Big Data is changing football. Liverpool, Barcelona, Arsenal, Manchester City are among the teams that are already using data analytics to improve performance, analyze rivals, prevent injuries and optimize the transfer market management.
This trip allowed us to exchange ideas and learn from people doing data analytics in Liverpool, Arsenal, Milan, Leicester City, Seattle Sounders & Toronto FC, among other teams.
On behalf of these meetings, we would like to explain what we think is the main objective of applying data in a football team and list some of the best practices we think that are important get the most out of it.
Big Data in football clubs
The main objective of using data analytics is to obtain additional information in the decision-making process.
This type of analysis allows teams to save time and their conclusions can be used to generate new questions about hidden aspects or as an objective source to confirm their thoughts, intuitions & diagnosis.
Even though incorporating Big Data in a football team can have a lot of advantages, using data in an incorrect way can easily lead us to inaccurate conclusions and decisions. Let’s now look at some considerations when using data analytics to prevent this from happening.
1. Use data analytics as a complementary tool
Considering data as a piece of additional information in the decision-making process, it is necessary to complement it with other analysis for it to be such a powerful tool. This means that data analytics conclusions should not be the only source of information when making a decision.
Even though data can help us to find new players, nobody is going to sign a player by only looking at their numbers.
2. Add context to data
Data by itself doesn’t mean anything. To obtain sensible conclusions, it is important to always take into account the context.
On one hand, benchmarks can help us to evaluate if some value is high or low. Depending on the objective of the analysis, this comparison can be made with historical values (for the same player/team) or with other comparable players/teams’ numbers.
For example, we don’t get too much information from a player that has covered 10km in a game. But if then we see that that same player covers an average of 14km per game, we can understand the value better.
On the other hand, it is key to maintain the football vision and not being dragged by mathematics. All data analysis will depend on the team’s tactics and strategies, so we have to take into consideration how these affect evaluation metrics.
It would be a mistake, for example, using passing percentages to compare defenders from teams with a passing game strategy with others that play with a more direct and long passing style.
3. Differentiate the important stuff over the interesting one
Football is full of statistics: goals, shots, passes, tackles, etc. The challenge is to know which ones are really useful to evaluate and make decisions and which ones are nothing more than a number.
On one hand, we have to pick metrics that are good indicators of what we are trying to measure. This implies, for example, to use Expected Goals (xG) as an indicator of offensive & defensive performance instead of just summing the number of shots made/received.
But on the other hand, the selected metrics must be aligned with the team’s playing style. Take the case of the number of rival passes in their defensive zone is not an important number if our team waits for rivals to pass through the half of the pitch to start pressing.
It also happens that some metrics will be important in some analyses but not in others. This is the case of the distance covered. For the physical aspect, this metric is key to preventing fatigue and injuries. But in the tactics aspect, it doesn’t give us too much information as we don’t know about accelerations, directions or locations in the field for example.
4. Communicate with simple visualizations
This aspect is key for data analytics to be used and applied correctly. We can have the best data analysis, but it comes to nothing if we cannot transmit effectively our ideas to directives or coaches.
Conclusions obtained from data must be communicated in a language so that any person can understand and use them in the decision-making process. In many cases, it is preferable to sacrifice some model accuracy to be able to communicate and explain how are the metrics built.
Visualizations take a big role in this aspect. They should be the most graphic and as simple as possible. The objective here is to show evidence of the main idea without confusing people with too much information.
For example, we can see the next graph showing the defensive intensity in an easy way to understand.
5. Know data analytics limitations
To be able to use data analytics in the decision-making process, it is not enough to know its benefits, but it’s important for us to know their limitations and weaknesses.
We have to take into account that, even though mathematical models can approximate most of the things, they will never be 100% exact.
For example, if we are using the Expected Goals metric to evaluate offensive and defensive performances, we must know that this model only considers shots. This means that we are not going to take into account dangerous attacks that, for any reason, were not finished by a shot.
Another information that is hard to detect, in the way that data is collected nowadays, is the body orientation. So if we analyze passing opportunities in a moment of the game, it could be that some of those passes were impossible to make due to the player’s body orientation.
In summary, for data analytics to be a powerful tool in football, we must use it as a complement to other types of analysis.
Also, to be able to obtain actionable and relevant conclusions, we must select metrics that (1) are good indicators of what we are trying to measure, (2) take into consideration the context and (3) are important to the team’s playing style. These conclusions must be communicated with simple visualizations, for any person in the team to be able to understand and use them in the decision-making process.
On the other hand, we must understand that mathematical models and data collection are not 100% exact and so we have to considerate their limitations when using them.
The data revolution is already happening in football and the best teams in the world are using it to make smart decisions.
If you’re interested in incorporating data analytics in your club, I invite you to write to us by mail to email@example.com or via Twitter.