Penny Wise Pound Foolish?

With the last derby now in the rearview mirror and the league wrapped up, I thought it a good time to begin the process of benchmarking the controllable versus uncontrollable aspects of the 2024-2025 season. Today’s column will begin with league play, where Celtic obviously currently stand on 85 points, with 103 goals scored versus just 23 conceded - not too shabby!

Of course, we amongst the spreadsheet shagging community are very interested in various performance measurement metrics and models, so to begin this exercise, I share with you the Wyscout expected points model since the 2015-2016 season.

But first, here is a quick reminder/explanation of what expected points (xPts) models attempt to measure:

Expected points quantify the number of points a team is expected to earn from a match, considering the probabilities of winning, drawing, or losing based on the quality of chances during the game.

It is a sort of hypothetical universe where each game is ‘scored’ using expected goals and done so on a proportional basis - i.e. if the relative xG in a game resulted in a 50% probability of a team winning, 30% of drawing, and 20% of losing, then their xPts for that specific game would be:

xPts = (0.50 x 3) + (0.30 x 1) + (0.20 x 0)

xPts = 1.50 + 0.30 + 0

xPts = 1.80

Let’s say the game result was 1-1, then the team would have had 0.80 more xPts than the actual result.