In #football#Analytics we often refer to the analysis of attacks as sequence analysis, implying an attack can be modeled as a cascading sequence of events. A thread on why the term “sequence analysis” can be misleading.
Despite the popular analogy (often used to describe strategic moves), #football is actually nothing like a game of chess. In chess, one piece can be moved at a time, after which the opponent reacts. In #football, 22 pieces can be moved simultaneously.
The movements of the 22 pieces are strongly bound together, and #Science has shown that attacking play is not the result of individual, sequential, actions, but rather the product of inter-player & inter-team interaction.
Chess players think in terms of action-reaction, and the best ones can “think 3 steps ahead”, if this then that and that and that. Research has shown #football isn’t determined by action-reaction, but rather by coupled interaction.
Thus “think 3 steps ahead” means something entirely different for #football players. It means deriving patterns from interactions, and seeing opportunities to break the coupling with an opposing team.
This has large consequences for the way we analyze the game. If a reaction cannot be attributed to a single action, how would that impact #Analytics?
Take the example of a pass, to whom should we attribute the decision to pass, and the chosen destination? Should this be the passing player, the receiver? In reality the pass decision is largely determined by the interaction of 22 moving pieces.
Off the ball, teammates try to get open, opponents try to block passing lanes, largely determining the resulting action. Ofc individual abilities & quality play their part, but any action can be seen as the product of at least multiple players on different teams.
In other words, 1+1+1 != 3 in #football#Analytics, just like pass + dribble + assist != goal. An attack is not the product of the sum of individual actions, but rather the product of interactions. But does this mean for you shouldn’t be doing sequence analysis?
The opposite, sequence analysis can be very valuable, when used to identify moments of interest.
Certain actions are likely to signal key phases of an attack, and can therefore serve as a starting point for analysis. One should be very careful however when attributing causality
Take an accelerating line-breaking pass, or a key-pass leading up to a shot. These actions are typically the product of a key-phase of interactions, and analyzing the preceding 5-10 secs can reveal valuable insights.
In #Analytics these moment are often framed as the action causing a result, while from the perspective of interactions, these actions are a result themselves.
In other words, if you want to understand the attacking determinants of succes of a team, finding the player that gives most key passes will not hold much meaning in and by itself.
If however you’d go back in time and study the interactions leading up to the action, you can start to identify patterns that determine the succes of attacking play.
Using #DataScience techniques and position tracking data, you can come to very interesting insights this way, unraveling the underlying tactical patterns.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
“Data don’t lie”. But it typically requires a process of defining #research questions, hypotheses, methodology, interpreting and #dataviz that can introduce subjectivity and #bias. Scientific rigor and objectivity are key in #DataScience. Some #Tips for #DataScientists 🧵
Don’t dive straight into a dataset, domain knowledge is critical. Good #Science requires a theoretical understanding of a topic while #ignorance introduces bias. Sound domain knowledge enables you to ask the right questions and give relevant answers with #DataScience
Investigate the alternate hypothesis. Business questions asked to #DataScientists are often directive, as there already is a hypothesis. Don’t confirm this hypothesis without properly investigating the alternate option.
Tactical behavior in #Football has a spatial and a temporal component, and results from interaction with the opponent. It’s key to account for all these aspects in data-driven tactical analysis, as well as to respect the complexity of the temporal and spatial dimensions 🧵
Two years ago I published a systematic review in @EurJSportSci on using big data in #soccer for tactical performance analysis that illustrates the associated challenges and provides a data-driven scientific framework. #DataSciencetinyurl.com/mrxky6ca
The most common analysis issue is the fact that spatial and/or temporal complexity is not respected. For example by aggregating data over multiple minutes, or constructing spatial features aggregating 11 player positions into a single variable.
Preparing for a technical interview for a #DataScience position? These are some of the questions that typically allow me as an interviewer to quickly distinguish between juniors and mediors, including some quick tips 🧵. #Python#pythonprogramming#DataScientist#Jobs
All questions about SQL. Not the hardest thing to learn, but many #DataScientists only start to learn the value of SQL when they actually become part of a dev team. I’m not only talking about SELECT * FROM table, but also about joins, truncates, partitions and constraints.
Interacting with an API. Make sure you know your requests (GET, POST, PUT, DELETE, PATCH), as well as the #Python requests library.
#DataScientist in a software dev team and #pythonprogramming code for production pipelines? You should think carefully about scalability and integration. One of the things to consider is datatypes, here are some helpful tips 🧵
#Python is a dynamically typed language, but that doesn't mean you shouldn't care about types. Know you dtypes, from "str" to "bool" to "int8" to "float64", and understand their memory footprint and restrictions. Especially when working with larger objects, choose wisely.
Loose the strings. 9/10 times strings can be replaced by categoricals (Pandas) or even better by Enums (docs.python.org/3/library/enum…). This can reduce memory footprint of large dataframes with >30%, and improves performance.
Yesterday I shared a small thread about getting into #DataScience. Today I’ll build on that and share a bit about my own journey into sports analytics, specifically as a #DataScientist in the #football industry. 🧵
My path began with a MSc in Sport & Movement Science @VU_FBW. It’s not computer science or anything, but it does involve quite some #Math, #Statistics and #Physics, as well as a course in programming. Mainly it learned me Science, and gave me a lot of domain knowledge in sports.
I wasn’t planning to become a #DataScientist, but I wanted to work in sports. I did various stints as an embedded sports scientist, mostly internships/part-time, before joining @ZZLEIDENBASKETB. Those jobs involved data & science, but it wasn’t anything close to #DataScience.