Accident Analysis for Goalkeepers.
Just like with goalkeepers, airplane pilots are often the first ones blamed for accidents, even when they might not be at fault. And although placing the blame like this might be easy, it does not help much in explaining why, or even if, the pilot / goalkeeper made an error. Without a proper explanation, it is impossible to understand why something happened and without this understanding there is nothing to prevent the error from happening again.
The Swiss Cheese Model
In the mid-nineties, as part of our training, pilots were taught about a risk-analysis model called the “The Swiss Cheese” model or the “cumulative act effect”. This model, which was originally constructed by Dante Orlandella and James T. Reason of the University of Manchester is sometimes deemed to be overly simplistic, but it is a very useful model for basic analysis.
The model states that most accidents happen because of the cumulative result of different acts. Imagine a couple of slices of Swiss cheese. Each slice will have air holes in it, but they are in different positions and of different size in each slice. If you put the slices together to form a block of cheese, some of the holes might line up, but it is not very likely that all the holes line up from the front of the block to the back. In the model, an accident occurs on the rare occasion that all the holes DO line up. Identifying what caused these holes to line up allows you to take action in order to prevent it from happening again.
In aviation each “slice of cheese” represents a barrier. One slice would represent for instance airplane systems, another standard operating procedures, yet another one external factors like weather or Air Traffic Control with the last slice being the pilot. Any of these barriers working properly would prevent a situation from deteriorating. For an accident to happen, all these barriers would have to fail, at least partially, lining up the holes in the Swiss cheese. Of course there are occasional exceptions where the situation has deteriorated to the point that there is nothing the “last slice of cheese” (the pilot) can do to prevent an accident from happening, or where everything works as advertised, only for the pilot to make a huge error.
Cumulative Acts leading to Goals
It’s the same with goalkeeping. If a defensive error leaves the goalkeeper alone facing two strikers, there is very little he can do to prevent a goal, other than covering the short corner and hoping that the striker will take the shot rather than pass it off to the side to the other striker. Alternatively, the team can do everything right only for the keeper to make a “howler” of a mistake leading to a goal.
Most cases aren’t this simple. Goalkeeping especially remains an area in which the available statistics often do not tell the whole story so analytical skills are required for coaches at any level. Game stats do not necessarily show where the first error is made, starting the domino effect that leads to giving up a goal. OptaPro has an “errors to leading to goal” stats category, which is “allocated when there is a glaring error that leads directly to a goal. It normally applies to players that have miskicked the ball into the path of the opposition or misplaced a pass straight to an attacker, but only during the phase of play that leads directly to the goal and would have to be very obvious” (Thank you, Simon at OptaPro for taking the time to explain it to me) As a result, in almost every game at every level there are situations leading to goals where an analysis using the “Swiss Cheese” model can be of great use.
Sample Analysis – High School level
A few seasons ago, we gave up a goal and in watching the highlight tape it showed the opposing striker chipping our goalkeeper, who was at about the penalty spot, from approximately thirty yards out. It would seem like a clear-cut case of the goalkeeper being out of position. Until you rewind the tape a bit further and take a closer look at where the action leading to the goal started.
Our goalkeeper came out to intercept a through-pass in the box on his left hand side a few yards inside the 18-yard line. He then ran across to the top of the box on the other side of the goal and threw the ball to our right full back. That took a second or two. The ball bounced high, and the fullback had a difficult time controlling the ball while also coming under pressure from the opposition’s midfielder. The fullback got a weak, high pass off to our central defender who couldn’t immediately clear it. The ball bounced in front of the other team’s striker who didn’t hesitate and took a shot on goal, which went over the keeper who was still backpedaling from the top of the 18.
Had the back been able to control the ball, or the central defender been able to clear it, there would not have been a goal. These were “contributing factors” but the real mistake happened before that: By running from the one side of the box to the other before throwing the ball to the fullback, the goalie gave the opponent’s midfielder time to close the gap and put pressure on the defender. It wasn’t necessary for the goalkeeper to do so, because his arm strength was enough to get the ball to the defender from further out, as evidenced by the fact that he overthrew the ball, which was the main reason the defender could not handle the ball to begin with.
