A Quick Recap
A Quick Recap
Just a follow up piece from last week’s pair of English Premier League articles and it is fair to say that if one was on the way to Championship winning glory, the other was set for data driven betting relegation.
Both articles were written last Sunday and that meant they were before the matches between Burnley vs Arsenal and Tottenham vs Manchester City, which means I just need to give a quick recap on any further statisical developments.
In this first instance, let’s look at the Over/Under goal market and see if there was anything in the way of continued success over the course of Super Sunday and to do so, lets look at what “GoalBot’ served up last week:
Burnley vs Arsenal – Over 2.5
Tottenham vs Manchester City – Over 2.5
A pair of outcomes that were no good and that means after testing this model on 56 matches, 40 have been correct. This subsequently has given me a success rate of 71.42%, so as the sample size is increasing the percentage is getting a little smaller.
However, that’s not to be too unexpected, but obviously we don’t want to decrease too much further. If I can get it stable at around 70% then we’ll know that this certainly a concept that works, but perhaps there needs to be an element of A/B testing.
For example, why was Burnley vs Arsenal at Over 2.5 goals – hindsight will tell us that was far too big a leap of faith after such a drab performace at Turf Moor and the probability may have been skewed by conceding goals rather than scoring them.
I’ve built a model that will try and filter out any bad defences skewing the probability and I will look to put that to the test on the other side of this split winter break (although I will still have the normal goal data available)
MOVING ON QUICKLY
As for the predictive model ‘PremBot’ it was anything but a Super Saturday with just three out of eight picks correct. However there was at least a bit of respite after a draw at Turf Moor as the model did predict that correctly.
Unfortunately there was also a draw selected between Tottenham and Manchester City last weekend and that meant, the week finished with a paltry four correct picks. Although at least that was a small improvement on the week before.
Then again with just seven of the last twenty picks being correct, it means something is not under the hood or there have been a lot more upsets in terms of results. Unfortunately, I think it is more the former than the latter and I’m not having a good run in terms of data driven football predictions.
While with the conclusion of that week, it means I have now predictively modeled 100 matches and 46 of them have correct. Now you don’t have to be the biggest maths fan to know that that is a hit rate of 46%.
A return that you would more than likely be able to get from gut instinct alone (if not better). Therefore ‘PremBot’ is going to need some long awaited wholesale changes and this split week, might just be the time to do them.
Happy punting and thanks for reading. Dan
If this has grabbed your interest and you would like to discuss/feedback then please feel free to drop me a message at dan@realfootballman.com. While I am always looking for new football/data projects to work on and if you feel that my skills would be of use, I can be contacted at the same address.
Follow me on Facebook at Dan The Stat Man
You can also check out my Premier League Podcast on Soundcloud