Daryl's Notebook: What I learned from the 2025 American Soccer Insights Summit (Part 1)
From MIT Sloan, where all sports and practitioners unite under one roof, to the ASI Summit, where the best minds and enthusiasts of US soccer analytics gather.
This year’s MIT Sloan conference once again delivered with plenty of inspiring keynotes, presentations, and panels where the best data practitioners from a wide range of sports came together and shared bits of insights from within the industry. The sheer number of panels and presentations made available on their YouTube channel made it impossible for me to cover all of them in just a single Notebook edition, so I encourage you (if you are interested in sports analytics) to go to their channel and find what catches your interest.
But, among the wide variety of sports analytics being showcased in Boston, Massachusetts, there was only one panel that discussed about football/soccer analytics, which was understandable considering the MIT Sloan is hosted in the US, where baseball, basketball, American football, and maybe hockey (?) dominate football a bit more. Even worse, at the time of writing, the panel that consists of the biggest names in football analytics like Sarah Rudd, Luke Bornn, and Raúl Pelaez is still not uploaded to 42 Analytics’ channel. With other, smaller conferences like the PySport meetup in London or the Field of Play conference in Manchester have also not made their presentations and talks available to the public, that leaves me with no football analytics content to enjoy amidst the sports analytics conference season, probably until the Opta Forum and the StatsBomb Conference are announced…
…That is until I found out about the American Soccer Insights Summit at Houston, Texas! A whole summit dedicated to football analytics? Sign me up…virtually! This year was the first edition of the summit, but they have received a lot of support from many prevalent names in the industry, including Rudd herself, Ravi Ramineni, and more. Combine that with a very reputable lineup of speakers, and I will be there (virtually) no matter what. At the time of writing, they have also made almost all of the presentations available to the public, which is just a great chance for me to watch them and take notes from within the comfort of my own home.
I was inspired and learned about the ASI Summit after listening to Akshay Easwaran’s podcast episode with the American Soccer Analysis team, which is a good episode that I highly recommend checking out. Akshay himself also wrote up a good summary of the two-day event, which I also recommend having a read. Instead of giving a summary for each presentation like Akshay did, I will borrow the presentation order that he used in his article and follow the same structure that I used for my MIT Sloan piece, where I note down some memorable and noticeable quotes along with my own thoughts and opinions. Without further ado, let’s get into the presentations!
Keynotes
Bridging Insights and Decisions in Soccer Operations
Presentation by:
Ethan Creagar (Data Scientist of Houston Dynamo FC) &
Pat Onstad (General Manager of Houston Dynamo FC)
Why did I pick this presentation?
There’s nothing more that I love about football conferences than when people who are working at clubs and within the industry sharing their own experiences and insights into what their club are working on. It is a rare glimpse into how data are being used within sport organisations and that usually is the best way to learn about practical applications of data analysis outside of just creating visualisations and generating insights from such vizzes.
And it is a rare thing because not many clubs are willing to actually share what is going on behind the curtains. While it is understandable from a competitive point of view, because no clubs or organisations would want to normalise the things that are giving them the advantage over their competitors, it hurts the growth of the industry as a whole. Without being able to know the practical applications of football analysis, it is hard for enthusiasts and hobbyists to learn the necessary skills to break into the industry and widen the talents pool. As such, keynote presentations like this are super helpful for enthusiasts like me to understand how clubs are using data and how I can improve myself to break into the industry.
My notes!
Houston’s recruitment process: Performance & Roster Review → Scouting Brief → Market Analysis → Player ID → Data & Video Review → Detailed Analysis & Live Scouting → Final Decision
PO (Pat): “We don’t just say ‘oh this guy is free, he’s a really good player’, but we have three right wingers, why would we bring in a fourth? We definitely will be specific in terms of what we are looking for.”
EC (Ethan): “One of the underrated parts of the process is market analysis. Before just jumping in, looking at players, and say ‘this player fits, let’s try to go and sign this player’, getting a holistic view of the market and say ‘what do we think is out there, what do we think the level that we should expect for this player in the price range that we have been given?’ If we think we are not going to be able to get the right level of player at this budget, maybe we go back to the scouting brief and say, ‘this is not the right budget for this profile’. If not, we will move to player ID.”
