Expected results Primera División Argentina (Fecha 13)

In this blog I give probabilities for games in the Argentine Primera División, show a predicted league table based on those probabilities and give a list of bets that corresponds to the given probabilities (note: you are free to use the ‘tips’, but I am not using them myself!).

I have replicated an existing prediction model and I am optimizing it for the Argentine league. The model is the so-called pi-ratings model by Constantinou and Fenton (2013). In this article they describe a system in which they assign probabilities to match outcomes (home win, draw or away win) based on historic results.

Based on past results it calculates two dynamic ratings to each team in a model: a home and an away rating. For each new game the model calculates a rating for both the home team and the away team. Based on those ratings an expected outcome is given (for example that the home teams wins by 1 goal). If the home team outperforms this expectation (let’s say they win by 2 goals) both his home and away rating (remember a team has two ratings) will be adjusted positively. Primarily the home rating as the team was playing in that condition, but you can assign a percentage to which extent the away rating is also impacted. The reversed thing will happen for the team that could not cope with the expected result (and lost by 2 instead of the expected 1 goal). Another variable, the learning rate, determines how fast new results will change the old rating. Or in other words; if the team ratings put a lot, or fewer, emphasis on the latest results.

I have optimized those variables and the expected results for the Argentine league. The most important change I made is that instead of using goals I use expected goals.


This is the table of predicted points per team at the end of the tournament (16 games). You can see the pre-tournament prediction and the updated prediction after each game (taking into account the result of that game).

Note: one thing I noticed is the small difference between the worst and best teams. This is mainly due to the nature of the tournament: there are very few easy wins, anything can happen. Each of the last 6 tournaments have been won by a different team. Tournaments are in general quite short (19 games) and another example is that often higher rated teams play some games with B players (those past results are taken into account in the model). Both make variation bigger.

Predicted final points in tournament, updated after each game
Predicted final points in tournament, updated after each game

Those are the model’s probabilities for the upcoming week, with Racing Club and San Lorenzo the biggest favorites, and a, quite surprisingly, underdog role for Estudiantes against Gimnasia.




Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s