Prediction model Argentine Primera División (weekly updated)

I am constantly dealing with statistics from the Primera División of Argentina, and last week I was comparing shot averages, shot effectivity and save percentages. Doing this I started thinking about some of those statistics and how to apply them to predict the upcoming gameday. As fond player of the local Proode (you pay $2 pesos (€0,15) to predict 13 games, and in case you correctly predicted all the scores (home win, draw or away win) you share a $500,000 peso price with the other winners) I wanted to see what the value of the historical ratios was when forecasting the future games.

In the small model I used:

  • average shots per game
  • average shots on goal per game
  • average goals scored per game (minus penalties and own goals)
  • goals / shot on goal (%)
  • goals scored (minus penalties and own goals)
  • average shots received per game
  • average shots received on goal per game
  • saves / shot on goal (%)
  • average goals received per game (minus penalties and own goals)
  • Strength of the past opponents
    • To adjust the prediction of the shots and shots received against the upcoming opponent

One of the next steps would be to include historic match data (e.g. X – Y always sees a lot of goal, or X never lost against Y, etc.) and home advantage.

For Weekday 7 the model predicted the following scores (with updated results):

Boca Juniors – Racing Club 1-1 (result: 1-2)
Arsenal – River Plate 0-1 (result: 1-1)
Newell’s – Olimpo 1-1 (results: 1-0)
Belgrano – Vélez Sarsfield 1-1 (result: 1-0)
San Lorenzo – Godoy Cruz 1-1 (result: 2-1)
Atl. Rafaela – Estudiantes 1-2 (result: 1-0)
Gimnasia – Def. y Justicia 1-1 (result: 0-3)
Tigre – Rosario Central 0-1 (result: 4-1)
Independiente – Quilmes 1-1 (result: 5-3)
Lanús – Banfield 1-0 (results: 1-0)

All those predictions came out of the model as the score with the highest probability (in general between 13%-18%) for that game and does not necessarily reflect the prediction of the winner of the game.

(E.g. in the case of Boca, Boca was a clear favorite considering all probabilities, but the score with the highest probability was 1-1.)

Update: only 1 of the 10 results came out. However, for 10 of the 20 teams the model correctly predicted the amount of goals.

Update2:

The predictions for Weekday 8:

River Plate – Independiente 2-1 and 49% of under 2.5 goals (Result: 4-1)
Rosario Central – Arsenal 1-1 and 62% of under 2.5 goals (Result: 3-1)
Godoy Cruz – Tigre 0-0 and 81% of under 2.5 goals (MATCH CANCELLED)
Def. y Justiacia – San Lorenzo 1-1 and 51% of under 2.5 goals (Result: 1-3)
Olimpo – Belgrano 0-0 and 91% of under 2.5 goals (Result: 1-0)
Vélez Sarsfield – Atl. Rafaela 1-0 and 66% of under 2.5 goals (Result: 0-0)
Banfield – Boca Juniors 1-1 and 57% of under 2.5 goals (Result: 1-1)
Quilmes – Lanús 1-1 and 67% of under 2.5 goals (Result: 0-2)
Racing Club – Newell´s 1-1 and 70% of under 2.5 goals (Result: 1-1)
Estudiantes – Gimnasia 1-0 and 67% of under 2.5 goals (Result: 0-0)

Update: 2 of the 9 results came out. However, for 10 of the 18 teams the model correctly predicted the amount of goals.

The predictions for Weekday 8:

River Plate – Boca Juniors 1-1 and 43% of under 2.5 goals (result: 1-1)
San Lorenzo – Tigre 1-0 and 76% of under 2.5 goals (result: 0-2)
Def. y Justicia – Arsenal 1-1 and 55% of under 2.5 goals (result: 2-1)
Rosario Central – Lanús 1-1 and 67% of under 2.5 goals (result: 1-2)
Godoy Cruz – Independiente 1-1 and 51% of under 2.5 goals (results: 2-2)
Racing Club – Atl. Rafaela 2-0 and 57% of under 2.5 goals (result: 0-2)
Quilmes – Newell’s 1-1 and 61% of under 2.5 goals (result: 1-1)
Gimnasia – Vélez Sarsfield 0-0 and 76% of under 2.5 goals (result: 2-0)
Olimpo – Estudiantes 0-0 and 97% of under 2.5 goals (results: 1-2)
Banfield – Begrano 1-0 and 66% of under 2.5 goals (result: 2-2)

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