The Model

Let’s start by asking this question. Let’s say Chelsea’s spending of 500 million euros didn’t devolve into one of the Premier League’s all-time collapses. How should it have worked out for the Blues?

# A tibble: 4,000 × 1
      `1`
    <dbl>
 1 -34.7 
 2 -32.6 
 3 -24.7 
 4  -4.69
 5  -8.48
 6 -22.0 
 7 -31.2 
 8  -5.35
 9  -7.73
10 -26.0 
# ℹ 3,990 more rows

To create this graph, I used the rstanarm package which allowed me to create a fit using stan_glm, which gave information on the relationship between a club’s transfer balance and the points they gained or lost. Using that fit, I told the model that the club had in theory spent 500 million euros and asked it to predict the points they would gain or lose.

Based on the last season, the model dictates that the team would lose around 10-20 points from the previous year. I was skeptical though, due to Chelsea’s season being very weird for a normal team. So I removed Chelsea from the dataset we gave the model and recalculated.

After removing Chelsea from consideration, it becomes more evident that based on the last season, spending big doesn’t lead to big results. The model says while Chelsea’s season was unlikely, it wasn’t improbable. The team should have points, but probably not as many as they did.

What does that say about how to manage a football team? It says you shouldn’t try to spend big to build a better team. Instead, focus on selling off the players you don’t need and reinvesting that money, while maintaining a positive balance. Oddly enough, this is what Chelsea is doing now.