Transforming data into information, knowledge & wisdom

Alex Marin Felices (pictured): I’ve been the Data Scientist at Nottingham Forest since early 2024, working mostly in recruitment. Outside of the club, I run a newsletter called The xG Football Club, where I summarise and get key insights from the newest research in football analytics.

For the purposes of this presentation, there is one particular piece of research, published in 2024, that I would like to share. In the study, the authors surveyed practitioners from federations that had qualified for the 2022 World Cup, as well as ones from a range of professional clubs around the world.

In total, there were respondents from 29 federations and 32 clubs. The idea was to explore the characteristics of the practitioners’ data analytics infrastructure, of their roles and also their value for elaborating player monitoring and positional data.

The findings were eye-opening. Only 58% of the respondents agreed that data was ready to generate decisions at their clubs or federations, and the consistency between departments regarding data was 36%.

So the problem is not the amount of data we have available, but what we do with it. As Matteo Matteotti mentioned in the previous talk, a single match with tracking data can generate three million data points.

We have match data – line-ups, who scored, red cards, etc. We have video of these matches or of training, from a tactical camera or from a broadcast feed or from a drone in training.

Then we have what has probably been the key of all football analytics up until now, which is event data. Of course, this only shows you what happens when one player has the ball though and ignores what happens in the rest of the pitch, when the players don’t have the ball.

This is why tracking data is now making an appearance and becoming more and more important, because it allows you to see what happens in every second for all players and the ball.

Then we have some physical and load data, that may be GPS, distance, speed, but also sleep quality, injury history of the players etc, and finally we have some contextual and off-the-pitch data, for example to analyse if a player is suitable for the culture of the club.

Some clubs are using AI or natural language processing to check news or social media to see if this player will be suitable for them.

We all want a data holistic approach, incorporating all of these different types of data, but there is a problem if this is not leading to actual useful decisions at clubs.

I got this image from Jan Van Haaren, defining data as a tool for wisdom. We need to gather data and transform it into information, knowledge and finally wisdom – namely insights that can be actually used.

So let’s use an example of centre-back recruitment as a simple high-level case for this ‘wisdom pyramid’. We are using tracking co-ordinates from the spatial temporal data, so these may be tackles, blocks, aerial duels.

The pool of players that come through in the data could be 1,000, which is is impossible for scouts. We can’t give a list of 1,000 players to scouts and tell them, ‘Look at the video and tell me which one is good.’

So how do we help them reduce this pool? We aggregate and summarise all the data to create some information. Now we don’t only have blocks, but we have blocks per 90, we have the aerial win rate, we have the areas of the pitch in which the interceptions happened.

This gives context. After all, it’s not the same if a player intercepts in the middle of the pitch with a high block, as if they intercept in the middle of the box on a low block. 

The next step is knowledge. We have all of these metrics, like expected goals, and then we need to think, ‘Okay, what does this really tell us?’ If we have a centre-back with data we like, we can assess what their league is like in terms of playing style and whether it is similar to ours, in terms of physicality, speed and so on.

Is what the player is doing in his team transferable to what we want to do in ours? Maybe the centre-back really likes to intercept in a high block, but we as a team play in a very low block – then his best skill may not be suitable for us.

Then we need to use video, because scouts have to check some of the things data cannot see. Is he regularly looking over his shoulder to check whether a player is coming? Is he a leader? Is he strong mentally? Does he bring his team-mates up when they start losing?

And then we can bring in some other kinds of data  – his contract situation, his lifestyle. If a player is very good, but there isn’t alignment on the contract situation, then it may be for nothing.

Once we have all of this – tactically we like the player, he has good physical attributes, on the financial side it’s doable and he fits the style or role we want him to play – then we can actually make a decision on whether or not to sign him.

In the end, this is what we want data to do. We don’t just want a nice dashboard to look at once a month, we want something that is driving decisions.

This is why you should combine data from different departments and from different sources to transform it into wisdom.

Gabriele Gnecco: In the last six years, since the inception of Hudl’s Customer Solutions team, we have been privileged to have tens of thousands of interactions with professional clubs across every major continent.

This outreach spans organisations of vastly different scales – from clubs operating with world-class budgets to those navigating tighter financial constraints; from organisations with deep technological expertise to those just beginning their data journeys.

Thanks to this extensive and diverse experience, we have been able to construct a clear Data Adoption Curve within professional football. This defines four precise and distinct phases in a club’s journey towards data maturity:

1. 1. Phase One is what we call outsourced analysis. The club uses raw data but relies entirely on external consultants or providers to perform and interpret the complex analytical task. In this situation, the internal capacity of the club is minimal and no extra staff are needed.
   2. In Phase Two, the analysis is internalised, while the engineering is kept external. So the club starts building internal analytical capabilities and hiring analysts, but the challenging foundational process of data engineering, cleaning, and transforming and storing data remains outsourced.
   3. In Phase Three, all data phases are internalised, from engineering to analysis but a major bottleneck persists, so different datasets are analysed in separate silos. Performance data is viewed independently of scouting ones and financial data is separate from medical reports and this prevents from the generation of cross-functional insights.
   4. Phase Four is the holistic approach. This is the destination, this is what we are talking about today. Data is centralised, unified and analysed, with an overall and comprehensive view. This phase is characterised by cross-functional collaboration driven by integrated data. It requires data engineers and a massive effort to map all the data services that are implemented in the cloud data pipeline.

