Goal scoring analysis – Continuation
In a post last week I discovered that you could make money by placing bets on over 4.5 goals in all 30 European League games last Thursday (1 unit bets on all games resulted in a 8.8 units win in total). My plan is to find a long-term strategy to bet on over 1.5, 2.5, 3.5, and/or 4.5 goals in two-legged home/away matches. Here are some plots that may be used to analyse which factors to consider when building a model.
First, it seems reasonable that there are, in general, many goals in games with a big favourite. In the following figure the inverse of the favourite’s odds (which can be interpreted as Unibet’s prediction of the probability of a favourite win) vs number of goals is plotted. The linear trend is also included. It can be seen that the trend is clearly positive, meaning that high probability of favourite win correlates with many goals. However, and as expected, the spread is large (the sample correlation coefficient is estimated to r=0.29, and it cannot be established that a positive linear trend truly exists) and more factors than favourite odds must be taken into consideration to build a good enough model.
Obviously, it also seems reasonable to assume that games with many goals in the first round will also involve many goals in the second round. Here is the number of goals in leg one vs leg two plotted. As I expected a positive linear slope is observed (r=0.41). Observe that each dot in the plot might represent multiple games.
A simple and interesting betting strategy could be to bet on more goals in leg two than leg one (but at most 4.5 goals in leg two). An example strategy would be: Leave out the two games with no goal in the first round. For the four games with only one goal in round 1, bet on over 1.5 goals in round 2 (all four of these games had more than 1.5 goals). For the games with two goals in leg one, bet over 2.5 in leg two (eight of ten games fulfill that). For the six games with three goals in leg one, bet on over 3.5 goals in leg two (only one of the six games had more than 3.5 goals). And for the remaining eight games, bet on over 4.5 goals in round 2 (which actually happened in four of the eight games). If one unit is placed on Unibet in each bet, this strategy would end up in a 7.19 units win altogether.
Finally, I believe that if the favourite team lost the first leg, they will attack more and therefore score more goals in the second leg (compared to if they would have won the first game). At the same time, if the favourite team won big in the first game they are probably much better than their opponent and might win big also in the second game. In the x-axis in the below figure the goal difference in the first leg is given. A positive goal difference indicates that the team that are favourite in the second leg won the first leg. It can be observed that two games with large goal difference in the first leg (five and six goals) also had many goals in the second leg (nine and six, respectively). Apart from that no real trend can be observed (the positive slope of the fitted line is heavily influenced by the two goal-rich games).
I will continue with this (and other) type of analysis as soon as my data base increase. So far no general conclusions can be drawn due to small sample sizes.