PO: “The tricky part of [the scouting] world is seeing clips. We are only seeing a short 20-30 clips on a player and are trying to make a reasonable judgement. So, one of the things we task the scouting group with is, don’t just show us 30 highlights where this guy gets an assist or a goal everytime we are looking at a progressive #8. Show us the strengths but also show us the weaknesses. […] There are different things we ask [the scouting team] to show, sometimes they show a good picture, sometimes they don’t. But in the end, if it is a guy that we think is intriguing, the data supports it, the scouts support it, then myself and Asher [Mendelsohn - technical director] will start watching full games.”
PO: “When we go live [scouting], we already know he is a pretty good player [because] we have seen enough videos and clips, and we already know what type of player [he is]. But the biggest thing, for us, at that point is the character. Are they [a] good teammate? Are they somebody that, when the team is down a goal, are they trying to win or are they giving up? Are they somebody that, when their teammate gets cracked from behind, are they one of the first guys who run over and argue with the referee, or are they going to walk to the side and take a drink of water? What happens when they score a goal, are they celebrating with their teammates or are they not?”
Closing Keynote
Presentation by:
Zayne M. Thomajan (General Manager of Chicago Fire FC)
Why did I pick this presentation?
Another opportunity to get a glimpse of how data is used at a club, and it is from a different club, so it is good to compare how Chicago Fire do things differently to Houston Dynamo. Each club have a different way of doing things, and some of those things work, and some do not. In both cases, those experiences are good to learn to adopt something that works and to avoid the things that do not.
From Akshay’s summary, I can also see the difference in both keynotes, where Pat and Ethan talked about Houston’s recruitment process and the steps that they take, and Zayne discussed the questions and the reasons behind that process. Two different keynotes, two different perspectives, more meaningful information delivered. It’s always a win in my eyes!
My notes!
Summarising the first section of Zayne’s talk: women’s football is growing at a rapid pace with on-field and off-field records being broken every season, which is also driving more investments to improve the quality throughout.
“NWSL clubs must be very intentional, strategic, and proactive about their roster strategy and any acquisitions,” because:
the league is at a unique position due to its position as one of the top women’s leagues globally with a salary cap,
limited and unique roster rules (max 26 players, up to 4 with minimum salary),
limited player development pathway (no second teams or academy structure).
“There is a need to be very intentional about who you bring in and try to maximise the identification of players who fit into [the club]’s system, style of play, has the right performance metrics. But are there any additional elements that allow them to be integrated into the team more quickly? Are there certain patterns and trends where a player coming from this league has an immediate impact, thus eliminating the risk of taking a risk on a player and feel assured of their ability to contribute straight away.”
Roster build process in the NWSL:
Defining the positional profiles & metrics,
Be aware of the succession plans, making sure that the distribution of ages, positions, minutes played so [the club] don’t have any gaps,
Continuous evaluation of performance of current roster,
Have a clear plan for the following windows, and communicate proactively to the club’s scouting and data departments,
Begin the player identification process for any positions of need,
Maintain a detailed database of players observed containing all necessary key information, subjective analysis from the scouts, and data points from the data analysis team.
“When [the club] bring in a player, it is a holistic process of making sure that they have the support because we have these data points for on the field, but also how can we support them off the field so they can perform to the best of their ability?”
Some questions to consider:
“How does a player’s goalscoring record translate from different leagues to the NWSL?”
“Do players coming from certain leagues or specific clubs adapt to the NWSL more quickly?”
“Do players that have already left home once to play in another international league adapt more quickly because they have already experienced a significant move to a new place?”
“Are international players that can speak English more likely to provide immediate impact on the pitch?”
NWSL player origin analysis using Statsbomb’s On-Ball Value (OBV):
Supports that there is an adaptation period for players new to the league,
Players that played in the NWSL for at least a second consecutive season have higher average OBV than those new to the league,
NWSL players joining directly from Liga F (Spain) adapt slightly better than those joining from the other leagues (English’s WSL, French D1, German Frauen-Bundesliga),
Italian Serie A Femminile has a notably lower average OBV per player than all other player sources (but sample size is very small at only 5 leagues).
