Pepe Reina & Liverpool's Defence: An Analysis
Is Pepe Reina doing a good job in goal for Liverpool this season? And is he receiving defensive support? Today on the blog, Mark Taylor breaks down the numbers and provides an analysis of the Liverpool keeper in season 2012/2013.
It was only a few short seasons ago that Liverpool were permanent members of the Premiership Big Four, but the emergence of Manchester City as a formidable new force has seen the Merseyside giants not only replaced in the pecking order, but fall into mid table mediocrity.
Andrew Brocker has graphically illustrated Liverpool's fall here, a decline that has yet to be fully embraced by the odd setters and that trend continued on Boxing Day when Stoke entertained the former European Champions at the Britannia. Stoke emerged comfortable winners despite the visitors being priced up as strong favourites , a price at odds with the pregame league position of both sides.
Bring the Data
The gradual release of more granular play by play data has begun to lay bare the reasons for a side’s successes or failures and while analysis of such data is still very much in its infancy, there are numerous key statistics that are universally regarded as both important and predictive.
The attacking side of the ball invariably attracts more attention because of the obvious connection between scoring goals and winning football matches. Also the chain of connection between recordable on field actions and the end product of goals scored is also more transparent.
Raw passing numbers for a side are more indicative of match situation and tactical approach and therefore overall, are poorly correlated with scoring goals. Assists and shots are the immediate precursors of goals and as such provide a strong indicator of how a side has achieved their current record.
Considering Shots On Goal
Midterm in the Premiership is the ideal time to use a team’s current season record to begin to draw meaningful conclusions about their present level of ability. Strength of schedule issues will partly be absent because teams will have faced almost all of their fellow members of the Premiership at least once and we have yet to see the January spending spree which can change teams, hopefully for the better.
Usually shooting data is used to assess the ability of the attacking players and by December strikers will have accumulated a reasonably sized sample of scoring attempts spread over a wide range of defences for tentative conclusions to be reached. However, even half a season of individual shooting statistics will require heavy regression towards the mean to allow for the considerable contribution from luck in these figures.
The position which can be most readily analysed using shooting data from a single incomplete season is that of goal keeper. Strikers have to share goal attempts between their fellow players, but a keeper who plays regularly is the first line of defence against shots on target and as such his databank is always the fullest.
Shots will vary in potency for a vast variety of reasons. Some, such as the origin of the shot and whether it was a header or not can be readily recorded, but others such as defensive pressure can only currently be guessed at. Nevertheless a workable model to investigate shooting efficiency can be constructed using primarily shot origin and type of attempt, matched with shot outcome over a representative sample of Premiership sides and opponents to produce an average expected outcome for shots from anywhere on the pitch.
Goalkeeping Analysis: Pepe Reina
It is possible to assign an average probability for outcomes ranging from likelihood that a shot will be blocked, on target or result in a goal for every recorded goal attempt. When this model is applied to actual shots faced by a keeper and the real outcomes are compared to theoretical expectations, we can perhaps begin to draw conclusions, not only about how the keeper is performing compared to league average, but also how well is his defence protecting him from the shooting attempts of opposing players.
Pepe Reina has been Liverpool’s first choice keeper since Benitez purchased him from Villarreal in 2005, but he has come under increasing criticism for his performances both this season and in 2011/12. Using a broad predictive model for shot outcomes, he would appear to be an ideal candidate to see if the critics are justified and we can then extend the investigation to see if part of his performance may be related to the wider performances of his defence.
Keepers, by their positioning and ability to close down attacking players can have a partial influence on whether or not a shot is on target, but it seems sensible to merge this with the overall duties of the defence and concentrate instead on how effectively they deal with on target shots.
