The SEW Sports Economics research group is happy to present its project, called Soccer Analytics. In this project we developed our own and independent forecast for the German Bundesliga. The forecast is produced using empirical methods of the so-called machine learning.
(last update 07.05.2018)
Updates of the results are announced on Twitter.
Model update February 2018:
The aim of Soccer Analytics is to test machine learning in a real setup. In the course of continuous development, a new method for the forecasts is used since February. The forecast of the final table has also been updated with the new methodology. Important for the comparability is that only information before the start of the season was used to build the prediction. The "old" predictions of the final table are still available under further analyses.
On this page we present answers to the following questions: What is the range of final rankings of a team? Who are the favorites and underdogs in each specific game of the upcoming round? How the final table of the season will most likely look like? Which teams are the positive and negative surprises of the season so far?
We provide short explanations how the predictions are calculated and provide more details under further explanations. Additionally, we present results for other interesting analyses. We provide further analyses and assess forecast quality.
Before the last matchday, there are still several duels for important season goals. Leverkusen (64%) against Hoffenheim (63%) for the Champions League. Leipzig (72%) against Frankfurt (15%) for the Europa League. Wolfsburg (78%) against Hamburg (14%) for the place 16.
Rank in the final table
The table below shows how likely every team finishes on a certain rank. To calculate these probabilities, we simulate over 10,000 different seasons by a computer program based on our prediction model. If you want to know how, please read our further explanations. The interactive graph shows how often each team finished in each position at the end of the simulated seasons. In further analyses, we illustrate these numbers in an interactive graph and show probabilities to achieve different seasonal goals.
Probabilities smaller than 5% are not displayed
ER: Expected ranks
EP: Expected points (graphical illustration in further analyses
↑↓: Difference in EP of the final table between current prediction and prediction before season (details in further analyses
Here, fans can check the probabilities according to which their team wins, looses or draws in the next round. It is often erroneously assumed that we predict the team with higher winning probability will win. This interpretation is misleading. You should think about the presented numbers in the following way: Imagine the same teams in the same situation play against each other 100 times. Then the reported probabilities show how many of these imaginary 100 games are expected to have the respective outcome.
Over- and underperformers
The graph below shows for each team whether its achievements so far are above (green) or below (red) the expectations before the season. The bars show the difference between actually achieved points and points that a team was predicted to achieve so far according to the model in the beginning of the season.
For questions and suggestions please contact us via email@example.com.