Out Of Control
Out Of Control
In last week’s introductory article I outlined the start of this project and how it has evolved from the design of league form infographs. An evolution which will see me build, test and tinker with a model, that will not only predict football results but hopefully predict the right ones.
Before we launch head first into this project, we need to lay some important structure to it and in doing so, it gives us a reference point to work from – one that will be the basis of all betting activity going forward.
What does the structure entail exactly? First it is a set of control bets that I placed last week and the theory here, is to see how successful ‘gut feel and instinct’ alone was last weekend.
As I mentioned in the introductory piece, the Russian Premier League is going to be the pet project to start with and last week I placed a series of shall we say ‘blind bets’, in order to give us the required reference point.
Here are the fixtures from the previous weekend and my selections:
FC Ufa vs Spartak Moscow – Away Win
FK Akhmat vs Krylya Sovetov Samara – Draw
Zenit vs Rubin Kazan – Home Win
FC Tambov vs FC Rostov – Away Win
FC Orenburg vs Lokomotiv Moscow – Draw
Dinamo Moscow vs PFC Sochi – Draw
Arsenal Tula vs Ural – Away Win
CSKA Moscow vs FC Krasnador – Home Win
(Correct picks in italics)
Each bet comprised of 1 unit and for the purpose of this experiment each unit will be 50p, for the simple reason that I am not prepared to bankrupt myself for the pursuit of excellence.
Admittedly eight single bets at 1 unit is not going to make me rich at this stage BUT an eight fold accumulator of the picks coming in, would certainly swell the coffers and this is where the model could EXCEL itself (pun intended)
With eight units place last weekend on these control bets (a total of £4), I returned £3.26 and although it is minus 18.5% ROI, it does at least mean that I’m not in the dark regarding Russian football and can identify a result or two.
While had I got one more pick correct, I would have turned a very small profit. However, blind betting alone is not the concept of the project and now it is time to get down to the nuts and bolts of it all, while explaining my first level of theory.
HOW IT WORKS
If we look at the example of last weekend’s fixtures, you will see that each team has a six-match run of form and the points that they have accumulated during that spell – something that will be the initial bedrock for my model.
What I identified is that the combination of how points are accrued could differ but the end tally could be the same, for example:
CSKA Moscow could have picked up nine points and won their last three and lost the three before that. While FC Krasnador, could have been unbeaten in their last five (two draws, three wins) but last their sixth game and still accrued nine points.
Therefore points alone as a comparative measure, is too weak a basis to try and predict who will come out on top in any given fixture, what we need to do is work out which six game result combination is better than the other.
Something I have managed to do by working out each of the 729 different six match outcome combinations and then ranking them. For example six wins would have a rank of 1 and six losses would have a rank of 729, all the other combinations would then lie somewhere between.
From here each team would be given a form ranking depending on how they have performed during the last six matches (this is just the control example – remember I bet on instinct last weekend) and then these two rankings would be compared.
The next logical step is to write some code that says if the ranking difference between the home and away team is more than a certain amount it’s a home win, less than a certain amount and it’s a an away win, while a certain threshold would mean a draw.
Once that code is in place, it will give me a first set of predictions to use for this weekend and that is what I’ll be using to test in my first round of model bets, where I will place the eight individual bets and a accumulator combining all the picks.
So I need to go off and put the final touches to the code, I might drop in a very quick article with the picks tomorrow (Friday) but if time gets the better of me, I will review the first week’s live testing early next week.
Thanks for reading and wish me luck. 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.