A Stubborn Refusal
A Stubborn Refusal
If you read my previous article on trying to offer a data driven betting approach to the English Premier League you will be aware that it was once again time to serve up another raft of top-flight predictions.
While after getting five out of ten correct with my last effort, it was time to see if the incremental progress could be continued and more importantly whether I could take my predictions out of mid-table mediocrity. With that in mind, let’s take a recap at the picks below:
THE MOMENT OF TRUTH
On the basis of the model above, it seems as if Chelsea and more importantly Manchester City are causing me some rather big headaches and if it weren’t for their two relative slips, an average week would have become a very solid one.
However, there is no time for ifs and buts, just a review of what was actually correct and overall it was another 5 out of 10 correct – no change from last week and that means after six weeks of operation, I have returned 29 out of 60 picks = 48.3% success.
To be honest, you could probably return a similar strike rate on gut instinct alone and especially if you consider Liverpool and Leicester keep winning, therefore you’ve already got two picks in the bank before you’ve started.
So perhaps this is the time to be a little hyper-critical and compared to the success of the goal probability bets that are also being placed, this is still wide of the mark. However, that’s not to say it is a total bust either.
For single line betting, this is not the worst return. Although to be testing this against the big boys, it is still has an element of EFL League One to it and there is certainly scope for more fine tuning along the way.
LOOKING LINE BY LINE
One recurring issue is the fact that I’m trying to build a model that fits across all ten matches in a week. Although on reflection this must run in trouble because, the variables for Watford vs Crystal Palace must be different to that of Manchester City and Manchester United.
Therefore, the question that needs to be asked is: How do you model this more on a line by line basis? Because if you look at a major high street bookmaker, they will not be draping a model over all of the matches, they instead will be analysed all individually.
Which means, what other factors are they taking into an account? Because just form and league placing as a basis for a predictive model does feel a little rudimentary at the moment, even if it is part of the charm.
That said, there is no reason why I cannot keep adding extra data driven elements to this, for me it’s more a case of knowing what functions to apply and more importantly how much weight they should be given.
While one other thing to take into account is the fact that this model is not reactive, in the sense that it cannot deal with team news (at the moment). That said I do have a mechanism to factor in team changes (although that would mean updating everything 1 hour before kickoff – a problem if I’m off watching Tottenham)
Then again, I’m not going to be put off by this stubborn refusal to move up the predictive table and as there was nothing in the way of a downturn, I think I’m going to bury my head in the sand for another week, unless I can think of some inspiration in the next few days (ideas always welcome)
On the topic of ideas, a shout to Lewis Chamberlain who suggested linking the Over/Under and prediction models together – that way you could generate scorecast bets, that in itself should not be all that difficult and is a fantastic idea. The same could also be done for BTTS.
Which means there are a raft of models to be made and a whole lot of football to be played in the next few weeks. Let’s hope there are some winners for you this weekend and I’ll offer up some more data driven predictions in the next few days.
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 email@example.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.