Football Result Predictions using Data Mining

With a little pride we would like to start by saying that we create our predictions completely independent of the predictions of others. In particular, we do not use betting odds for single matches provided by any betting office or a similar provider. We only use publicly available information for our statistical analysis using data mining.

We have collected and modified a lot of historical information relevant to football games. These data are analysed by us with the data mining software [Data.Mining.Fox (DMF ) by Easy.Data.Mining]. We do so in two steps. Step 1 generates a so-called model which is statistically calculated on the basis of historical data. In this model, the data mining software extracts all detectable, multi-factorial relationships of the historical data. And in step 2 we apply this model in order to predict the results of future football matches.

We perform this for all matches in three prediction cycles. Each cycle consists – as explained in the previous paragraph – in the creation of a model on the one hand, and the application of this model on the other.

In the 1st cycle, we always first answer the question of whether a game is a draw or if the home or away team wins. In the 2nd and 3rd cycle we then calculate how many goals the home and the away team score. From all three cycles together we then make up the predicted match result.

Note, of course , that neither statistically unpredictable circumstances nor short term changing factors can be reflected in our calculations – e.g. wrong decisions of referees, exceptional conditions of the playing field, injury or disease of players, light and weather conditions, red card bans of major players, effects of changing coaches, club disputes, financial problems of the association, unresolved contract negotiations, interruptions by fan riots, the extra burden some international players might have, etc. And about match manipulations we do not even want to talk ;-)

Our DMSC (Data.Mining.Soccer-Confidence) is a percentage for statistical confidence which provides insight about the statistical validity of our prediction. This percentage is an artificial hybrid ratio which is composed of several elements, for instance: the statistical confidence level calculated by the data mining software regarding the result class of the match output; the historic expectancy value for a home victory, draw or away victory; the average of the success rates that we could achieve for both teams in terms of all prediction to date; a quota calculated by us manually regarding the outcome factors home victory, draw and away victory.

Last but not least we would like to draw attention to our transparency. All predictions are made available before each match on this website. Our predictions remain transparent in direct comparison with the real results, even if our predictions might have been very poor one matchday (not always the case on other websites ;-)). We also provide statistical overviews regarding our prediction validity for a given category (i.e. usually for a season of a specific competition such as e.g. the Premier League).