So using the “Swiss Cheese” model allowed us to determine that the most important thing to fix in this situation was the goalkeeper’s awareness in regard to restarts, not necessarily his positioning skills. By throwing the ball earlier, he would not have signaled his intentions to the midfielder. It also would have given him the chance to let the ball bounce earlier and stay low, making it easier for the defender to handle the ball. This simple action could have been enough to prevent the situation from escalating the way it did. But even if it had escalated, the keeper would have been in a better position to get back to his line quickly, minimizing the possibility of getting chipped. This one action would have made the holes in every slice of the Swiss cheese smaller.
Sample Analysis – Professional Level
Things like this happen in the professional game as well. For instance, in their 6-1 loss to Man City, Tottenham’s Hugo Lloris made a couple of great initial saves only for the rebounds to be converted. But this doesn’t tell the whole story.
In the first goal, a backpass was poorly cleared by Lloris and retrieved by City. The ball came to Negredo who had enough time to take a shot. Lloris made the save but palmed the ball back into play, which was tapped in by Navas. There are “contributing factors” here, like the Spurs defense failing to put pressure on the City players on several occasions, but the cycle leading to the goal started with Lloris’ failure to properly clear a backpass.
The second goal was very similar, a bad clearance by Lloris caused a loss of possession, insufficient pressure by the Spurs defense allowed City to counter and take a shot on goal, leading to a palmed save by Lloris falling to a City player who scored.
Plenty of blame to go around on both of those goals and a better performance by various Spurs’ players in any of those areas would have prevented these two goals. But using the Swiss Cheese theory, two things tend to stand out to a goalkeeper coach: poor foot clearances by Lloris followed by him allowing rebounds to fall back into play.
I have only watched a limited amount of Spurs football this season and I don’t have any empirical data to back it up but from my observations it seems like Lloris gives up more rebounds back into play than the other EPL goalies. If anybody has any data on that, I’d love to see it.
The other problem area suggested by the analysis clearing the ball which could be particularly concerning for a keeper that likes to come off his line. Of course, the example is only one match so these two errors could just be a fluke. But historical data might suggest otherwise.
In an article in the Guardian a few weeks back, Jonathan Wilson wrote that Lloris had “successfully swept up behind his defence 33 times, more than anybody apart from Newcastle’s Tim Krul (39).” He also noted that Lloris’ pass completion rate was only 42.5%. I don’t know how that pass rate compares to the other goalkeepers in the Premier League, but in this week’s Bayern-Arsenal game, Manuel Neuer completed 14 of 15 passes in the first half alone for a completion rate of 93.3%. (http://www.theguardian.com/football/blog/2014/feb/13/tottenham-hugo-lloris-premier-league-sweeper-keeper)
To me, it seems that if you are going to sweep up behind your defense, your pass completion rate should be (much) higher than average for two reasons. First and foremost, a keeper successfully coming off the line to intercept a through ball leads to a change of possession, which gives his team a great opportunity to start a counter attack. But when you are playing that far ahead of your goal, you are going to be out of position to make a save, if required to do so. Therefore, not giving the ball away, especially in your own half is crucial. Blindly banging the ball long or out of bounds gives a keeper time to get back in position, but at the expense of losing the possession you just gained, and should therefore be avoided whenever possible. That’s why it is so important for modern goalkeepers to have the ability to make an accurate pass, both long and short.
An interesting stat to track would be how many of a goalkeeper’s inaccurate passes or clearances lead to loss of possession in the team’s own half, leading to shots by the opposing team
These examples are not meant to pick on Hugo Lloris. I think he’s a solid goalkeeper who has not yet realized his full potential. He has made his share of mistakes this season but some of them have come as a result of his willingness and courage to play the way the modern game requires. I’m convinced that improving his passing skills will greatly improve his effectiveness.
The examples above show that, regardless of playing level, the use an analytical model like the “cumulative act effect” can be very useful in interpreting keeper performance, identify probable weak areas, and fix them.