In Seventh Heaven
In Seventh Heaven
If you read my previous article, you will be aware that I once again offered up a raft of English Premier League goal picks in order to continue my pursuit of data driven betting excellence and with the dust now settled, it is time to see how I fared.
With a slightly curtailed fixture schedule last weekend, it meant that there were potentially only eight games to choose from and once the numbers had been crunched, the percentage probability looked something like this:
Seven matches up for grabs, with Newcastle and Burnley not clear enough in terms of probability – something that turned out to be a good thing, when you consider that it was an absolutely dour 0-0 draw.
While of those seven, six of them were the correct outcome – the only game that let me down was Norwich vs Leicester. Therefore, it must be said that this was a pretty decent weekend and more importantly it cleared the target of five correct picks.
Which when you consider the percentage probabilities for the games in the green section, those games were not necessarily a clear result in terms of Over 2.5 – as usuaully games are nearer 60-65% for this benchmark.
Still with that said, there is little complaint from me and that means that the overall strike rate has increased and more importantly, it has increased with the sample size as well, something that further validates its relative success.
WHERE ARE WE AT?
This means that after testing the model on a total of 74 recommended bets, 51 of them have been correct, giving me an overall strike rate of 68.9% – up 1.8% from the previous week and further solidifying that we’re onto something.
Of course, we have not hit data driven football predictions utopia just yet, but with this being the first season of operation, it is a little difficult to ascertain what the exact benchmark of quality should be – therefore the whole 19/20 season will be our control run for next season.
However, if these good weeks continue, it does put us in good stead for next time around and that’s because the logic does not change for ‘GoalBot’ – it is a simple measure of goal probability, unlike when trying to predict final match results.
This means, that there could be a slight variance or you could run another model with perhaps a different set/and or additional criteria (perhaps running parallel) but ultimately the logic will always stay as is
And if it’s turning in nearly 70% success, that means it is something that is can be ported across all the other models that I have (Euro Major, Scottish) for next season and the more I think about, the more this is where the focus should go. Hopefully there’s more of the same this coming weekend.
If this has grabbed your interest and you would like to discuss/feedback then please feel free to drop me a message at firstname.lastname@example.org. 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.