When the Pyramids’ lifted the African Champions League trophy last year, it was the result of calm, focussed consistent work over many years.

Despite not winning titles in its first few years, the incoming management of the club always avoided rushed decisions. While other clubs may have greater resources, the key to Pyramids’ long-term success has been their ability to utilize theirs effectively. They have an agile team management structure, which avoid much of the bureaucracy that other African clubs face.

And they increasingly rely on data to drive those decisions.

What I didn’t write in my previous article about Twelve’s Data Expert Mohamed Dawoud is that, not only has he succeeded in front of the camera, but he has also been working behind the scenes to help build the Pyramids, as their Chief Scout.

And a large part of that work involves data scouting with Twelve.

He uses our more advanced tools, such as evaluating a player’s fit for a specific team or predicting performance across leagues, to make sure that signings are justified.

A recent example of this success is bringing Nasser Maher to the Pyramids. ten matches since joining, Maher has scored 5 goals and provided 2 assists, compared to 10 goals and 5 assists in his previous 70 matches. The data had shown that he was the missing piece of the puzzle the Pyramids needed and, after signing him from their title rivals, Zamalek, the Pyramids quickly found themselves chasing them down for a new title.

The title is still to be be decided, but this week, Mohamed shared a new triumph in the Twelve group chat. The Pyramids won the Egypt Cup.

The story at the Pyramids is the same as the story at Hammarby and many of the other clubs successfully using data. Once the processes are in place, the data becomes a natural component of the decision-making. Then the decisions get smarter and the teams win more.

It doesn’t matter if you are from Egypt, England, Sweden, Greece or Cambodia, it is all about getting processes in place, getting everyone aligned and prioritising informed decision-making.

The Pyramids are two points behind, Zamalek, who have the advantage going into the final game of the season on Wednesday. But with the right result, the cup and league double could still happen.

Not because I thought that being a qualified dentist was essential for the job as data expert in football, but because there isn’t much job security in football. I didn’t want to have to answer to his parents if he dropped out.

Mohamed reassured me that he had completed his degree, with high honours, so the possibility of becoming a dentist wasn’t closed to him. But he had decided that the skills he had learnt during his studies — “the principles of evidence-based practice”, as he called them — could be put to better use.

“Every analysis must be backed by data”, he told me, “not just theory or personal opinion.”

Mohamed had spent his final year of study balancing exams with preparing and presenting TV segments, after onTime Sports, Egypt’s leading sports channel, invited him on to present his analyses. Now he wanted to put Twelve’s analyses on Egyptian TV and Socials.

Having made sure he had gone into his new career aware of the dangers, I was more than happy to give him a position as a Twelve Data Expert. It seemed a bit strange connection— applying dentistry to football — but it had led him to exactly the right approach. When Mohamed builds an analysis, he starts with an idea he had about a player or team, he reviews the video clips to test it, and finally, brings in data to confirm or challenge it.

“Combing video and data builds a story”, he told me more recently, “you show the idea, support it with evidence, and the audience grasps a complete concept.”

Sometimes Mohamed’s analyses start with data, catching the audience by surprise, then they break it down through video, keeping them engaged while ensuring the insights come across.

“The beauty of data is that it doesn’t always confirm your hypothesis — it challenges it, forcing you to dig deeper and understand the game more comprehensively.”, Mohamed told me, “Data is loyal, honest, and impartial; it tells the truth, whether it agrees with your initial idea or not.”

As a result, Mohamed has spotted several young talents like Lamine Camara, who is now at Monaco, and Emmanuel Agbadou, at Wolverhampton.

Mohamed approach to TV punditry is seldom replicated in the UK and Europe. Data is either wheeled in as a straw-man, which the former football players blow over with hot air. Or it is used to confirm some narrative, as in recent media discussions about set pieces, without ever going deeper.

A year into our collaboration, millions of viewers now follow Mohamed’s insights. They know that every claim is thoughtful, backed, and entertaining.

“Data is the bridge that makes the audience rely on your analysis”, he told me, “especially when important decisions are at stake.”

