Stoke’s GK coach explains difficulty of using stats to assess goalkeepers

In Depth

We don’t need statistics to inform us that Claudio Bravo has struggled during his debut season in the Premier League.

Pep Guardiola described the former Barcelona goalkeeper as “amazing” after last week’s FA Cup win over Huddersfield Town, but demoting him to the role of Manchester City’s domestic cup keeper is a pretty clear sign that he too has been less than impressed with some of the Chilean’s recent performances.

In an effort to illustrate Bravo’s struggles, the Daily Mail last week published an article which led on the statistic that he has conceded from the first shot on target he has faced in 13 of his 25 City appearances so far.

They could have sexed up the headline even further by focusing on Bravo’s last 11 appearances, in which he has been been beaten on all but three occasions by the opposition’s first accurate attempt.

• • • •

• • • •

But what does that information actually tell us? In isolation, very little. Perhaps nothing.

Take a look at the 13 goals highlighted and you will probably conclude that Bravo was culpable for two: Zlatan Ibrahimovic’s strike just before half-time in the Manchester derby – Bravo’s debut – and Burnley’s reply in City’s 2-1 home win in January.

Even then, neither goal was consequence of failing to deal with a shot on target. Both came from high balls into the box that Bravo could not gather or punch clear.

The goals also range from arriving anywhere between three and 70 minutes.

• • • •

• • • •

Six of the 13 came in the opening 10 minutes, but Bravo as an individual appears blameless on all of those.

As a unit, City’s defence certainly isn’t, and the numbers might provoke Guardiola and his backroom staff to further investigate why they seem to make quite so many individual errors or mistakes as a group in the early stages.

You might look at how Bravo is organising his back line, but you certainly can’t apportion personal blame towards the keeper.

The flaws with this particular statistic sum up the problems with attempting to assess goalkeepers by data. Without the proper context, the numbers are almost useless.

• • • •

READ: Formations: Are they worth the whiteboard they’re written on?

• • • •

The position of goalkeeper and what is required from those who occupy it is probably the least understood by fans and pundits. So in the absence of technical know-how, the simplest way to judge is from statistics. Goalkeeping by numbers.

The oldest and most accessible metric is clean sheets. A keeper who doesn’t concede goals must be doing his job right. Right? Then came save percentages, followed by more and more nuanced save percentages.

Prior to the weekend, of the keepers who have 10 or more Premier League appearances to their name this term, the five with the lowest save success rates according to whoscored.com are Bravo (53.4%), Fraser Forster (55.4%), Maarten Stekelenburg (57.6%), Artur Boruc (59.1%) and Lukasz Fabianski (59.5%).

• • • •

• • • •

But it would be wrong to draw any conclusions without at least first scratching the surface to find out why these keepers have apparently under-achieved – if, indeed, they have.

These save success rates consider none of the variables a goalkeeper must contend with in almost every action they perform as a shot-stopper, but still they are still presented as a method of appraisal. From what distance is the attempt made? From what angle? Opposed or unopposed?

Adding such context to goalkeeping statistics continues to keep the number crunchers awake at night.

• • • •

• • • •

One improved model comes in the form of expected goals (xG) which assesses the number of shots on target allowed in the context of the quality of the chance. The variables considered include the distance and angle from shot came from; the method of assist; the type of shot and the play that led to the chance being created.

What it doesn’t account for is the position of the goalkeeper or his defenders, or who was taking the shot – deadly centre-forward or lumbering centre-back?

Goalkeeping statistics should certainly be taken with a fistful of salt, especially when presented in isolation.

That does not mean, however, that such data is dismissed by those within the game.

• • • •

READ: Why is it taking football so long to embrace sports psychology?

• • • •

Andy Quy, first-team goalkeeping coach at Stoke City for the last 10 years, will use the statistics to see if there is anything the naked eye might be missing.

