“Failing the possibility of measuring that which you desire, the lust for measurement may, for example, merely result in your measuring something else - and perhaps forgetting the difference - or in your ignoring some things because they cannot be measured.” - George Udny Yule (Kendall, 1952)
The Canadian Men’s National Team (CanMNT) will begin a new chapter with two preparatory friendlies ahead of the Copa America. With Jesse Marsch at the helm for the first time, in what is also his international managerial debut, Canada will face the Netherlands on June 6th followed by France on June 9th. The calibre of this opposition is high with the Netherlands and France representing two top 10 nations in the FIFA rankings. As Canada is currently ranked 49th, many wonder how this iteration of the CanMNT will fare at this level of play.
This post will serve primarily as a brief exploratory visual overview of the selected Canada team and how it may compare relative to their opponents. A secondary objective is to make good football use of some great packages such as reactable and reactablefmtr.
Measuring team strength in football is a big challenge yet measuring team strength in a valid and reliable way is an even bigger challenge. By using simple public data sources, the table and figures below will attempt to paint a logical picture with the hope of having some tolerable level of face validity1. However, this is a reductive approach that fails to capture a number of key components that contribute to successful team and individual performances. A player’s ability is multifaceted and dynamic. It cannot be accurately captured by a few digits on a spreadsheet or a few points on a plot. It goes without saying that there are more sophisticated and robust methods of estimating football team strength that go beyond the scope of this article. As the saying goes…“all models are wrong but some are useful”(Box, 1976). I’m not sure this approach is useful at all, but it may pass the just-for-fun-blog-post-test.
The table and plots include data from FM Inside and The Opta Analyst. FM Inside is a public source derived from the popular Football Manager game. FM Inside includes an ability rating for each player on a scale of 0-100. Details regarding the makeup of their proxy variable for measuring individual quality are scarce (this Reddit comment indicated that they won’t declare the secret sauce). Ideally, I would substitute this with a likely more reliable source such as Twenty First Group’s player ratings model but this data is outside of my reach. Opta provides public data for worldwide club ratings. For details regarding Opta’s model see Introducing Power Rankings: Your Club Ranked.
The following packages were used to produce the table and figures in R:
ggplot2tidyverseshowtextggtexthtmltoolsreactablereactablefmtr
(Cheng et al., 2024; Cuilla, 2022; Lin, 2023; Qiu, 2022; Wickham, 2016; Wickham et al., 2019; Wilke & Wiernik, 2022).
Select a table column title to sort in ascending or descending order. For mobile viewing, drag the table from right to left to reveal additional columns.
Canada Squad: June friendlies
Last update: June 4th 2024
Data from June 4th 2024; NT: National Team; APPS: Appearances
Data from June 4th 2024
Figure 1
Figure 2
Figure 3
What does it all mean? Looks like what we knew before. Canada likely has a group of players who, in many instances, may not be as experienced and skilled as their upcoming opponents. Quelle surprise. Good luck to Jesse Marsch and Les Rouges.
Far better an approximate answer to the right question, which is often vague, than an exact answer to the wrong question, which can always be made precise (Tukey, 1962).
Christina is a football (soccer) data analyst based in Canada. To get in touch: 6ysblog@gmail.com
References
Footnotes
Face Validity: The extent to which a measure appears to assess the construct of interest. This is not a formal type of validation or part of the psychometric development or evaluation of a measure (Kazdin, 2003).↩︎