I am sure many European and UK viewers would agree, and wouldn’t mind seeing a data-inspired dental student (or similar) on their screens.

How can companies be competitive in this environment? How do you avoid getting left behind? What happens if a competitor just copies your product?

Since Twelve Football have been working with novel AI products for the last two years, we have had been thinking about these question for a while. It has been a continual discussion between myself, our CEO Marcus Lådö and our Chief Data Officer, Dominic Jordan. We knew these changes were on the way, we have worked to make sure we were ready when they came

First, at the start of 2025 we clearly defined our mission statement as shown below.

The sentence and the five highlighted words capture what is important to us. We don’t just build AI, but we use the most advanced cutting-edge methods to help teams get ahead. It don’t just deliver data, it is about the insight which can be gained from that data. We realise that every club is different and we always identify the best way to help teams win.

The second step was to adopt, what we call, the U-shaped approach. Pre-2024 and the “AI revolution”, we did data science consultancy work in football. This involved prestigious collaborations with Champions and Premier League clubs, as well as leading national teams and some smaller clubs.

But it was here we had to think about how we were adding value. We realised early on that data science consultancy would become easier and that our customers would expect more (this is exactly what we see now, during 2026, in both programming and consultancy). So started to push our work in two opposing directions.

In one direction we built AI Analysts. So clubs with a smaller budget can get immediate help with scouting, performance and opposition analysis.

Here the focus is on delivering better automated insight than anyone else. When you ask Twelve Earpiece a question, it answers on the basis of data and our models of that data. It communicates in a way that anyone working in football can understand.

In the other direction, we worked to be more personal in the way we worked with our strategic partners. We make on-site visits, we work closely with staff members, we were agile and quick in addressing their needs. This the human approach.

As a result we have helped teams win big matches, survive relegation and build for long-term success, both through consultancy and through use of our AI tools.

The two directions — more automated or more personal — make up the U-shaped model illustrated below.

When done correctly, there is a virtuous feedback between automation and personalisation. By being the best at personalised help, we also build the best automated tools.

Being stuck in the middle, where your job can be automated by a generic tool is the most dangerous place to be today. For Twelve, we see that our plan is working: we are increasing both our long-term strategic partnerships and the users of Earpiece.

The old-world of consultants charging customers vast sums of money for nice looking Powerpoints filled with generic advice is gone. The era of AI solutions accessible by everyone and personalised is here to stay.

But before that, the challenge is to identify the numbers that will get the scoreline we want.

At Twelve Football we have built an approach around data-driven interventions. We identify the metrics that lead to a team reaching their goal and identify the actions which will get us there.

This can be a long-term strategy, like the one we have implemented at Hammarby. But it can also be a short-term. Many of the clubs who contact us are focused on survival. They have heard about the “miracle on grass”, where our Chief Data Officer, Dominic Jordan, helped Burton Albion beat 100–1 odds and survive in the Premier League. And they want the same solution for them.

One of these stories is Degerfors, a team promoted to Allsvenskan in 2024, but who found themselves struggling going into the last 8 games of the 2025 season. They had 16 points from 22 matches and were 5 points away from the safety of13th place (there are 16 teams in Allsvenskan).

It was then they booked in a meeting with our Head of Decision Science, Ida Höglund Persson.

“Degerfors asked us if there was anything we could do to help…”, she told me, “I booked in an initial meeting to go through their season so far. I expected a small group. Instead, the people showed up — the entire staff — all eager to find out how we could save their season!”

Ida set to work. She identified a number of key metrics essential for survival in Allsvenskan and plotted them to show to the staff. We have hidden most of them in the plot below, but one we have revealed is recoveries.

These metrics were coupled to changes in how Degerfors trained. Getting the ball back in midfield was now a focus, along with a number of other interventions. The whole team, players and coaching staff, were aligned behind these changes.

It paid off.

Not only did the metrics vastly improve, but Degerfors got almost as many points in the last 8 games as they had in the previous 22!

Not only did they survive, but they avoided a 14th place play-off. Degerfors fans will be enjoying Allsvensk football in 2026 and, with help from Twelve, are planing how to work towards a stable place in the league.