Salary efficiency:
International transfers tend to demand a higher transfer fee and higher salaries, plus an international spot.
There are consequences with paying a higher salary for an international player, especially within the confines of a salary cap.
“How can we ensure that the increased salary that [the club] paying are put towards players that are going to have an impact?”
“Where is it smart for teams to be more generous with base salaries if it indeed translates to better on-field performance?”
Salary & Productivity Analysis:
Goalkeepers: Positive relationship between On-Ball Value per 90 and salary, suggesting that more highly paid GKs have better performance.
Defenders: No relationship between OBV p90 and salary. Best performing DFs are at mid-to-low end of the salary range, but results should be taken accordingly due to the difficulties of measuring defensive performance.
Midfielders: Weak, but positive relationship between OBV p90 and salary.
Forwards: Strong, positive relationship between OBV p90 and salary.
A player’s ability to adapt or have an immediate impact varies due to multiple on-field and off-field factors, including suitability to a team’s style of play.
Is there a partner club to send players on loan to, especially when the NWSL have no extensive reserve squad setup?
Industry presentations
Building an Analytics Operation: Lessons Learned through an Oral History of American Soccer Analysis
Presentation by:
Brian Greenwood (American Soccer Analysis) &
Tyler Richardett (American Soccer Analysis)
Why did I pick this presentation?
American Soccer Analysis, or ASA, have been one of the pioneers in introducing football/soccer analytics to the US soccer audience and even to MLS and NWSL clubs. They have also acted as the platform for many of American data practitioners to kickstart their career in football analytics. Having been around since 2013, they have experienced everything from working with the most basic tech stack to now evolved to having one of the most impressive data infrastructures in the football/soccer world.
I have admired ASA from afar and envied of the work that they have done for American soccer. So, I am hoping that, by being able to listen to Brian and Tyler on how they have built ASA’s data analytics operation, I can learn some tips and tricks that I can use in the future.
My notes!
TR (Tyler): “None of us have the answers, nor do any of us think we are going to come up with them. But there are a lot of smart people out there and sometimes go without being heard, and our self-proclaimed job is to find those people, give them a voice, and have a bit of fun doing it.”
TR: “Everyone starts somewhere, starts small, put these things out in the world, and you will learn best by doing it.”
TR: “It is one thing to publish a piece of analysis, and it remains static, but it can be a lot more impactful to put something interactive into the hands of the consumer and allow them to lead their own discovery and their education.”
BG (Brian): Regarding scaling up the operation:
Introduction of SQL database, CI/CD, configuration as code, etc, enables quicker iteration, easy rollbacks, and further expansion of coverage.
Orchestration of data pipelines allows one-off analyses to mature into self-sustaining metrics.
Use configuration files, makes it easier to add support for new competitions over time.
Do not assume data will be flawless. Add quality checks and/or automatic fixes for common issues. Isolate errors, and quarantine bad data.
Version control data, code, and assets.
Not all data needs to be accessed at the lowest levels of latency. Warehousing & infrequent analytics are different from performant web apps.
Reserve high-cost, low-latency stores for only “production” data, and use low-cost (e.g., blob storage) stores for everything else.
Consider the needs and abilities of all consumers.
BG: Start with success criteria
Define evaluation methods and success criteria early.
Especially valuable when your target is ambiguous/intangible and you risk chasing shadows.
Building a Data-Oriented Culture on a Budget
Presentation by:
Alejandro Dávila (Director of Analysis of Mazatlán FC)
Why did I pick this presentation?
It can be hard to build a data team and a data culture with a small budget or little resources. But as football evolves, the need for a data team is growing with clubs exploring many ways to build such team of their own while also stay within their means.
It seems like Mazatlán is one of such clubs who do not have the resources to compete with the big names in Mexico while also being one of the newer franchises, yet they have managed to build a data team in-house. While they have not experienced much success, they are still a good blueprint for smaller clubs to look towards and take away some lessons from Alejandro.
