the most or joint most shots from free kicks for his team (in the league) that season. Here are the players:
Included is the players success at shooting from outside the box alongside the free kick statistic, as they are clearly comparable; a free kick is a shot from outside the box with the added difference of a deadball and a wall in the way. It is a real surprise that player’s have an increased conversion rate in these deadball situations, you would have assumed the inclusion of a wall would have drastically reduced the success rate. The evidence suggests that the benefits of a stationary ball and extra time to prepare have a greater influence than having a wall and the added mental pressure of a deadball scenario.
That 11 of the players have not converted a single free kick, along with 5 of the teams, highlights the need to study this area further. It is clear that shooting from free kicks is clearly very inefficient, unless your name happens to be Seb Larsson.
Only 1/18 shots from ‘specialists’ and 1/19 shots from free kicks overall directly result in goals.
It is of course possible that by including only free kicks that are taken as shots we are excluding other free kicks which are already taken as passes. However from personal experience and observations I do not believe this happens with anywhere near the frequency to interfere significantly with this investigation. I believe that, in the Premier League at least, most free kicks close to the box and in central locations are taken as shots most of the time.
The ideal solution for analysing the alternatives would be to look at free kicks that are already passed in positions where a direct shot is usually taken. Unfortunately this data is not readily available and it would take a massive effort to gather the raw data. The only solution is to manufacture this stat using readily available statistics.
The most obvious difference in passing a free kick is that possession will most likely be retained, and will be retained in the final third. What is the relationship between keeping the ball and scoring goals?
It is interesting to note that possession stats are based on ‘passing volume’, that is a team’s total passes over total passes in the game, and has no time element, Opta’s possession stat is calculated in this manner. This type of stat could be construed as misleading as most people understand possession as a time based statistic ‘how long a team has possession of the ball for‘ but the relation between the two is almost equal. Using this interpretation of possession, a player physically shielding the ball in a corner or embarking on a mazy dribble up the pitch is not adding to possession, and in the same way teams who play lots of short, quick passes (such as Barcelona) will receive an inflated possession figure. This hampers analysis slightly. It would be difficult to analyse this problem using possession statistics anyway, but it does bring up an interesting idea.
Though we can assess free kicks using this ‘interpretation’ of possession, in a very rudimentary manner, by calculating how many goals are scored ‘per pass’ in the Premier League. Using data on the average number of passes per team last season (5189.6) and average goals scored (53.3) we can calculate, using regression analysis, that an additional pass in total would contribute 0.003goals. So a pass cannot be interpreted to have much of an effect on goals at all (although it is statistically significant).
If we use this method and merely counted passing a free kick as just keeping possession, ignoring the quality (ie. location of the ball), a passed freekick would have to result in 269.8 passes to ‘create’ a goal or 15.11 passes to equal the chance of scoring from a direct free kick (0.056). It is clear that this is not the stat to use. Although the fact that 14.03 passes are ‘equivalent’ to a shot really puts the issue into perspective.
The next step is to include the quality of possession. Handily data for passes in the final third is available, and by using the same method as before we come to a figure of 0.016 goals per final third pass.
To put this into context a team choosing to pass from a free kick would need to make 3.5 passes in the final third to equal the statistical chance of scoring directly from the free kick. This is a huge simplification, there are many influencing variables (defenders will be deep and clustered in front of the goal), but you can begin to see how ineffective the choice of shooting has been.
Using regression analysis again a through ball is calculated to contribute 0.14 goals – If a team manages to play a through ball from a free kick instead of shooting they have a chance of scoring 2.85 times more likely than directly shooting. Again, there are obviously other interfering factors here, but still, even without precise data it is clear that it is very possible that passing is much more effective than shooting from free kicks. For reference a through ball is defined as “a pass splitting the defence for a team-mate to run on to.”
We can also try and control some of the influencing factors.
Whilst browsing the through ball data it became clear that some teams who play a lot of through balls also don’t score many goals per through ball. These teams are Arsenal, Bolton, Chelsea, Liverpool, Manchester City, Tottenham, and Swansea. It is not too great a leap to make the assumption that the reason for such a low conversion rate is due, at least in part, to the team’s playing style which is relatively more possession based (Bolton perhaps being the exception), and whom thus could be argued to ‘encourage’ teams to sit deep against them. If this is true then perhaps we should base the through ball conversion rate from the data that these teams give us, as it will more accurately represent a deadball situation – defenders sitting very deep inside their half and inside their own box. Surprisingly using the data solely from these teams the result is actually greater at 0.18.
It is evident that these teams are more reliant on goals from through balls relative to the other teams in the division, therefore of course through balls will ‘explain’ more variation in goals (in fact the r2 for this regression is 0.94 compared to 0.79 previously, so statistically it does explain more of the variation). This makes them even more ideal – not only do they remove some of the problems with this analysis, but their through balls also explain more of the variation in goals relative to other teams in the division, who score more goals in alternative ways. For instance teams lower down the table may score more goals from set pieces, through errors and opportunism, or from gambling from distance (Stoke who are set piece / long ball specialists have the highest goals per through ball: 23 through balls, 36goals).
To be an effective alternative to taking a direct shot a team would have to be able to manufacture a successful through ball every 3.27 free kicks that they previously shot from (based upon 0.18goals per through ball). This is isn’t a great ask for instance I do not believe it would be impossible to practice a free kick and effectively implement it once every 3 free kicks.
According to Opta the average number of free kicks per game in the Premier League is 23.1. We however only want to look at free kicks where a shot is taken, for which there are 0.72 per game. Using my estimate of a through ball being 3.27 times more effective than a shot we could estimate that a team who played a through ball every free kick they would have previously shot from would score 5.07 goals per season, up from just 1.44 from direct shots, which is an increase of 3.62goals per season.
To put this further into context a team can expect 1.015 points for every goal they score, therefore they could potentially increase their season total by 3.57 points, more than an extra win.
I do not for a second believe that a team could play successful through balls every time they have a free kick – that would be ludicrous. However this analysis has highlighted the fact that teams could be much more efficient with their free kicks.
Obviously a team who began passing every free kick should expect opponents to adjust their defence against them, perhaps by sitting deeper and make sure there are no exploitable gaps. It is even theoretically possible that should a passing tactic become successful there may be a decline in application of ’the wall’ or an adjustment in the goalkeeper’s positioning, which would almost certainly increase the chance of scoring with a direct free kick.
It is interesting here to recall that free kicks have a greater conversion rate than for shots from outside the box. Considering how efficient free kicks are in comparison to open play shots I wonder how the absence of a wall would affect things.
Of course this analysis requires much more research to know for sure, but there is a good argument that a mixed strategy of shooting and passing would be optimal for most teams. Certainly though, in the current climate there is definitely an opportunity for a team to benefit from attempting more through balls or passes since most opposition teams will be anticipating a direct shot on goal.
This analysis is barely scratching the surface of what is possible with free kick analysis.
The research could be vastly improved by increasing the robustness of the dataset by including data from previous seasons and from a variety of leagues.
If we could isolate the effect of a wall and a goalkeeper’s subsequent positioning this would open up the possibility of game theory analysis of free kicks. However it may be troublesome to find situations which compare to a free kick without a wall. Perhaps we could assess the success rate of ‘push’ free kicks where the ball is passed laterally, allowing for a shot with a trajectory which bypasses the wall, perhaps shots from open play without any obstacles or find conversion rates for free kicks which are already passed as a through ball, but again this requires a lot of data gathering.