Twelve are working with several clubs on turnarounds. This involves implementing the steps aimed at achieving just one aim: survival.

First, we identify where the team have the potential improve most quickly and most effectively, based on the reality of the squad, the league and the schedule.

Second, we reduce analyse opposition behaviour, identifying exploitable weaknesses and avoiding time spent on projects or ideas that are unlikely to deliver in the short term.

Third, give the club a clear, evidence-led path forward. One that prioritises what actually could move the needle, aligns staff around a shared understanding of what good looks like in those cases and increases the probability of success over the remaining fixtures.

If you are interested, you can get in touch with Ida, Dominic or myself today. It is never too late for us to help you out.

Advice like this is easy to come by. But what impressed me is the point in time Adrian said it: five games in Hammarby’s 2024 season. They had set up metrics to measure our performance and they were not looking good. The mens team were neither winning nor playing the type of football we wanted to see.

And the fans were getting worried. Comments about Adrian’s inexperience, were first muttered and expressed out loud. “Hammarby looks like they are running a nursery school” was one of them: a comment made not only about the young team that was fielded, but also about the 27-year-old technical director himself. Some disgruntled fans snuck in to the training ground and put up a protest banner comparing Hammarby’s leadership to a Sunday league team.

What the these particular fans didn’t know was the details of the process Adrian had set up. During the winter leading up to the 2024 season, Adrian, together with data scientist Agust Palmason Morthens and video scout Jacob Hjelte (both also in their mid-twenties, Jacob was still banging in goals for Gefle in the Swedish second tier at the time), had revamped the recruitment strategy on both the mens and womens side. They had created clearly defined role profiles for the type of players they want to have in the team. To achieve this, Jacob explained what he was looking for in players, while Agust converted these into metrics. This led to the creation of the Hammarby Scouting Qualities, which are a weighted combination of 3–10 metrics.

This reorganisation of the principles did not lead to a lot of signings. In fact, to be in line with the strategy, the key was to give the younger players — sitting on the bench the year before — a chance to show everyone what they can do.

But now, with 3 losses out of 5 games played, the new coach Kim Hellberg hadn’t got the team playing the style of football we had (using data) recruited him for.

It would have been so easy to panic at that stage and make a change, but Adrian stayed calm. They had made a plan, now they had to see it through.

The composure paid off. Hammarby finished the 2024 season in second place. The team was now playing in the way we had set out to play. And the young players got to prove how they could adapt and improve over time. The team consisted of a small core of older (over 25) leaders and a majority of young (17–24 year old) up and coming talent.

A club strategy can’t be evaluated over one match, or even a run of five games. This is something which many clubs — and very notably many of those playing in the Premier League — seem to find really difficult to grasp.

The constant in-fighting for power, the failure to align values, the complete lack of a strategy grounded data, leadership giving their incompetent mates jobs in areas they don’t understand, the catching on to one fad after another, the banging on the tables and shouting at each other, the backbiting and gossiping, the list goes on ….

In my position I see this and much more all the time at clubs at all levels.

But not at Hammarby. Here I see a maturity from a 27-year-old natural leader. I see someone who puts a trust in data scientist Agust and a knowledgable young player in Jacob; someone who has an ability to learn from and work closely together with mens sporting director, Mikael Hjelmberg; someone who always without a pause lifts up the achievements of the women’s team whenever he hears someone praise the men; someone who values insight from data.

If you are a Hammarby fan and you have an opinion about the process the club follows (although not about daycare, thank you) you are welcome to send an email to Adrian. He answers every one of them.

• Large language models (LLMs) wouldn’t get better just by making them bigger. They were already big enough to contain most of the Internet.
• LLMs will help coders code, designers design and data scientists do their job, but they can’t replace the creative aspects of these professions. They only replace the repetitive, established activities.
• Despite these limitations, I still thought that AI was important. Applications based on LLMs can get a lot better (they could then and they still can now). We can automate repetitive tasks done by specialists. But we need a new way of building software to achieve this.
• We thus need to find new ways of structuring and fetching data that work together with LLMs. This required us to become better programmers, not to replace programmers with AI.

