What an exciting start to the season. At nearly three goals per game, there have been plenty of great matches. Outlandish score-lines such as 8-2, 1-6 and 3-5 have led commentators to scratch their heads.
Short Term Variation
The power of statistics is that it accounts for short term variation. When information is limited, we should be wary about drawing conclusions from our observations.
So when we get an “unusual” result, such as 2.98 goals per game so far this season, we can use statistics to tell us if that really is a surprise, or if it’s to be expected – after all, we’ve only seen 99 games.
How to Predict Goals Scored
We use our team strength model to predict Premier League matches. Total goals is something we can predict. At the start of the season, for example, we predicted Liverpool to score 1.69 and concede 1.02 goals at home to average Premier League opposition.
So on average, Liverpool home matches will produce 2.71 goals. But that’s only an average – any individual match may produce 3, 4, 5, or even 10 goals. In this example, a “Liverpool V Average PL Team” match has a 22% chance of producing three goals and a 19% chance of producing two goals.
I took our pre-season team ratings and simulated total goals this season 10,000 times. What was the result? We predicted that the 99 matches so far produce 2.82 goals per game on average. That’s may be a surprise in itself – in fact we have consistently predicted the Premier League to produce more than 2.5 goals per game since 06/07. Suddenly 2.98 doesn’t seem such an extreme number of goals.
We can also look at the number of times more than 2.98 goals per game were simulated. The chart shows that it’s not unusual to see 295 or more goals in 99 matches. In fact, 17% of our simulations resulted in 295 or more goals.
In other words, “2.98 goals per match is not statistically significant. We would expect a more extreme result 17% of the time, simply due to natural variation in goal scoring”.
Trends in Time
We have been predicting more and more goals in the Premier League since 06/07. The chart shows how many goals we predicted and how many were actually seen. We were surprised about the dip in goals in 08/09 but, apart from that, the predictions and the observations have trended upward.
The final chart shows the the distribution of goals for 06/07 and 11/12. Even though 2.98 goals per game is no great surprise today, it would have been highly surprising – predicted never to happen – in 06/07.
We’d love to hear your feedback – let us know your thoughts in the comments section.