We are very happy to report that over the years the average level of technological sophistication across the professional game has clearly risen and, most importantly, the number of clubs aiming to make that decisive final step to a holistic approach is growing exponentially.

The necessity for this unified approach stems directly from the sheer volume and variety of data that clubs now have to manage. They find themselves dealing with a true multitude of different data types.

In a high-tempo week, a top club can easily have to deal with more than 22 million distinct data points and accumulate more than 100 hours of video content per week. The current defining challenge is therefore not collecting the data, since this battle is largely won, but using all these disparate elements as a single cohesive source.

We must ask ourselves why this is important. Firstly, every single source contains value but, due to its nature and main purpose, it has some intrinsic gaps. Using different types of data within the same process significantly increases the possibility for analysis. The sum of the integrated data is greater than its parts.

For example, the integration of event and tracking data allows for the calculation of metrics such as line-breaking passes or analysing the space that is left to players at the moment they receive the ball.

And this of course is valuable both to analyse the next opponent, but also to evaluate our team’s capability of occupying the space and for our attackers to find free space to receive the ball.

Or, to provide another example, in order to minimise risk of taking bad decisions during recruitment, physical and performance data are not enough – they have to be analysed together with the reports coming from the scouts, including subjective capabilities like leadership or mentality that at the moment can’t be measured.

The ultimate phase in this holistic journey, which only the true pioneers are now undertaking – the integration of the raw data insights with video footage. This is the bridge between the technical analytical team and the decision makers on the grass.

Integrating videos makes the complex numerical insights dramatically simpler to understand. The less technical figures such as Sporting Directors, coaches and more importantly players are more easily influenced and improved through the immediate impact of images rather than by being presented with mere abstract data tables, statistics or dashboards.

Multiple scientific papers and research studies have confirmed that visual learning tied to quantitative metrics accelerates comprehension and facilitates behaviour change on the training ground.

By unifying data metrics and video, we transform analysis from an abstract report into an actionable visual feedback tool, unlocking the full potential of your data and consequently your team.

At Hudl, we have developed a solution called Insight that allows clubs to seamlessly and automatically integrate event and tracking data with video without manually mapping between different services or custom coding.

Jon Ollington: I’m going to finish by showing you a dashboard that’s built in Hudl Insight that leans on what we’ve heard from Gabri and Alex.

What we’re doing is combining the physical and tracking data with the event data and video. This is where we move on from simply integrating single-source data to actually understanding playing style and tactical behaviours.

What Insight does is take the raw information – so the video tracking feeds and event logs – and layers this into something more contextual. So instead of viewing metrics in isolation, we can start to answer questions about how teams actually behave.

I’ve divided this dashboard into playing style, physical performance and set pieces. This is only my set-up – Insight is a fully customisable tool, so it can be shaped around whatever you want.

The first dashboard is playing style and positional set-up. What we can see here is the structure of the team without the ball. 

If the team is in a low block, in a mid block and in a high block, we can switch it round easily. We can also see how they are in the same situations but from an attacking point of view.

The most important part is that this is driven automatically by the event definition. So once the possession state is identified, Insight aggregates the tracking data to show the team’s true positional tendencies. This makes it more intuitive.

So instead of manually searching through video, analysts get a clean, consistent visual of how the team occupies space in both defence and attack.

Next, I’ve created a dashboard which shows what happens when you lose possession of the ball under pressure. You can instantly see who lost the ball, where it happened and what the tracking shows about the surrounding pressure.

If I use an example, it’ll bring up a video. In this case we can see FC Copenhagen playing out from the back and how, when put under pressure, they eventually go up the field and lose the ball.

With a larger sample of matches, if a trend like this was to continue, it could become a clear indicator of a player or an area of the field to target. Because the widget is mirroring the movement in the video, you can see the entire sequence playing out.

Next I’ve created a dashboard called attacking play and it splits out runs behind the lines, line-breaking passes and take-ons. For each of these – and for all of the definitions – you can build custom rules.

So if we look at the run behind the lines, you can define exactly what constitutes a meaningful run for you and your team. You can change the maximum number of defenders ahead of the runner, the minimum number of defenders overtaken or change the maximum time between the run and the player who receives the ball.

This picks up the precise behaviours you want to see – and every example goes back to video, so you can see the movement, the timing and the outcome in context.

The final one of the playing styles is attacking passes. This goes beyond looking at simple pass counts.

I’ve broken the team’s attacking actions into meaningful categories. I’m looking at passes that progress the play, any attempts that have come from balls that enter the danger zone (showing how often the team looks to penetrate from these high-value areas) and then any attacking sequence that has started from the ball being played from the keeper in open play.

Next I’ve created a dashboard around physical data, so you can see how much distance each player has covered per game and the average number of sprints. We’ve got burst runs, sprints in behind.

Instead of zones, we can change this to players and select any player.

Lastly it’s set pieces. Here I’m only showing a single game. This would be more useful with a season’s worth of data, so you could analyse goals that have come from corners, shots that have come from the left-hand side and so on.

What Insight really gives you is a single environment where video tracking, physical data and event data all speak the same language.

Analysts can move from selected metrics to a holistic understanding of a team behaviour. Coaches get a clear visual explanation for when certain moments happen. And players can see those behaviours in context, as it’s linked back to video.

Ultimately, Insight turns raw data into actionable football insights. You’re helping teams prepare better, train smarter and make clearer tactical decisions.