I immediately set to work on building based on this realisation. The first thing I did that summer was to build a David Bot. I took the pdfs of all of my books, created a system that searched through them and could answer questions in the way David Sumpter the author would answer them. That was fun, but I also realised that this was something that was too easy to do. I realised that Google would soon build similar tools and better than I could. And they did: first with LLMNotebook and now with integration into Gemini.

What was more difficult, and thus a much better idea to pursue, was to get an LLM to talk about data. One idea that was floating around at that time was to convert user queries (in natural language) into SQL (a database language) so an LLM could retrieve information. This was clearly (to me) a daft approach for most applications, because it was prone to error (hallucination) and natural language very seldom specifies what we want exactly from data. So, when I saw a Chief Data Officer from one big analytics company present how his team was trying to solve the ‘language to SQL’ problem for football, I realised I was on to something. I realised that solving the data/language problem was both non-trivial — i.e. this Chief Data Officer didn’t understand the problem — and extremely important — if we could talk about football data, then we could build something people would actually want to use.

There was something else going on in Summer 2023 which also influenced my thinking. In my work at Twelve Football we were increasingly overwhelmed with requests from clubs to help with scouting, match analysis and season analysis. A lot of this involved writing reasonably standardised reports which converted data into words. I started to think seriously about how this work could be automated. How could I automatethe boring parts of my work in football data science.

I would later call the idea I came up with Wordalisation. While visualisation turns data into visuals, the idea behind wordalisation is to turn data into words. Here is an example: visulaisation on the left, Wordalisation on the right.

The generation of this text is entirely automatic: you can scout any player in the world in this way. You can easily make a full report on the player. It is taken from our Earpiece product, which is free to try out on WhatsApp. We have also published a scientific articles explaining how to build Wordalisations and given a step by step guide for an example of evaluating shots in words. We also provide an open source version to help others (even other Chief Data Officers!) implement this approach themselves.

The non-trivial challenge, on which the Wordalisation solves, is finding a prompt to the LLM that will produce reliable output. A prompt is the text you send into ChatGPT (like “What does postmodernism mean?” or “Why was Sam Altman wrong about the progress of LLMs?”). The prompts we use for Wordalisations consist of thirty or forty paragraphs of information, carefully structured in order to get the best possible and most reliable description of the player in terms of data.

Using the Wordalisation approach, we already had a prototype (of what later became Earpiece) ready in early-2024 and made it available to the clubs we worked with. It meant we could more quickly answer their scouting and performance questions.

At this point, we had not fully addressed the challenge I set myself in 2023 of answering a user’s query based on data. To create a Wordalisation we need to first identify the data and background information that will best answer the users question. If the user selects the data from a dropdown menu then this is relatively easy, but working out what the user wants from text is more difficult.

To give a feeling for the approach we took to solve this part of the problem, I often use the analogy of a choose your own adventure book, like the Fighting Fantasy novels pictured below.

In these books, the reader is given options — such as “fight the Dragon” or “run away” — and by turning to a different page number can follow a different story. We exploited a capability of LLMs known as function calling, where a user query can be used to then call a function, which in turn assembles both the background information and the data needed to answer the users question. By pre-defining ten or so functions (comparing players, looking at transfers between leagues, etc.) and insisting that our language model only uses these functions, and nothing else, we limit what the user can do, but ensure that it only provides reliable information.

From summer 2024 we started building a choose your own adventure tool for football. In Spring 2025 we released the first version of Earpiece, not on the Web, but on WhatsApp: the favourite form of communication for those working in football.

The Earpiece product is now used at over 50 clubs worldwide to ask questions about players and get reports on their performance. Next week we will release an individual tier, which will allow anyone working in football to talk to our AI analysts about the game.

When I think back to what I thought Summer 2023, I feel I was mainly correct in my analysis then. LLMs have not solved the problems some of the hype of 2023 said they would. Sam Altman was wrong (although he did hype his company to the stratsophere along the way).

