THE SPORTS DATA REVOLUTION
‘Data’ is one of the most contested terms in sport. In this discussion, Nathan Leamon and Tim Wigmore – two of the expert lecturers on the Leadership in Sport MA at the Institute of Sports Humanities – reveal how sports data informs high performance today.
Nathan is senior data scientist for the ECB and author of Hitting Against the Spin (with Ben Jones). Tim is a writer for the Telegraph and author of Crickonomics (with Stefan Szymanski).
Is the whole sports data revolution exaggerated? Simon Kuper ended his review of Rory Smith’s recent book on data in football by saying, basically, that the data revolution in football will be interesting…it just hasn’t happened yet.
There has been an arms race in professional sport about the next ‘edge’. But employing people doesn’t guarantee they have influence. How much influence does the analytics department really have?
Nathan: No, I don’t think so. I’ve recently done a tour of different sports looking for insights and ideas we can apply in cricket, and I think in the last ten years the application of data has been transformative to the thinking in a large majority of sports.
In general, two forces fight each other in this regard. The complexity of the sport, and the resources you have available. Applying data is easier and more effective in fairly simple, linear sports. One-dimensional time and distance sports lend themselves to it. (Try running a track cycling team without data and you’re not winning many races).
On the other hand, the impact you can have with data is limited by the resources you have to hand to collect, collate and analyse it. Each sport has a different set of coordinates on that 2D plot – Simplicity v Resources.
Football is probably the most complex sport to analyse, and despite its considerable resources, still the sport where data has had the least tangible impact.
Tim: Of course, the impact varies significantly between and within sports, but zoom out and data’s impact on modern sport is significant. The whole sport of basketball has been transformed by the three-point revolution, which is founded on a simple calculation: three-point shot attempts are less reliable than two-point shot attempts, but, because they earn more points, they have a higher expected return.
In English football, the two clubs who overperform their spending most consistently are Brighton and Brentford: two teams who use data exhaustively to underpin their recruitment. At a macro level, the average distance of shots has decreased in recent years – reflecting that such shots have a very low chance of going in (as shown by expected goals metrics) so it normally makes more sense to try and pass to someone closer, even allowing for the chance of such a pass being intercepted.
The use of expected goals in the media – even on Match of the Day – has also had an impact in changing the discussions around sport. There’s more acknowledgement of the huge role of luck on a game-to-game basis. Brentford once removed a manger who had come fifth in the Championship (ostensibly a very impressive result) because their ‘table of justice’ – a table of what the league ‘should’ have looked like, based on underlying performance and expected goals – showed that they had been lucky.
Most teams don’t seem to use it as much, and it’s notable that most Premier League boards – again Brighton and Brentford are notable exceptions – continue to ignore the finding that, on average, sacking a manager makes no difference, and clubs would be better off saving money spent on hiring and firing to invest in players and facilities instead.
In general, the impact of data is easier to see at the level of recruitment than in-game. Data is more obvious as a tool when it comes to evaluating how good a player is than on the optimal strategy to adopt on a particular day – which doesn’t just depend on the players involved, but form, the state of the game and the conditions.
To what extent does being an effective data analyst rely on personal relationships with key decision-makers?
Nathan: At the first level of working, I think it comes down to trust. And although the shortest, surest route to that is having a strong personal relationship with someone, it isn’t the only way. If someone trusts me, then I can have an impact on their thinking.
But that can also happen if they trust the data I’m standing in front of, or the person I’m standing next to who is vouching for me. You can show someone something that they didn’t already know if, and only if, they believe what you’re showing them. Trust is the necessary and sufficient condition.
At the deeper level of working, yes, I think it must be an established relationship over a period of time. Because it becomes a conversation, an interplay of ideas. Analysis done well is a two-way process. The players can see and intuit things that the laptop can’t, and vice-versa. The bouncing of ideas back and forth from players to data and back again is where the best, most impactful insights happen.
“Football is probably the most complex sport to analyse, and despite its considerable resources, still the sport where data has had the least tangible impact.”