My notes!
“For the money that we spend, you have to be very precise and not waste any single dollar.”
Training analysis:
Evaluations & individual plans
Documentation of individual plans including physiological and technical, tactical objectives
Workload management
Aims: Improve performance & injury prevention
Objectives:
Standardise the protocols into categories
Make the data capture and storage efficient
Generate relevant insights for decision making
Pillars:
What do we train? (Objectives): Document all training sessions and turn them into valuable insights, which will benefit the development of both first team and academy players
How much we train? (Volume): Measure training intensity & volume and visualise via dashboards for the coaching staff. Perform holistic analysis of each training session.
Creating performance data:
Objectives:
Standardise protocols for 1st team and all youth teams
Facilitate qualitative analysis of coaches from all teams
Make the data capture and storage efficient → generate relevant insights for decision making
Provide reports and forms for coaches to fill out with a rating system → allows for more effective player/position profiling and style of play profiling
Takeaways:
Create valuable data without cost
Data interpretation and storytelling is crucial
Ask the right questions
Partner with great organisations
Outsource at a lower price
Speedrunning an Analytics Department in 50 Days
Presentation by:
Arielle Dror (Director of Data & Analytics of Bay FC)
Why did I pick this presentation?
Getting started with a fresh and new soccer franchise who have already started their first season is probably one of the toughest challenges to walk into as a new director of analytics. Arielle had experienced such challenge when she joined Bay FC and was trusted to use data and analytics to turn the club’s fortune around, who were at near bottom of the NWSL at the time.
Since then, they have become one of the stronger teams in the league and have remained in the conversation for at least a playoff spot last season and the current season. Arielle has also expanded her analytics team to include a performance analyst and a data analyst at the time of writing, after the success of the first season. With that, she definitely has a lot to share from her experiences and a lot for me to learn from.
My notes!
Data & analytics does not only contain the usual jobs like recruitment (creating dashboards/reports), supporting the coaching staff (create pre & post-match reports, in-game feedback, building internal metrics & models), building scalable infrastructure. It also contains unusual tasks like vetting vendors/organisations, tech support for members inside the organisation, or anything that related to touching a computer.
“If you cannot get more help, you need to focus on a limited number of tasks that will allow you to be more successful.”
“If you ever find yourself working in a small department, which is really common in soccer, my big advice is working with your boss (usually your General Manager) to understand what is important to them and then work on that.”
Getting external hands and advice is another way to get an outside perspective and have more people working on projects. Capstone project is another way to work on projects that have unclear impacts and do not want to spend a lot of time on, while also provides many students with hands-on opportunities.
“If you want to have influence, you need to be in the room where it happens.”
“Our coaches have a lot of thoughts in their heads, but they struggle to get some of those thoughts out in a way that was meaningful for [the data & analytics team]. So, being able to listen to them and how they talk about what they wanted to do was really meaningful.”
“Having conversation that [the coaching staff] can hear [the data & analytics team] have and conversations between [members of the team] together allow for some level of informal teaching to take place and it brought [the team] some level of comfort.”
“We emphasised early on that we wanted to keep data as neutral as possible. […] Data is just merely information, and we are all working together to create a product on the field together.”
“Maybe the most important [factor], patience is everything for an expansion. There were so many moments in this season where we just felt like we were running into a brick wall repeatedly, and no progress was being made, but the reality is progress is incremental until the season is over, and you can see where you grew.”
“We are not where we are going, but we are not where we have been” - Matt Potter (Bay FC General Manager)
Building Shiny Apps for Rapid Prototyping
Presentation by:
Lydia Jackson (Machine Learning Engineer at Teamworks/Zelus Analytics)
Why did I pick this presentation?
I saved this quote from An, who attended this same conference, and, presumably, said this after watching Lydia’s presentation. Brian and Tyler from ASA also talked about this topic briefly, and I wholeheartedly agree with every quote said about the topic. A dashboard is always much more effective at answering questions, and when you put that dashboard into the hands of the coaching staff, they can answer many more questions that they might have not even thought of. Dashboards are so much more impactful than a traditional report, paraphrasing what Tyler Richardett said above.