We have focussed during the last two years on a very different approach than Open AI, and looked at how to harness the power of existing LLMs to build reliable applications. Along the way we have published articles explaining our work and given seminars on the methods. But now the Earpiece product is coming out, I will also be communicating more about what we have learnt: starting with a soon to be announced course in early 2026. The name of the course will follow the name the industry seems to have adopted for the approach we have used since 2023: Context Engineering for Football.

I like the term Context, because it really focusses on the central challenge in football and in many other applications: without context data doesn’t mean anything.

The possibilities for a Context Engineering approach are endless, stretching well beyond football. I hope that more people will start to pick them up and I will be working to explain this approach in more detail in the new year.

A key reason that Slot fitted so well lay in his ability to create danger during attacking transitions. The plot below shows how much expected threat his teams produce in the 10 seconds after they win the ball back. He took Feyenoord into the top 25% on that metric when he became manager and kept their performances at a level comparable to Klopp for most of his time in charge.

Slot did not continue to produce attacking transitions at the same level as Klopp, even during hisPremier League winning season. This metric and many others associated with attacking transistions continue to decline. In the context of Liverpool’s current challenges, though, we should remember that even during Klopp’s reign there were variations in performance.

What was important from the perspective of Hammarby’s search for a coach in 2023 was that we could identify a manager who has an impact on a team. It is here that time series like the one in the figure above help the most. We can see that when Slot started as Feyenoord, he had an immediate and sustained effect on their ability to create danger when they won back the ball.

Agust and technical director Adrian von Heijne had, during 2023, conducted a study showing that successful young teams are defined by having a high defensive intensity and quick counterattacking style. Another requirement was that the manager was able to get the most out of young players. It was also important that they wanted to attack with the ball: Hammarby fans always want their teams to play on the front foot and they were unhappy that the team had (not only underperformed) but also been too defensive.

The manager Hammarby recruited was Kim Hellberg. He had taken over at Värnamo two years previously, after they had been promoted to the top tier. His first season at the club had been all about survival, but during 2023 he had established them as a ball playing team, who had more possession in the final third than Hammarby had had that same season. This is illustrated by field tilt, which is larger when a team ha more of their possession in the final third.

The key to interpretting these graphs is to look at the change. Hellberg had shown during 2022 that he could keep a team with a tiny budget up, during 2023 he showed that he could make them play his way. The trend at Hammarby during 2024 and 2025 shows we made the right choice. We now wish Kim all the best after his move on to Middlesborough.

At Twelve Football, we have used this approach to help a wide range of clubs, from smaller budget leagues inScandinavia and North Africa to the Premier League to, find new managers. There is definitely a qualitative component to the approach: it is about understanding the story of a managers time in charge. But looking at what happens when a manager takes charge is a very powerful way of judging their effect on players and the team.

I was curious then about the other manager we (and many of the fans) identified for Liverpool back in 2024, especially because the rumours are flying around about his possible departure from Real Madrid. Xabi Alonso has produced a slow but steady improvement in attacking transition (which is an aggregate of several metrics we use to measure how teams transition when they get the ball), while Bayer Leverkusen have dropped in that measure since he left.

In fact, Alonso has improved Real Madrid over a whole range of metrics, exactly as he did in all his earlier managerial jobs. If the rumours really are true, and Real Madrid fire Alonso, there are lots of Premier League data scientists who will be explaining to their clubs that he is a manager who really does make a (statistically) significant difference.

This was the question that Adrian von Heijne posed to us when he started as technical director at Hammarby in 2023. Like many clubs, we play in a league where other teams have better financial resources than we do. Furthermore, we can’t compete in terms of salaries and transfer fees with most other leagues in Europe. But what we can do is build an environment that develops young talent, ensures that the players in our squad increase in value and in this way create stable long-term success.

It is this logic which underlies the question. By identifying how young successful teams play, we could bring in talent which would entertain the fans, win matches and bring long term financial stability.

The first step was to identify the teams which fit the profile. These are given in the table below.

