From Ice Hockey to Football: Mean Even Strength (MESF)
From time to time I like to look at other sports in the world and see what we — football enthusiasts — can learn from other disciplines in elite sport. By doing that we can learn about innovative ideas that we can transition into football, but also to recognise that some models related to data, are already working very well.
In December 2024, I wrote about average attention draw and defensive entropy, defensive models and metrics from Basketball. In my articles I explained how to translate this research into football and how we can use it:
In this article, however, I will look at a different sport: Ice Hockey. And, yes I call it Ice Hockey, because where I live just hockey is called what’s officially field hockey. Are you with me? — Anyway, I want to use Mean Even Strength in Hockey (MESH) to see how we can translate, emulate and perhaps improve towards football.
Mean Even Strength in Hockey (MESH)
In hockey, “even strength” refers to gameplay when both teams have the same number of skaters on the ice, typically five skaters plus a goaltender for each team (5-on-5). This is considered the standard playing condition during most of the game unless one or both teams are serving penalties, which leads to power play or penalty kill situations. Even strength is crucial for evaluating a team’s overall performance since it excludes the influence of special teams.
Several key statistical variables are associated with even strength play. One of the most important is ESG (Even Strength Goals), which measures the number of goals a team scores under these conditions. ESGF (Even Strength Goals For) tracks goals scored by a specific team, while ESGA (Even Strength Goals Against) counts goals allowed by that team at even strength.
Data collection
Before we are going to look at how we can transform the data and metric into football analysis, let’s have a look at the data we are going to need in this specific little research.
We are using shot data for this particular article. The data comes partly from Opta and partly from my own model and was collected on Thursday, February 6th, 2025. The data focuses on La Liga 2024–2025 with emphasis on players who have played in over 5 games, to make it more representative.
The expected goals model are of my own making and are therefore different from other data providers such as Opta, StatsBomb, Wyscout or else.
Translating data
Football of course, doesn’t have power play so we have to find something that we can use as an even strength state. I have chosen to opt for gamestate. The gamestate can either be winning, drawing or losing for the specific team we are looking for or player we are looking at. By doing so, we have an even strength based on gamestate, when we use draw.
We can look at different sort of data, but I’m going to look at expected goals data because I want to focus on shooting. You can focus on other metrics as well in the same gamestate, but for me this is what I sought out to do.
Methodology
With the selected data I have I will calculate for the filters I’m using:
- Gamestate == Draw
- xG
- From RegularPlay
As you can see I also selected from RegularPlay or open play. I have made sure to not look at set pieces. They are not nearly the same as power plays of course, but they are not standard and can be infrequent as well, so that’s why I made that decision.
If we have that data, we are going to calculate the mean. What is the mean?
The mean is the mathematical average of a set of two or more numbers and be seen as the average. The mean provides a quick way to understand the “typical” value in a data set, but it can be sensitive to extreme outliers.
I calculate the mean of different variables in Python and Julia, which you can find on my GitHub. I then get the average xG and PsxG for the players and teams in La Liga under the event strength. The new metric is called Mean Even Strength in Football (MESF).
Analysis
As you can see in the bar graph above, I have the mean strength value for the xG per shot in La Liga 2024–2025 so far. We can see a few interesting things:
- Barcelona has the highest xG
- Sevilla and Getafe have the lowest xG
- The most common value is 0,11 xG per shot (6 times)
- Real Madrid surprisingly has 0,10 xG per shot
When we look at the players we can see a few interesting things too for xG per shot for MESF values:
- Hugo Duro, T. Douvikas and Borja Iglesias all score above 0,3 xG
- After Koundé, the xG stabilises, so the top 4 are outliers
Final thoughts
Even strength can show us a more even playing field — literally — and can measure how impactful a player or team is when there are even variables. That leads us to that players are less likely to benefit from more difficult or easier situations.
In the future I will look at the even strength and expand it to other areas of play like passing.
