The objective of this tiresome exercise is to find out how the clubs have fared based on their average ball possession. Method: 1.This excel (you can download it from HERE ) contains the possession vs goals scored numbers for of the BuLi clubs for the 2009-2010 season. Each sheet has the record of each club. 2. The assumption made is that each club will score a goal only when they have possession of the ball. Self goals, obviously, are exceptions to this rule, but I am neglecting them because their proportion is negligible. So, for each team, (eg Bayern Munich) I note the opponent of each match day, Bayern's possession and goals scored by Bayern in that match. You can see that in the excel given above. 3. After compiling the possession vs score for each team of the Bundesliga for the entire season, I collate the data as shown in the figure below. In the excel sheet, it is on the sheet titled 'main'. http://img708.imageshack.us/img708/6094/tableac.jpg (please zoom in to view clearly) 4. You can see in this table, that there is a column called R (Average Possesion/Total Goals scored in the season). This is simply how likely it is for a team to score in 1 unit time of possession. The other ratio that you can see in this table is the Poisson coefficient of possession efficacy. To compute that for a club, for each game I have taken the goals/possession ratio and taken an average across the season. The reason I am calling it a Poisson ratio, is because I expect this to be distributed in a Poisson pattern for clubs. I am unsure of this. This ratio has just been computed for now, and hasn't been used at all. It maybe used in prediction purposes. In a nutshell, what is done is that there are two ways of taking a possession parameter. One is taking an average of all possessions and then taking a ratio. The other is taking a ratio and then taking an average. Since, I am not professionally trained in statistics, I am unable to decide which is the better option and hence have gone ahead with the first one (i.e the non Poisson option). In the image, its the highlighted column. 5. I have co-related this possession parameter with the points tally of each team at the end of the season. Here is the chart (zoom to view clearly) Inferences & Findings: 1. Bayern has the highest average possession and Bochum the least. 2. Werder and Wolfsburg have the highest possession to goals ratio this season in the BuLi. 3. Cologne, Nuremberg and Hertha make the least of their possession. Hertha is relegated so I won't talk about them. In spite of having a near 50% possesion, Nuremberg have scored just 32 goals this season. This indicates that their possession is more in their half/ back passes and officially (or at least mathematically) they are the most boring team of the season. 4. The line in the graph is the best fit line for R (possession ratio vs points). Going by this metric, Wolfsburg, Werder and Leverkusen have the best season. They have gotten maximum points for the possession/goals ratio they've maintained this season. 5. Bayern on the other hand, have slightly fallen below the Bundesliga average. 6. Hertha, Cologne and Mainz perform poorly on this metric. Your thoughts? Any suggestions?Any other inferences? Should I use a non linear regression equation between R (possession ratio) and points and then see which clubs lie above and below the regression curve??
Interesting analysis. I would argue one small point - shouldn't teams below the line of best fit be the "boring" ones, and teams above the line be the "exciting" ones? It seems to me that a team with lots of possession but not many points is like Arsenal - nice football but no results, whereas a team with little possession but a lot of points might be playing longball tactics - playing for the result, not for nice football. This matches the anecdotal evidence - Leverkusen and Werder are considered two of the more attractive sides, while Schalke is generally considered pretty boring. It also makes sense that the underachieving teams - namely Hertha - would be playing more boring football. After all, if a team like Bochum gets relegated, it's to be expected (well, except they call themselves the "unrelegatables", but objectively speaking), so they can be a bit more gung-ho - but Hertha getting relegated was pretty much a disaster.
As a whole, this is an interesting analysis. However, I would disagree with this point. While Wolfsburg, Werder, and Leverkusen had the best goals/possession ratios, they look to have gotten less points than they should have based on the graph. Based on that regression line, for Werder and Wolfsburg to have gotten maximum points, they would both have finished with about 72 points. Instead, they finished with 61 and 50 points, respectively. Bayern and Schalke got the maximum points for the goals/possession ratio they maintained. If they had gotten the expected number of points for their goals/possession ratio, it looks like they each would have finished with about 5 less points.
Can't really generalize that those below the best fit line are boring. The problem with that generalization is that some of those teams have scored ( cough cough Bayern) the maximum number of goals in the BuLi. The other problem with such a sweeping generalization is that many a times after a stronger team goes into a lead, it snuffs out the life from the opponents play by keeping possession. This adversely affects the R of the team that is leading and the one that is trailing. The former because the possession time increases without any increase in goals. The latter because they neither score nor keep possession. This is a hypothesis of course. I will go through some match records to see how much of it is actually true. You raise a very valid observation. I like your interpretation better than what I had earlier concluded. The only explanation that I have to offer is that, maybe I shouldn't have gone for a linear best fit model. Just maybe then this discrepancy would be addressed.
I downloaded the file and played with the data a little bit myself-I think you were right to do a linear best fit model. It would be interesting to see why it is that Bayern and Schalke got more points than expected while teams like Werder and Wolfsburg got less points than expected. I didn't watch the Bundesliga on a regular basis last year so I'm not qualified to give a good answer. Any thoughts?