With the exception of Premiership matches against Reading, Everton, Newcastle and Chelsea, Reina has started all of Liverpool’s remaining EPL games. Up to and including Wednesday’s Boxing Day night game at Stoke he had faced over 50 on target goal attempts, three of which were penalty kicks. His save percentage is just over 60%, but to add much needed context to this figure we should try to estimate the type and distance of shots he has faced.
|Keeper||Shots On Target||Expected Goals||Actual Goals Conceded|
The average on target shot faced by Reina has originated 15 yards from goal and four yards wide of either post, but there is a great variation spread over the 50+ on target attempts that have been fired at him. The aforementioned penalty kicks are likely to be converted close to the long term rate of just over 65%, while the most ambitious on target effort came from Southampton, originating from wide of the area and from midfield and carried a goal scoring likelihood of just one in 50. Cumulatively, an average keeper would expect to concede 17 goals allowing for the shooting position of the on target attempts faced by Reina. The reality is that Reina has allowed 22 goals.
There is mitigation in some individual efforts he has faced. Bale’s free kick at White Hart Lane appears to take a deflection, wrong footing the Liverpool keeper and making the save much more difficult than the raw positioning of the shot would imply. But equally some efforts from seemingly dangerous positions may have lacked power, so the hope is that these effects will begin to balance out in largish sample sizes.
In short, Reina has conceded more goals than the model would predict given the difficulty of the shots faced. As has his understudy, Brad Jones, on a much smaller sample size.
To conclude this study we can use our extended collection of Premiership shot data comprising all 20 teams and defensive units to see how likely shots are to be blocked or on target from any distance.
Defensive pressure, as exemplified at its best by Stoke under Tony Pulis in the Premiership can greatly reduce both the accuracy and potency of goal attempts. Stoke are an extreme case, sacrificing attacking intent for disciplined, defensive cohesion that relies on pressure and shot blocking rather than active intervention, but it is likely that all defences vary in their ability to make life difficult for attackers.
We need a second regression to analyse all attempts on goal because the one for keepers related exclusively to shots on target. When a player sets himself for any shot we don’t know whether it will be blocked, off target, result in a goal or a save. Therefore, a second regression is required to account for this extended choice of outcomes.
By comparing the number of blocked and on target shots Liverpool are allowing with the number expected by an average team once shot origin is accounted for, we may see if Reina’s defence is helping or hindering his cause. If a team is allowing opponents to hit the target more often than expected, this may indicate that the task of shooting is easier and the shots may also be better placed or more powerful as a result. This may partly explain Reina’s poor concession rate over the first half of 2012/13.
|Defensive Ability||Expected Total||Actual Total|
|Allowing Shots On Target||65||69|
The model has taken the x, y pitch coordinates of every shot faced by Liverpool so far this season and produced a likelihood that such a shot would be either blocked or result in an on target attempt by a league average defence.
By adding together these probabilities we can get an idea of how many such events an average defence would experience had they faced the range, distance and angle of shots encountered by Liverpool.
Liverpool’s defensive setup has blocked eight fewer shots than predicted by the model and allowed four more on target attempts. These results of course must come with many caveats, but generally they appear to show that in some aspects of defence, Liverpool should be regarded as no better than league average. Certainly below the standard traditionally expected from Liverpool sides of the past.
Similarly, Reina is also likely to be performing below the standard expected from an Anfield keeper, especially one touted as “the best in Spain” as he was upon his arrival. Around seven tenths of a goal per game was the benchmark concession rate for Liverpool defences when they were consistently competing in the top four.
Since they fell into the top ten, that figure has risen to more than a goal a game and the trend is continuing this term. Shot figures so far from 2012/13 seem to indicate that the demise should be shared between Reina and the defensively responsible members of the Liverpool team and the general criticism of Liverpool’s defensive unit is probably well founded.
The numbers seem to indicate that Liverpool concede more on target shots than average, block less and Reina is below average when dealing with the shots that are allowed.
For more articles relating to shooting efficiency, especially in relation to Stoke City, search my blog.
Read more of Mark's work on his The Power Of Goals blog
And follow Mark of Twitter: @MarkTaylor0