The standout team is RB Salzburg, which had an average squad age of under 24 for most of the time period, and have dominated in Austria. Ajax, during that period, were also young and won a lot, as did several other Dutch teams, like AZ Alkmaar and PSV. In Sweden IFK Norrköping 2015–2019 were a team to admire, but the also provided a warning for what happens when the transfer books aren’t balanced and there is a move away from investment in young talent.

Once Agust Pálmason Morthens, out lead data scientist, had identified these teams (all of them have over 1.75 points per game and average age under 26 over at least 3 years), he set about studying how they play. First he asked the scouts and staff at Hammarby to list the aspects of play which they thought were important. Although, he told me that, if he would repeat this today, he would not make this initial filtering, but look into all metrics. We have created hundreds of metrics at Twelve, and it is important to make sure we account for all possible explanations of success.

What do I mean by a metric? Here are some examples: defensive line height, recoveries per 90, shots for and against, penalty area touches for and against, passes per defensive action (PPDA), expected threat generated within 10 seconds of a recovery, time to a defensive action…. Some examples for defensive and attacking transitions are shown for Salzburg this season (full report here).

There is a lot to take in when we look at all these metrics and these are just for transitions! We have further metrics for chance creation, attack and defence. The question Agust was asking is which of these metrics correlate with success in young teams.

I am not going to give you the full list of all of the most important metrics. We have to keep a few secrets in what we do at Twelve! But let me tell you some of the key findings. Young and successful teams generate high xT and get closer to the goal quickly after recovering the ball. They are quicker to defensive actions and recoveries following a ball loss. And, of course, they have high xG for and low xG against.

In order to tease out what marked the success of young teams in particular, Agust compared younger and older (average age > 26) successful teams. Here we could see, for example, in the plot below that older teams could be successful with low defensive intensity and high PPDA, but for young teams, success came from pressing hard all of the time.

This analysis set the basis for Hammarby’s strategy in 2023. We now knew how a young team should play. The next step would be to find a manager who could lead a young troop and shape this style of play.

In reality, I probably don’t have any choice other than to be honest, because when I asked Twelve’s lead data scientist Ágúst Pálmason Morthens to write some notes about Twelve’s work over the last three years with Hammarby he was very honest with me.

He started his debrief by writing: “Before Adrian came in, there was no proper structure on our collaboration with Hammarby and it was not clear (at least to me) how we were contributing.”

Agust’s assessment is a blunt but fair representation of how things were at Hammarby in summer 2023. I had been embedded at the mens team in 2019 and some of 2020, working very succesfully with manager Stefan Billborn and the then technical director Ola Larson. But then in 2021, when I went back to work at the University, Stefan got fired (after winning the Swedish Cup!) and Ola left (“to pursue other challenges”) there wasn’t the same impetus around using data. We did get random requests here and there from people in the men’s team, mostly from Abel Lorincz, the match analyst. He would ask about line breaking passes, formation changes of the opponent, set pieces and how we could measure style of play. This led to lots of interesting problems for myself, Agust and Jernej Flisar to work on. But there was no structure to it.

The benefit of giving an honest account now is that I know that many clubs (even those in the Premier League) are still in this sort of situation I describe for Hammarby in 2023. They have some data scientists, but they don’t know quite what to do with them. Decisions are driven by the strongest voices at the club: be it scouts, the board, the coaching staff or by agents, without a clear decision-making structure. Data is seen as part of a power game, and data scientists roped in to “win arguments”.

The exception to this rule at Hammarby in 2022 and 2023 was in the women’s team, coached by Pablo Piñones-Arce. Maybe it was because they didn’t have the capactiy to look at as many players as the men. Maybe it was because Pablo had very successfully worked with Fran Peralta during 2020, before he went to Athletic Club. Or maybe it was simply because Pablo had my number on WhatsApp and liked to disturb my evenings! But in any case, Pablo would text me about potential players.

One player I was very proud of helping bring to Hammarby during that period was Maika Hamano. She was offered to Pablo on loan from Chelsea and he texted me one Tuesday evening to ask if she really was the player they needed?

