Pre-split statistics

Now that we've reached the split, I thought now would be a good time to put together a brief analysis of the season so far, since all teams have played each other an equal number of times.

A note on calculations: the "raw" over/underperformances given for each team all consider whether each opponent was played twice at home and once away, or vice versa, using each opponent's average points conceded at home and away.

For example, Aberdeen's biggest raw overperformance was against Hearts, where they picked up 1.4 points more than the league average despite only taking 3 points in total across the three games. This is because they travelled to Tynecastle twice, where Hearts only conceded an average of 0.235 points per game. Therefore, a team playing away at Hearts twice would take an average of 0.47 points, while taking an additional 1.125 points against Hearts in their single home game (due to Hearts conceding that average number of points when playing away), making a total of 1.6 expected points across the three fixtures. This is why Aberdeen's 3 points represent an overperformance of 1.4 points vs. the league average.

The "weighted" performance statistics take into account both home and away differences *and* the relative strength of the team in question by applying a strength multiplier to their opponent's average conceded points. The strength multiplier is calculated by simply dividing each team's points per game by the league average points per game. To use the same example as previously, Aberdeen's strength multiplier was 0.737, meaning they accumulated 0.737 points for every 1 point accumulated by the league average team. Therefore, they would only be "expected" to gain 0.737 times whatever their opponent's average conceded points are for a given fixture. This often, but not always, results in a different opponent being identified as the most significant over-or-underperformance.

I am by no means accomplished in data processing and visualisation, but I hope you all find it interesting!