- NATHAN LEAMON
Do you think the sports media – in approximate terms – under-estimates or over-estimates the influence of data in sport?
Tim: This is hard to generalise because the tendency works both ways, which often exaggerates divides in attitudes to data between sides. Teams who present themselves as ‘old school’ are often thought to have abandoned data completely, when the reality is more nuanced: to a lesser or greater degree, data is integrated into teams’ thinking. And, when it is being used openly, there is certainly a tendency to either present data as a panacea or the cause of a team’s problems – even when the truth is almost always somewhere in the middle.
For instance, there’s never been as much interest in a managerial reign at Southampton as Nathan Jones’s last season. Jones was hired, Southampton said, largely based on data – and how his sides overperformed their expected results, based on their relative spending. When it went wrong very quickly, this use of data was blamed – probably rightly, in the individual case, but also often to attack clubs using data more broadly, thereby ignoring all the instances when it has helped. Sometimes it can feel as if every data-informed choice is depicted as a referendum on data in sport.
I referenced Brighton and Brentford earlier. It’s notable that they remain relatively resistant to giving the media access to their inner workings: clearly, they still believe that they have secrets to protect.
Do you think data is more central/useful within tactics (e.g. match-ups) or strategy (e.g. recruitment)?
At a push, if you had to choose between the two and they were mutually exclusive, would the data department have an ‘in’ to the dressing room, or sit independent from it?
Nathan: This depends entirely on circumstance. If you are in a relatively free marketplace for talent, then the most direct impact is strategic. Putting a better price tag on effectiveness, exploiting inefficiencies in the market for players is normally a bigger edge.
However, in the absence of a market for talent – if you work with a national team, for example – then it becomes a question of time scale. The spending of scarce resources to develop and deploy talent is still the most important part, but it is the aggregate of decisions taken over up to a decade. In the short-term, smart tactical use of the available playing strength is more impactful.
To what extent can sports learn from each other in how to use data?
Tim: Sports have always learned from each other. Historically, it has been US sports that have led the way. The Moneyball story was only possible because of the amount of money in Major League Baseball and the maturity of the recruitment market; its impact was also multiplied because of the cultural reach of the US.
The most common area, perhaps, relates to attitudes to risk. One of the common findings across almost all sport is that teams are better-served by playing in what is perceived to be a riskier, more attacking way. Football sides who are losing have more chance of getting a win or a draw when they make substitutions earlier. In ice hockey, losing teams who ‘pull the goalkeeper’ – replace their goalkeeper with an extra attacker – also have more chance of drawing level. Basketball teams do better when they take more riskier shots (3-pointers). In T20 cricket, teams who lose early wickets do better when they continue to be more aggressive, rather than consolidating for long periods of time.
These examples shows that the argument ‘data stultifies players’ is simplistic. Properly used, data can be a way to liberate them – creating a culture that understands that playing more aggressively is generally the ‘smart’ thing to do too.
What does sports data look like in 10 years’ time?
Tim: In ten years’ time I strongly suspect that the direction of travel will continue. Data will be more integrated in teams and probably a little more a part of media coverage – in general, data is covered more in US sports media than in the UK. The coaching staff of tomorrow will have been more accustomed to using data in their own playing careers, which will have a bearing on how much it is used. Non-playing experience might well be regarded as less important for those who evaluate and sign players, though sports teams are much too complicated to be run by algorithm.
One prediction in Moneyball (published 20 years ago) was that the number of scouts in Major League Baseball would dwindle; in fact, the number of scouts have increased. So, for all the temptation to construct oppositional stories around how data is used, the successful teams will be able to integrate it in a way that complements, rather than replaces, other skills.
Source: SIG
The Institute of Sports Humanities and Loughborough University London have partnered to deliver The Leadership in Sport MA, a course for sports industry professionals to study alongside their full-time jobs.
The course starts this autumn and applications are now open.
Find out more at www.sportshumanities.org