For today's exploratory data analysis, let's look at the stats from the German Bundesliga and see how they compare with the leagues covered earlier - EPL and La Liga.
| Total Goals | Count of Match | % |
|---|---|---|
| 0 | 20 | 6.9% |
| 1 | 43 | 14.9% |
| 2 | 80 | 27.8% |
| 3 | 60 | 20.8% |
| 4 | 44 | 15.3% |
| 5 | 22 | 7.6% |
| 6 | 13 | 4.5% |
| 7 | 4 | 1.4% |
| 8 | 2 | 0.7% |
| Grand Total | 288 | 100.0% |
Looking at the distribution chart of total goals in the Bundesliga, it's clear that the most common score is two-goals and again, i would have thought that more games would end with less than 2.5 goals.
Instead, the outcome is more even, with 50% of matches finishing below 2.5 goals and 50% finishing more than 2.5 goals.
| Under 2.5 | Over 2.5 |
|---|---|
| 49.7% | 50.3% |
The Bundesliga is a lower scoring league than the premier league but on par with La Liga. I did a cursory check of blindly betting on odds greater than 2 when it comes to Over/Under 2.5 goals and no... that's a bad idea.
| Full Time Result | Matches | % |
|---|---|---|
| A | 78 | 27.1% |
| D | 82 | 28.5% |
| H | 128 | 44.4% |
| Grand Total | 288 | 100.0% |
As expected, there is a home ground advantage. Draws are also more common in the Bundesliga with 28.5% of matches ending in draws compared to only 23% in Spain and 27.5% in England.
For consistency, let's define favorites as having odds to win at less than 1.8. There have been 79 games where the home team was favored to win according to the bookies. Here's how they ended:
| H | D | A |
|---|---|---|
| 52 | 20 | 7 |
In terms of predicting the matches accurately, it seems like the bookie are doing a good job. But does their advantage leave anything for punters. Here's your return if you had bet a dollar on H/D/A for each of the 79 matches
| Initial Outlay | H | D | A |
|---|---|---|---|
| ($79.00) | $76.08 | $91.20 | $44.75 |
Again, just as with the premier league, betting on the home favorite would have netted you a loss despite 66% of matches ending as wins. If nothing else, i think the exploratory analysis done on the three leagues have shown that blindly betting on the favorite team is a money loser.
Betting on draws remain positive, yielding a 15% return in this scenario.
Something i'm working on right now is trying to predict draws better. One of the simpler inputs is to use predicted low scoring games as an input. Filtering for games where the odds of < 2.5 goals is less than 1.75.
This leaves us with a pool of 46 matches. Of these, 16 ended as draws. This works out to 35% instead of 28.5% in the general population. Betting on draw for each of these matches would have given you $53.42, or a $7.42 profit. This works out to be a 16% profit over the course of the season.
An important question we have to ask ourselves is whether filtering helps at all? If we were to have bet a dollar on a draw for each of the 288 matches, our return would have been $306.89. This works out to be a $18.89 profit or 6.55% profit. While filtering reduces our total winnings, it does help to increase the winning percentage.
For a more accurate model, i'm looking at incorporating H2H, number of draws in previous seasons. What is encouraging so far though is that predicted low scoring games have a higher chance of ending in draws based on the EPL and Bundesliga.
A quick check on La Liga shows that this strategy would be bad there and that's probably because the number of draws there are rather low.
All data from football-data.co.uk
