A conversation with: Daniel Krueger @DanielKrueger96

I was fortunate enough to be invited to the Opta Pro Forum in February. These type of events can be awkward for newcomers, but luckily for me, the first person I spoke to was Daniel Krueger. He had possibly traveled the furthest of anyone to attend (from Oregon) and had the unusual claim of being perhaps the only American born Huddersfield Town fan. If there are any challengers to this title I will correct the record.

Daniel represents to me the new breed of young analysts, used to working with big data, highly technically proficient and keen to learn. We had a twitter chat about the work he has done so far and his interest in working in football analytics in the future.

Daniel in action for Colorado College Tigers

2. 
   

   
   

   Q: Hi Daniel, tell us a bit about yourself, your footballing background, and how you ended up as, very possibly, the first US born and bred, supporter of Huddersfield Town?
   Mar 28Sent
3. 
   

   

   Thanks for having me! I grew up in Eugene, Oregon playing soccer since I can remember. I came up in the us club system and saw first hand the inefficiencies such a development program provides. I then went to play college soccer at Colorado College, where I just graduated from last December. I started supporting Huddersfield around the time when it was clear they were going to be promoted into the prem. I had never supported a European club before and didn’t want to be the typical American who picks a top 6 club. David Wagner being “American” helped as well. I then got a chance to see Huddersfield play Tottenham at Wembley last season and - despite a resounding defeat - I was all in.

    
   Mar 29
4. 
   

   

   Tell us about your playing college soccer career, what type of player are you?
   Mar 29Sent
5. 
   

   

   I played center forward in a 4-3-3. Playing competitively for so many years, I think I lost a little bit of the joy I used to find playing the game. There were days it felt more like a job than a passion. Kinda fading off into the Sunday league distance here playing wise which I’m really enjoying. Overall, no regrets. I think it’s remarkable how many American college athletes take for granted to opportunity to play their sport at a high level while getting a degree. Not really something you can do in most of the world. As far as soccer for me, really looking to transition to the analysis side of the game these days.

    
   Mar 29
6. 
   

   

   If I remember rightly you gained some experience of sport analytics working with the US Olympic team. What did you do with them? And what lessons can you take from it into football analytics?
   Mar 30Sent
7. 
   

   

   Yeah I worked with the strategy and business consulting department of the USOC. In that role I worked on the Medal Expectancy Model that the usoc has built in house. The model uses results from all international competitions across all olympic sports every week and runs a variety of R scripts to do Monte Carlo simulations to build expected medal count scores. Not only did this experience improve my hard coding skills (c# R SQL) but it also gave me an understanding of large scale Monte Carlo sim models such as the ones employed by fivethirtyeight for all their club and international soccer models. I think large scale Monte Carlo sim models have a lot of possible applications in the football analytics world.

    
   Mar 30
8. 
   

   

   Interesting, you then moved in to working at the University of Oregon Athletics department in data analytics. What sort of data is available to you at college level? Have you been able to implement anything you are proud of?
   Mar 31Sent
9. 
   

   

   As the first ever data analyst in the department there are a lot of positives. I have been able to start building out the use of data for all departments (marketing, development, ticket sales etc) from the ground up, which is both intimidating and exciting. Movement on many fronts is slow but we’ve been able to find some low hanging fruit that I’m proud of in the first few months of the job. Like many analysts who are one analyst of very few or the only analyst in a department, I’ve found collaboration outside our department has been key for our planning and use of data as a department. I’ve also become the de facto CRM manager in my department which is a whole job in itself, taking away from some of the larger models I was hoping to build out. I think I might start looking for some interns from the university to pick up some of my slack haha!

    
   Mar 31
10. 
    

    

    And on the footballing side I think you've done some work looking at passing networks? What is involved in that?
    Apr 2Sent
11. 
    

    

    Yes so for my thesis I created a rating using a network theory concept called degree centrality. I built a weighted network individual and team score that lives on a sliding scale based on where on the field the pass was completed. The rating also accounts for how many unique teammates the given player has passed to throughout a match. I used the 2012 MLS season of f24 files for my study. I am working on getting my .r file up on github and my thesis idea up on my blog. The pinned tweet on my Twitter page is an example of my passing network graphic that uses average passing position for each player to build a weighted network. I can build that graphic for any f24 file so those of you reading this that want to see a certain team’s passing network that you may have a f24 file for, feel free to get in touch with me and build a passing network graphic of for the game.

     
    Apr 3
12. 
    

    

    That sounds really interesting, but I've got a more provocative question...what would you say to people who don't see the value of this sort of information?
    Apr 3Sent
13. 
    

    

    Passing networks are certainly not a part of mainstream media yet. They still exist verymuchso in twitter football analytics limbo. Some analysts may look at my version of passing networks and see use cases for studying teams. I still wouldn't feel comfortable bringing my passing networks into a boardroom and trying to explain how they are actionable. The reason I started working on passing networks is my background in network theory analysis. I find passing a fascinating element in football that needs to be further threshed out through analytics in order to gain insights into how one team's passing may be superior to another's. Passing is a facet of the game that at face value is simple. Progress the ball towards the opponent's goal in order to score. Passing in itself is extremely telling as far as the style a manager and team hope to play. I had a really interesting conversation with @karun1710 about the current state of passing networks. We both agreed, though not currently extremely actionable in a team boardroom, passing networks give us insights into team strength, directional habits and strong connectivity between players. My version of passing networks, which use average passing position, allow for improved insights on a basic static graphic. Strictly looking at number of completed passes or basic possession statistics, is extremely pointless when trying to understand the game as a whole. Maybe passing networks, as an extension of basic passing stats, are the next step in what could be a breakthrough in this element of the game.

     
    Apr 5
14. 
    

    

    I can certainly see a use for passing networks. I'd like to see more done with passing at various game states. You only have to watch a match to see that losing teams typically pass longer when time is running out in a game. However, I'm sure there are more subtle changes you can't see. The idea of being able to load up the F24 data for each opponent in every match, and run algorithms to see what you can pick out would surely be a great help for coaches? Of course you'd need coaches involved in designing the algorithm and once designed they would only have value if kept private. I guess this is the big frustration for those outside the professional game?
15. Apr 7Sent
16. 
    

    

    When I first started exploring a career in football analytics, I really thought I wanted to work for a club. I still think that’s something I would be interested in in some respect but in many ways the best collaboration can be done at firms (opta, statsbomb 21st Club etc) and by hobbyists and bloggers (obviously within a club like barca there are exceptions to this). There’s this weird thing for everyone who works directly for a team in sports analytics where you want to attend conferences and show your colleagues that you are doing interesting groundbreaking work, but you also have stake in the game where you don’t want your innovation to be copied by other teams. I plan to publish my .r file publicly at some point in the near future, but that is because I want to collaborate with other analysts. If people use my ideas, I would be happy that I was able to help them achieve what they were hoping to accomplish as people helped me form/code my idea. I eventually want to enter into the football analytics world as an analyst at a firm, but right now I’m just enjoying the football analytics world with no real stake in the game. To your point, I don’t get to collaborate with a coaching staff, but I also don’t have to produce things that are actionable for a current season. This allows an analyst to form big ideas that push the envelope