The English Need Control
The English Need Control
In my previous article, I offered up the latest round of picks for the Russian Premier League and the preparation of whether or not ‘StatBot’ can return a profit for the third successive week in a row. While also in that article, I made reference of the fact that I need to start thinking about weighting.
The reason I have been pushed into this decision, is that even on gut feel alone, I am not confident that ‘StatBot’ is geared up correctly and I have a sneaky feeling, that the beginners luck is going to start to come to an end.
Which means we need a different plan of attack and it is one that comes in the shape of the English Premier League instead.
The best league in the world?
Now as much as would love to test a different league each week for money, I cannot and will not bankrupt myself at this early stage of proceedings and therefore when working away from the Russian Premier League, I am going to work on a pure percentage strike rate only.
With the theory being that with each week, I’ll track the success rate and if I can get it to a workable level, I can then scale across multiple leagues and more importantly pull the trigger on then betting for cold hard cash.
Which means that just like when I started off the Russian Project, I need a control week for the English Premier League and before I get started on Week 9 of the season, I have noticed a potentially fatal flaw.
Form guiding me down the wrong path
Now if you read the previous article, you will be aware that I have set up a developer notes page and this is will track all the major movements, in between the myriad of betting logic and thought pieces.
While in that section, you will see that r1.01 was all about tweaking the form index – which fundamentally is the bedrock of the model at the early stage and on a deeper dive, it looks as if I’ve tweaked it incorrectly.
What does this mean exactly?
Without getting too technical, the way it was ranking the six-match form combinations was slightly incorrect and not also taking into account the amount of points won – which subsequently has the potential to cause a lot of problem.
What’s the fix?
For the start of this sub-project, I have resorted the form index from 1 to 729 and including points won in this sort. Something that hopefully should give me more success in the weeks and months ahead – the best way to explain is by showing the predictions for this week
The reason I spotted an issue was the Leicester vs Burnley game – with the previous index it was predicting a Burnley win, which in itself is not the biggest scandal ever. However, when you consider that Leicester are at home and are higher up the table, then something cannot be right.
Where did it go wrong?
If we take Leicester and Burnley’s last match before the international break:
Leicester lost to Liverpool
Burnley beat Everton
Therefore my index was looking for a wins first regardless of points scored and ranking that higher than any club that lost their last game – even if they were better off in the table and had accumulated more points in the past six matches.
Ultimately this means, that there are going to be far more incorrect picks than correct ones and this had the potential to cause a lot of heartache along the way.
Which means, now I’ve ironed out that kink – I now have my control picks for the English Premier League and what I am going to do in this division, is be a lot more rapid with the tweaks and changes, if only because there’s no real financial pressure and any success can be fed back into ‘StatBot’
The only thing that concerns me is the fact that there is only one single draw out of ten matches (and it’s Crystal Palace vs Manchester City), but let’s see what it throws up come the end of Monday night, as we can get an idea of how successful this will be.
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 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.