In the last few years football analytics has made inroads into the media, with expected goals featuring on Match of the Day, and Sky Sports regularly featuring more in depth statistics when previewing and reviewing matches. With clubs increasing their focus on analytics and data science there is no doubt that interest and appreciation of the usefulness of data has never been higher.
The final frontier for analytics is coaching. Do coaches value the data? Does it influence their coaching? Is it only useful at the elite level?
To answer these questions Chris Summersell, a UEFA B coaching licence holder and English FA coach educator, joined me for a chat.
Hi Chris, tell us a bit about yourself, what is your footballing background?
Hi! I’m a football coach, coach tutor and analyst based in London.
I’ve been coaching now for about 7 years and hold a UEFA B Licence. I’ve recently finished a job working for a Premier League club development centre, and also still coach a grassroots team.
Recently I became a tutor with the English FA, tutoring the Level 1 coaching qualification and soon will be tutoring the Level 2.
As well as this I am an analyst for Total Football Analysis.
Playing wise, I was reasonable without playing at a level of any note, but importantly it was an experience that has shaped my journey in football. I have also followed Spurs home and away as a fan since I was 7 years old.
You could say I have grown up in a very ‘traditional’ football environment. When I played I was always in a 4-4-2 as one of the strikers but I spent half the time dropping in to midfield to collect the ball only for it to sail over my head down the channels. I spent almost my entire time playing the game fighting against bad ‘coaches’ who only knew one way of doing things. This experience led me into coaching, where I wanted to do things differently.
Latterly I have become extremely interested in the growing influence of data and analytics within the game. It has helped me understand the game a great deal better, but I can’t help think a lot of it exists inside an echo chamber – as a coach first and foremost, I’m interested in how we take the knowledge off of the laptops and onto the grass.
The echo chamber bit is really interesting. Do you find a willingness among the coaches you work with to embrace data? I know a lot of analysts would love to work with coaches to see what they actually want from data, and to chat about what is and isn’t possible with various types of data.
By echo chamber I mean that there is some absolutely astonishing work done out there with data, but I wonder how much this is really influencing coaches throughout the game, other than a notable few exceptions.
I sometimes think analysts need to read what they’ve written back to themselves and ask, ‘what is a coach going to think of this?’. It’s not about dumbing things down, it’s communicating in a language that coaches speak themselves. I too fall into this trap.
This also works both ways, and it frustrates me hearing coaches dismiss data and analytics so readily.
I must stress that I work predominantly within youth football, so my interaction with coaches is generally within this area. I do not see a huge deal of interest in data and what information it can provide at all. It is a common refrain to hear coaches say ‘I don’t mind statistics, but…’ – this is in my experience generally code for ‘I agree with all statistics that support my pre-existing views’.
One view of mine is that the insight and understanding that data can provide can be used at all levels of the game. No matter what level or age group you work with, concepts such as xG, packing data etc can help us understand the game more, make better decisions and ultimately help individuals and teams develop and improve. This isn’t saying you should be hammering your U12s for shooting from low value locations, but more how can you work with them to understand how they can create shots from better situations.
The other day I read on twitter that Dele shows up very highly on his pressing data, for both England and Tottenham. Why as a coach wouldn’t you then go and watch him closer to see what exactly what he does that makes him this good? If you watch him and disagree with the numbers then challenge it. This is how the game develops – if you put coaches and analysts together on the same page, then the potential is limitless.
Ultimately it’s a two way street and I don’t think either coaches or analysts have quite yet worked out how best to complement each other.
I have to try and remember that before I wrote about football I used to write a lot about how bad use of data – produced by people with no on the ground knowledge of the sector – was terrible for the school system. Just because Eton gets great exam results it doesn’t follow that teaching kids in the lowest performing school in exactly the same way, wearing the same uniform would get exactly the same results. Context matters.
It is the same in football. I can look at Liverpool or Manchester City and then say “I want Cambridge United to play like that” but the reason that those two can play like that is they have some of the best technical players in the world and a huge team of staff catering to their every need. Cambridge won’t ever have that but you can look at Liverpool’s pressing or City’s interplay and use elements of it to improve your own game.
Data cynics would say “Yes that is called video analysis and good coaching” but where I think data goes further is that you can pick out teams / players at similar skill levels who are achieving the results you want to. Much like a normal secondary school in a poor area might be far better learning from a school with similar intake with higher levels of achievement rather than Eton. You can then use the data to see if what you are doing is actually working. Are you pressing in the right areas? If not why not?
So if you were going to try and bring in analytics informed coaching at a lower league or junior level what type of data do you think would be most useful both for individual player development and for the team as a whole?
I like that schools analogy! Yes I agree context is very much key – and I think mainstream audiences are still subject to statistics that are meaningless like distance run/number of passes etc. This does have an effect, I think, and doesn’t help the cause of advanced analytics gaining a foothold. It’s obviously more predominant now, but still seen with great scepticism within football circles.
Speaking of context, I remember not so long ago there was a story that when he was Man United manager David Moyes told the players that he expected them to make at least 600 passes in a game. It was like saying in order to be a successful club you need to buy a large trophy cabinet, because Barca and Madrid have one.
I think at the top level we’re starting to get a lot smarter than that now, but it wouldn’t surprise me if this still goes on as you drop down the divisions. Also, just because a club has an analytics department and smart people, it doesn’t mean this translates to the pitch if you don’t have an open line between data and coaching staff.
In terms of using data at lower league and youth level, firstly I think coaches need to understand simply what data tells us about the game – they don’t have to be data experts themselves. Most obviously is xG, and at the fundamental level it is telling us to better think about where we shoot from – this is something that coaches need to know and plan for when working with players.