“I won’t look at the stats on a day-to-day basis,” Quy told us. “But over a period of time, perhaps six to 10 games, we might try to build a picture, whether that be individually with the goalkeepers or within units of the team.

“Defensively, we’ll see what is happening in the games and break those down into what areas of the pitch the shots have come from. Crosses, too.

“But you have to give yourself a period of time to assess what those numbers mean. It’s difficult from a goalkeeping perspective to read too much into one game but we might identify a trend over a period of time, whether it be a strength or a weakness.”

• • • •

• • • •

Quy says his keepers, who currently include Jack Butland, Shay Given and Lee Grant, are “very receptive” to listening to what the data may be telling them.

For example, over recent seasons, statistics suggested goalkeepers were having more success saving shots from distance by staying closer to their line, therefore allowing more reaction time, rather than moving down the line of the ball to narrow the angles the shooter has to aim at.

“Playing deeper in the goal for shots around the 18-yard box – that’s something we’ve changed a lot over the last two years,” Quy says. “We have to build up the data over a period of time and make it work to our advantage, but that’s definitely an area in which which the stats have made a difference to our methods.

“It’s something Shay has looked at in the two years he’s been here and he’s altered his start position accordingly, which demonstrates that you’re never too old to learn.”

• • • •

• • • •

Performance analysis is not the only area of Quy’s role where data proves useful.

“I can’t get to all the games that I need to see everyone, so there’s a definite use for the statistics within the recruitment process. We can look at certain criteria that we’d want,

“Height was an attribute that has often been put forward by the managers I’ve worked under. Tony Pulis wanted a goalkeeper that was of a good size, someone with a good strike who could kick the ball a decent distance but with good accuracy. That’s obviously something quite easy to measure.

• • • •

• • • •

“Once we’ve got that raw data, we break it down and put it together with what we might see on a video.

“That process might sometimes throw up the odd gem or anomaly that might not be at a fashionable club or in a fashionable country.

“We’ve got a good team of scouts and they feed back keepers that they’ve seen wherever they might be. But we can’t cover every game everywhere so there’s a good case for using statistics to start that process.”

• • • •

READ: Peter Crouch in the 100 Club: A hero on and off the football pitch

• • • •

It is not just goalkeeping but the defensive side of the game as a whole in which the analytics community is looking for more reliable and accurate metrics. Defensive actions are generally reactive, whereas those in possession control and dictate the play, therefore making their contribution easier to record and measure.

As Guardiola’s desire to recruit Bravo and release Joe Hart illustrates, the attacking side of goalkeeping is more of a consideration for managers than ever before. This is one area in which Quy mines the data to measure how his keepers are working within the team.

“At Stoke, the priority has always been that we need a keeper who will keep the ball out of the net. But now the game dictates that we need to be better at building play up and be more confident on the ball. I think we’ve definitely gone that way.

• • • •

• • • •

“The team that we had when we were first promoted, there’s no doubt about it, we were a direct team. But we were successful doing that. We had players in the side and goalkeepers that could put the ball into good areas where we could be successful.

“Over the last three or four years, with Asmir Begovic, Jack Butland and Lee Grant, we’ve been more confortable on the ball at the back. With the defenders we’ve got, we’ve been able to play a more possession-based game and move the ball more comfortably.

“We look at the videos every day, constantly analysing training – are there things that can be improved or changed?

• • • •

• • • •

“Post-game also, we’ll sit down with the keepers for probably 45 minutes, seeing how it played out for them. We’ll look at some of the numbers – how many contacts with the ball, long kicks, short passes, whether or not they were successful – but, crucially, we have to put that in context of the game.”

And that’s the key to goalkeeping analysis – context.

It might not be convenient when so many use only a headline or a shareable graphic before rushing to their judgement, but the art of goalkeeping is far too nuanced and sophisticated to be assessed by numbers alone, even when they appear as alarming as Bravo’s.

By Ian Watson

Promoted Stories