Introduction What is a ‘neutral’ gamestate? The two most common definitions I’ve seen for “neutral” gamestate are periods of match time when the score is tied, i.e. goal difference of 0 the absolute goal difference (GD) is 0 or 1 ...

Introduction “The best predictor for future performance is Expected Goals”, authored by Sander Ijtsma of 11tegen11 in 2015, is one of the most notable articles in public soccer analytics. The write-up provided compelling evidence for the superior...

Introduction Towards the end of each soccer season, we naturally start to look back at player stats, often looking to see who has performed worse compared to their past seasons. We may have different motivations for doing so–e.g. we may be t...

Introduction Towards the end of each soccer season, we naturally start to look back at player stats, often looking to see who has performed worse compared to their past seasons. We may have different motivations for doing so–we may be trying to a...

Introduction I’ve written a lot about expected goals (xG) in soccer, but I haven’t yet talked much about possession value (PV) models1, another big topic in soccer analytics. What are they? Well, every PV model is different, but they all generall...

Introduction I’ve written a lot about expected goals (xG) in soccer, but I haven’t yet talked much about possession value (PV) models1, another big topic in soccer analytics. What are they? Well, every PV model is different, but they all generall...

Introduction If you’ve played (American) fantasy football, you’ve probably felt that you have been unlucky and unjustly lost matches at some point. “I had the second most points of any team in my league this week, and I only lost because I happened to play the team that ...

Introduction If you’ve played (American) fantasy football, you’ve probably felt that you have been unlucky and unjustly lost matches at some point. “I had the second most points of any team in my league this week, and I only lost because I happened to play the team that ...

Introduction “Should we account for team quality in an xG model?” From a purely philosophical point of view, my opinion is “no”–I think that xG models should be agnostic to player and team abilities, even if we were to find that accounting for sh...

Introduction “Should we account for team quality in an xG model?” From a purely philosophical point of view, my opinion is “no”–I think that xG models should be agnostic to player and team abilities, even if we were to find that accounting for sh...

Introduction Soccer is a game defined by its discrete, low-scoring nature, where the dynamics of a match are often dictated by the “game state”. Whether a team is leading, trailing, or level with their opponent at a specific moment can often infl...

Introduction Soccer is a game defined by its discrete, low-scoring nature, where the dynamics of a match are often dictated by the “game state”. Whether a team is leading, trailing, or level with their opponent at a specific moment can often infl...

Introduction Ever grappled with a classification model that consistently over-predicts or under-predicts an event? Your first thought might be to re-evaluate the model’s features or framework. But what if tweaking the model isn’t an option, eithe...

Introduction Ever grappled with a classification model that consistently over-predicts or under-predicts an event? Your first thought might be to re-evaluate the model’s features or framework. But what if tweaking the model isn’t an option, eithe...

Introduction This blog post demonstrates how to calculate the discrimination and stability “meta-metrics” proposed by Franks et al. (2017) for an array of soccer stats. Before the computations (“Inference”), I start by defining these meta-me...

Introduction This blog post demonstrates how to calculate the discrimination and stability “meta-metrics” proposed by Franks et al. (2017) for an array of soccer stats. Before the computations (“Inference”), I start by defining these meta-me...

Updates Wording in this post was updated after the original posting. Originally an outperformance ratio greater than 1 was mistakenly called “underperforming”, and a ratio less than 1 was mistakenly called “overperforming”. The wording ha...

Updates Wording in this post was updated after the original posting. Originally an outperformance ratio greater than 1 was mistakenly called “underperforming”, and a ratio less than 1 was mistakenly called “overperforming”. The wording ha...

Introduction Recently, I pointed out what seemed to be a bug with the expected goals (xG) data shown on FBref. In particular, the difference between non-penalty goals (npG) and non-penalty xG (npxG)1 seemed to be an outlier for the 2021/22 season...

Introduction Recently, I pointed out what seemed to be a bug with the expected goals (xG) data shown on FBref. In particular, the difference between non-penalty goals (npG) and non-penalty xG (npxG)1 seemed to be an outlier for the 2021/22 season...