Weight On My Shoulders
Weight On My Shoulders
Now if you read the previous article, you would be aware of the fact that I’m now going to try parallel models, in order to get a better idea of what is going to work and hopefully in a much quicker time (that and in all honesty, I’m enjoying working on this and therefore want to write more content)
In doing so, it will hopefully allow me the ability to scale up the amount of betting in the test phase and the more fine tuning that can be made, means that we get closer to ‘StatBot’ being rolled out across multiple leagues.
Not only that, but because it is the international break (please end) it has led me to do some more thinking and that is how much weight is going to be needed to be placed on a couple of different factors.
When working on the first couple of weeks of the Russian Premier League results, I’ve been using a form index and one that has been tweaked ever so slightly from week 1 to 2.
However, fundamentally the index is just comparing one set of form against another and does not take into account anything in regard to home advantage. Which begs the question, does an initial amount of weight then needed to be added to the teams playing at home?
Logically you would have to say it should, but the danger is whether or not you end up adding too much weight to the eight Russian clubs playing at home each week and whether or not, you are then getting too much in the way of home wins as predicted results.
Too little weight added and you won’t see any benefit
Too much weight added and you will see the prediction logic heavily skewed
Therefore the question is, just how much weight needs to be applied? That is something that I will answer while working on the parallel version of my models and by only working on a percentage hit rate basis, it means I can afford to be a little more experimental on competitions such as the English Premier League.
Spreading the load
When we consider that the first two weeks of my Russian Premier League betting series has returned profit in both weeks, it means I’m not going to make a wholesale changes in terms of modelling. But one could be made in regard to the amount of unit placed on each game.
With each game consisting of 1 unit (8 in total) it means that I am currently betting on 8 singles at all the same level – something that has the ability to hamper my ability to make profit on a weekly basis.
Therefore a smarter solution could be to place more units on a banker (odds of less than 2.0) and less units on games with longer odds – for example down to just half a unit on a game with odds of 3.25.
In doing so, if the bankers come in as a correct prediction, I then make more of a return
While if the longer shots come in as a correct prediction, I also them make more of a return
Of course this is not a fail safe plan and there is also the fact that we’d be getting away from the fact that all games have been treated equal in the first two weeks of the test. Something that has the potential to skew the experiment in terms of pure profit and loss.
Ultimately this all food for thought and something to think about for the days and weeks ahead. However, I will leave it there in this piece, if only because I can expand on the idea a bit further in my next blog article. So until then.
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.