There are many hidden aspects of building a dashboard, however, and it sometimes can overlap with the traditional field of app development. I recently explored developing and deploying a Streamlit dashboard onto Snowflake, and while the process was challenging at times, I learned a whole lot throughout the whole thing. It also motivated me to dive deeper into Streamlit after a few years of neglecting it for other stuff, and I also want to expand out to Shinyapps in R as well. As such, Lydia’s presentation is the perfect place for me to learn more from someone with industry experiences.
My notes!
Some questions that Lydia received from teams that she worked with, internal team members, or even leaderships:
How good is this player in this role?
Are our players getting worse?
Why don’t I see my team’s most recent match report?
What manager would be a good fit?
How does this player’s performance in that league translate to this league?
Who should we sell this player to?
Typical workflow for any of these questions:
Deciding to build a report to answer the question
Delivering the report to the relevant stakeholders
Retweaking/rebuilding reports to reflect changes from data vendors or stakeholders
…But this workflow is not fun and time-consuming!
“There is a constant demand for fast, insightful, and accessible answers. […] As demand grows, our ability to deliver suffers.”
Solution: Shiny apps (using R, ShinyProxy, and Docker)
Skillset fit: existing expertise in R
Self-sufficient: instead of relying on Lydia’s team to generate reports, users can explore the data and find the answers themselves
Interactive: foster better engagements and feedback with users
Cost effective: does not require significant investment and infrastructure
New challenges:
Reduce development time to build fast
Deploy and manage many apps
App template:
Use `golem` to create, build, and deploy Shiny apps in R!
Build reusable logic
Standardise common UI components
Create functions for common data queries
Implement custom logic
Allows the developers to focus on solving specific problems instead of reinventing the wheel and figure out how Shiny works.
ShinyProxy acts as an application gateway, which handles user authentication, app routing, and manage lifecycles of Shiny apps. Combines with AWS EC2 as the server.
Lessons learned:
Leverage data table (in R)/data frame (in Python) for speed and efficiency, especially working with a large dataset.
Include a lot of comments in your code for readability and maintainability.
Separate your app logic (the UI and the server codes) from business logic (core functionality and data processing that the app depends on).
Incorporate logging for better debugging and monitoring, for tracking user interactions, simplifying debugging, and informing optimisation.
Keep It Simple Silly (K.I.S.S.): focus on primary question or objective, limit reactivity, and iterate gradually.
Recap & reflect
It was eye-opening to hear the experiences of many data practitioners across different club levels in North America, from well-established clubs like Houston Dynamo and Chicago Fire, to newly-established like Bay FC or Mazatlán. But one thing that hits home for me that everyone have mentioned is the importance of having a well-defined process and a list of achievable targets in place. It is beneficial when starting out because it gives a sense of direction and allows for better focus on aligning the coaching staff with the process, the vocabulary, and the use of data in general. It is also beneficial for clubs who are already established since it reinforces the core principles of what they do and how the club want to play, which is very helpful when it comes to recruitment.
When it comes to technology, a theme that both Lydia and the ASA team share, along with Alejandro and Arielle, is the embracement of dashboards when it comes to delivering insights and helping relevant stakeholders answer their own questions. As mentioned earlier, I think dashboards are very useful for thinking outside of the box, because it encourages exploration of the data and finding questions that one might not have even thought of. It brings and creates multiple perspectives depending on how the data is viewed and how they are filtered, instead of being static and limited to just the data that are delivered in a report.
Overall, I still believe that these keynotes and industry presentations are invaluable when it comes to picking up experiences that you would not normally get from just doing projects and working on ideas. They show you a behind the scenes look of working in a club or an organisation, the struggles that they faced, and how they overcame such struggles. In a world where many clubs and organisations are keeping their cards close to their chest, these presentations are a good glimpse into the industry and I found them super helpful as someone who is trying to make the step up from just making visualisations and generating insights from them.