At non-league level you are not likely to have access to xG data, but knowledge of the concept can help. I hear people all the time say ‘he’s got to score from there’ etc – often these are really difficult chances. In the absence of xG data it isn’t too difficult to record where your team takes and concedes shots from each game, you can divide it between six yard box, penalty area, outside the box etc. Then you have some form of basic notational data to work from, understand your players, and then come up with your individual and collective coaching plans.
All you need as a lower league coach is a set of flat discs to mark out danger zones to shoot (or deny shots) from and you can start training the idea of generating good shots.
Set pieces are another completely under appreciated aspect of the game, which is funny because in UK football crossing and heading is so much a part of the game. I really like what the guys at Statsbomb are doing with their set piece work, and any club that can afford them should be banging down their door trying to hire them. There is so much to explore in set pieces, and I think any club – from Premier League to non-league should be looking how to gain advantages in this area. If they’re not, then they’re doing something wrong.
With more limited time at Non league level on the training ground, then a playbook of set piece routines can mean you’re adding X amount of goals a season which can be massive.
Individually, I’d encourage coaches to consume as much information from analytics as possible. Find out what players show up high on certain metrics, see what they do differently, why are they showing up so well on these numbers and then try and relate it to your own coaching, and see if it works.
I also think that with data only going to become more of a central part of the game, National FAs may look to run their own courses on data in football, and its most certainly should be an aspect that is included in the UEFA B/A licence courses. If you want to work at that level, understanding how to apply data driven insight surely has to be something you are exposed to as part of your coach education.
Where do you stand on pragmatism versus “philosophy” management. I generally think recruitment works best if you are trying to implement a style and can recruit accordingly but I have a huge admiration for managers who can find a way to squeeze out points and adapt their game to counter the opposition. I also think data can help massively with creating game plans.
I like this question. It’s something I think about a lot – I think in an ideal world you have a joined up philosophy and style of play which in theory should make recruitment simpler in terms of identifying players who match the profiles of each position.
It was these clubs that got me into coaching in the first place, watching that Barca team for instance was obviously a huge inspiration and like a lot of others became insistent that this was the ‘only way to play’. I’m glad I don’t think like this anymore, and I have so much admiration for coaches who are not wedded to a certain style and adapt to their circumstances (Benitez, Ancelotti), or even those whose style is seen as regressive but continually get results like Tony Pulis and Sean Dyche. I remember a few years ago I listened to Pulis talking about using the correct body-shape to defend crosses and I realised there was a whole load of detail behind how they do things. It might not be pretty, but it often gets results, often in very trying circumstances. I’d love to see how these guys recruit players, whether they use analytics, and what analytics could do to recruit players to fit their styles.
This season Fulham’s story may serve as a cautionary tale in the future as they came to play expansively and have been dumped unceremoniously – I’m really interested to see how Norwich and Sheff United get on next season in the Premier League. If Leeds go up, we know Bielsa won’t change his style so that will be really fun to watch but as much as I love how his teams play, my gut says they would be chewed up and spat out again. They’re already overperforming relative to player quality in the Championship, which is testament to his brilliance as a coach.
Going back to the question, one thing that interests me in terms of using data to recruit players based on style/profiles is how does a team with a defined style identify and recruit players that play for a team that play in a completely different manner. I know there are adjusted stats which have tried to combat this to a degree, but I do sometimes see player radars and articles about players and I’m there wondering – is this metric like this because the player in question is simply following the coaches instructions, or simply isn’t able to demonstrate their full skillset because of the way their team plays? I guess this goes back to getting analytics and coaching experts working together better.
I think a really interesting case study into ‘philosophy vs pragmatism’ is at my club, Spurs. Many people would have Pochettino down as a ‘philosophy’ type coach but I think it’s more complex than that, and this season I think we’ve really seen him be adaptable in squeezing every single ounce out of this current squad.
When he took over we were absolutely wedded to a certain style of play, very rarely changed shape (4231 for the first two seasons) and set-up faithfully to press and possess against everyone. Even when we were individually inferior to the opponent, we’d rush off head first into the fire and get burned to death. At the time I said it was a price to pay in implementing his style of play, where he had a young committed bunch of players willing to learn, but we’ve really stepped it up a notch now in terms of adapting to opponents, and to the state of the game, that has meant we are now playing in a Champions League Semifinal.
One such player that has become integral to our season is Moussa Sissoko. This signing was roundly panned by the analytics community – and I’m not saying they were wrong at the time – and not long after that Poch came out and praised Sissoko for how he helped us deal with transitions. We now play more transitional type games, and Sissoko has more than come into his own. Obviously my take on this is that Poch is a magician.
I agree that analytics can really help with gameplans, and I think having data to work with can really increase your chances of spotting weaknesses to explore in your opponents. Let’s face it, even a team of analysts watching videos of their upcoming opposition a week before the game isn’t going to pick up every small detail. I’d want to use data to look at how opponent players tend to react in certain scenarios. Let’s say an opposition full-back – you can use pressure event data (if you have it – I don’t!) to see how that player tends to deal with the ball, where they turnover the ball under pressure. Do they go long, turn backwards, play the ball blindly inside? Use this data with video analysis and you’ve got a data informed pressing trap. I’m certain this level of detail does go on, clubs aren’t that likely to divulge what they do, but I’m also pretty sure lots of clubs leave lots of things to chance, or the quality of the eye test. Data can help.
It may well be happening already, but it struck me that as soon as Spurs drew Ajax (as an example), how much would it cost them to go out and hire a number of analysts to go and extensively scout each Ajax player, have a real deep dive into the data and report back with areas that you can exploit, or must stop? This relatively small investment could mean you find that one weakness that gets you into the Champions League final.