EURO2020 Predictions by konakalab
Prediction of EURO 2020
This prediction is based only on the international match results from 2018 to 2021/3. The data source is "International football results from 1872 to 2020" disclosed in kaggle.com.
Summary: prediction performance
The correlation coefficient between the predicted and the actual modified goal ratio was 0.638 ("Modified" means that g_ij=(g_i+1)/(g_i+g_j+2) where g_i and g_j denote the number of goals of teams i and j, respectively). These values has moderate correlation.
In 51 matches, the total W/D/L probabilities of highly-rated teams are 32.2/9.1/9.7. The actual values are 33/8/10. My prediction is very close to the actual results.
The winner and runner-up are Italy and England. They are ranked 4th (12.7%) and 2nd (16.5%) in my prediction. The combination is the 13th popular scenario (1.9%)
Outline of prediction model
This prediction model outputs Win/Draw/Lose probabilities based on rating values of the teams. The rating values are optimized to minimize predicted and actual modified goal ratio, i.e., (g_i+1)/(g_i+g_j+2) where g_i and g_j denote the number of goals of teams i and j, respectively.
Importance based on match date is included. Matches held 1 year ago have 75% importance. If the duration is n years, the importance is 0.75^n. No other importance is included for simplicity of the model.
0.75 is selected because this value could improve prediction performance for past FIFA World Cups 1998 to 2018.
This model can outperform to the official FIFA ranking.
Home advantage is also included. In closed matches, i.e, matches without spectators, home advantage is reduced, but is not disappeared. Therefore, the home advantage value is calculated based on the match results from 2020/1.
Prediction: Final Standings
Prediction: Most popular scenarios in our prediction
Written by Eiji Konaka, a.k.a. konakalab (https://twitter.com/konakalab ). Last updated:2021/6/24