World Cup 2014 is going on at the full speed and we are witnessing a joyful tournament. Despite the time difference for the fans and climate difference among cities for the teams we had a chance to watch very good games. Some teams surpassed some boundaries and brought surprise to the group stage games. On the other hand some teams could not reflect their strength appropriately and head back to their homes early. All in all, every team deserves applause for bringing excitement to probably the most prominent football organisation of this year.
Following that, I would like to present an evaluation of the tournament statistics very broadly. During the group stage games 136 goals were scored and average of 2.8 is above the 2.3 goals per match statistics of the World Cup 2010. When it comes to red and yellow cards; averages dropped to 0.2 from 0.3 for reds and from 3.8 to 2.7 for yellows. These two indicators tell us that teams are playing more open and fair. Thus we observe more and more goals. Or it can also be interpreted as that FIFA encourages referees to foster more goals and less cards. The other two statistics reveal that tiki-taka is wide spread and teams improved their passes from the average of 353 to 384. Finally, we see that actual playing time drastically dropped from 69.8 minutes per match to 55.5 minutes. This can be due to the hot weather and mental change in the tactical play.
Enjoy the “Road to the Cup”!
An Evaluation of my Forecasts
Now the group stage is over and teams qualified will have a rest day, but there is no rest for supporters. As I disseminated my match-end result forecasts before the kick-off and previously added note about my forecasts, now it is time for to reveal how accurate those were. In the group stage 48 matches took place and I managed to hit 23 correct results on the winner or draw, and 2 matches of it with exact scores. Roughly 50% of success ratio! However, when you make forecasts as an economist it is not so easy, there are always complications. 😉
First of all, if you claim a result before the match and would like to measure the discrepancy of the actual result with your forecast you need to calculate a variation and standard deviation of your forecasts. But in order to keep things simple (as much as I can do) I developed two methods: Namely, “Goal Average” and “Goal Difference” respectively.
With the first method, goal average in a match is calculated both for forecasts and realisations then difference of these two has been taken. Therefore, if you claimed that match result would be 2-0 and it ended as 4-2, then it is possible to conclude that the home team was strong enough to win by 2 goals in average. On the other side of the coin it is not logical to claim that your forecast was accurate with such a calculation, because there has been six goals in the end while you claimed only two. Thus, I added another calculation parameter to consider this handicap. This time, difference of realized total goals and the estimated total goals in a match has been computed. As in the above example if you claim 2-0, then the number of total goals is two and if the result is 4-2, then the number of total goals is six. Consequently difference is 4 (6-2=4), which shows that your bet was not that accurate in the end.
Having this information at hand I can briefly summarize my forecasts with some graphs (You can click on the link for the PDF format of graphs).
In the first graph; horizontal axis shows 48 group stage matches in line with the match numbers appointed by FIFA, and vertical axis indicates the possible match results (1: Home team wins, 0: Draw, and 2: Away team wins). When it comes to content of the graph blue dots represent my forecasts and red circles represent the actual outcomes. In order to keep visualisation clear I connected actual outcomes with red lines. As it is obvious, some dots are within the red circles, which means that forecasts fit the actual outcomes. 2 of this overlaps are perfect and 21 other overlaps are correct by the categories, which makes 23 hits.
In the second graph, discrepancies have been screened by the two methods I already explained above. Horizontal axis is shared with first graph and vertical axis redesigned for the number of goals. In the graph; red bars represent goal averages and green bars stand for goal differences. As an example, for the perfect hits (match number 23 and 31) these figures are zero and for the first match there is only green bar. Because I estimated as 2-0 for Brazil and actual outcome was 3-1 which makes the goal average zero and good estimate by the strength but incorrect estimate by the misspecification of number of goals. Instead of 2 goals in total there was 4 goals which is indicated with the goal difference of 2 in the left axis.
As a result it is required to combine these two methods in order to penalise such estimation errors. So, I decided to use equally weighted average of these two calculations and showed the results at the third graph. On the left axis “0 goals” is the base point and indicated that forecasts are satisfactorily in line with the actual outcomes. In the graph discrepancies calculated with the weighted average of goal averages and goal differences are showed with blue dots and connected to each other with a blue line. Any dot above the base line indicates that forecasts were under-shooting of outcomes, and dots below the line indicate an over-shooting of forecasts. While the maximum value is 2 and minimum is -3 and most of the dots scatter around the base line with a small discrepancy. In other words, this means that I could have done better but I am still satisfied with my forecasts.
When it comes to Round of 16, I managed to predict 9 teams to qualify with two exact parings. From that standing, I have forecasted 5 out of 8 quarter-finalists with also two exact pairings. For the semi-finals, success rate is higher with 3 out of 4 due to underestimating Dutch power / overestimating Spanish tiki-taka.
Ergun UNUTMAZ, 27/06/2014