The data was limited, but I could show him that she had excelled at the Womens World Cup U20, in particular in terms of her ability to receive the ball in the final third. That was just what Pablo needed to see: it was on the final ball into danagerous areas where the team had struggled the season before. Maika looked like the answer in the data.

She also turned out to be the answer on the pitch. Maika only stayed for half of the 2023 season, but she became a club legend in that short time, and played a big part in the cup and league double the team went on to win (she is represented on the left in the fans’ Tifo at the 2025 cup final).

Agust took over the Pablo hotline requests when he started at Twelve in summer 2023 and has continued to man the phones ever since. Again, in his rather dry style when summarising his work at Twelve, Agust told me that “I became the Hammarby responsible person in Twelve. That did not happen formally, it just happened organically.”

This is not quite my memory. My memory is that Agust kept hanging up Hammarby shirts in the background every time we met on Zoom or kept wearing Hammarby scarfs to our offices. So that (combined with the fact he is an amazing data scientist) it was obvious to me, and to Twelve cofounder and die-hard Hammarby fan Marcus Lådö, that he should be doing their scouting. So thats what he did.

Agust worked together with Nicole Kozlova to develop a method of scouting we have built on ever since, which we call team fit. It is a clever, but easy to explain method. First you compare the team the player is in now and the team they will move to in terms of style of play. In his report on Emilie Joramo, Agust first noted that Rosenborg, the team which she played for in 2023 had a similar style to Hammarby.

Then he went on to compare her to other players in the Norweegian league in terms of a radar plot, where we can see both her defensive and attacking qualities and her shots (very good at getting in the box).

There was little doubt that she was a player who would fit in well at Hammarby. And, together with Anna Jøsendal and Stina Lennartsson, who Agust and Nicole also analysed at that time, she has become a central player in the team.

Avoiding silly mistakes is not the same thing as building a successful recruitment process. For that we needed more than a clever graduate from the Soccermatics course with a Hammarby jersey hanging on his wall. We needed a technical director at Hammarby who really believed in using data and knew how to integrate it into the club as a whole.

And that story will continue in my next post…

If you look at the y-axis on these graphs you will see it is marked Outcome. Rather than simply being points, Outcome is an average of xG for and against, as well as xG per shot for and against. It is these metrics that Adrian first set up when he took over the club and asked the fans to judge him against. Adrian also set up style of play metrics to measure how the teams played: whether they dominated possession, attacked quickly when they won the ball back and got the ball into the final third. The idea was to measure both how well the teams performed and whether they played the way the fans want us to play.

What made Adrian’s approach unique was that he asked me to present them to the fans and then put up the numbers (updated every week) so that they could monitor progress.

Adrian’s approach has created a culture around the club where, whatever happens on match day is put in a larger context of how the teams are developing. Hammarby fans say that it is important that their team plays beautiful, free flowing, attacking football, as well as winning. Now they can hold the manager, Adrian and me (!) to account.

The presentation of data to the fans also makes data-driven work within the club easier to carry out. Scouting is done through a combination of data based approaches and watching players live and on video. Opposition analysis starts with a data profile that predicts how the team will play against Hammarby and suggests what games and situations the video analyst should watch. Players receive personalised performance analysis after every match. And post-match analysis is helped by a data report, outlining what really happened in the game. Data flows through everything that happens at the club.

There is one obvious question that the headline might lead you to ask: “How can you say Hammarby are best when they didn’t win Allsvenskan?”. Here i should give rightful respect to Mjälby. This small club from the south of Sweden won Allsvenskan, also using data and research to power their success. It is hats off to Dr. Karl Aksum and his team in pulling this off.

But as Adrian von Heijne pointed out at Twelve Day, from a purely statistical point of view, the performance across our teams, sustained over a longer period, does support the claim. But I also hope that, when I stand in front of the fans next year, we have secured double gold and not just two second places.

In the next post, I will write more about how we used to data to set up Hammarby’s style-of-play metrics, recruit the managers (the men’s manager, Kim Hellberg, has just moved to Middlesborough so we are already in the middle of that process) and put the team on the path to success.