English Premier League: 12/14-12/16

The first post of this blog will be something to take with a grain of salt, and a great time to look back on to reflect on how far I’ve come in the future. I spent most of this week figuring out how I was going to get a clean data set to work with. To save myself some time or money web scraping, I ended up getting some match CSV’s from footystats.org for $20 per month subscription. They have access to just about any soccer (football for the league in question) league’s stats you can think of with clean downloads available.

Enough with the unpaid promotion. I quickly got to work cleaning this data just a bit by separating the teams which were on the same row. This gave me two entries per game so both teams stats could modeled accordingly. Before you get upset about that being incorrect, I soon realised that maybe that was pointless?

Once I had a dataset I was happy with I got to work…. Well chatgpt got to work. I haven’t used it an excessive amount but I will probably be using it a bit more, as uploading the dataset and telling it what to do saves a tremendous amount of time. You have to have an understanding of the models inputs and outputs, as well as the conditions required to draw assumptions from that model or you’ll have lots of trouble. The code it gives you will definitely not work, but the framework it lays is incredibly helpful if you know what you’re reading (IF YOU DON’T UNDERSTAND THE THEORY OR CODING LANGUAGE YOU WILL PRODUCE A NOTHING BURGER.)

Boom. We got ourselves a model after some tweaks….. and then a few adjustments….. and then a few more tweaks. And I can confidently say without testing it that it is awful. BUT it is a model. Essentially it boils down to predicting future goals scored and processing the difference as the result. I decided to use .5 goals as the limit for a draw, as it would be silly to say that an assumed difference of .02 goals would be closer to a win/loss than a draw. As I said above this is awful and will be adjusted.

The current goal for the project is to create a model and then run simulations of each match individually to get sample distributions of each available line, to check for the best possible odds to trade. This will allow us to choose from over and unders and spreads as well as our basic match result.

Now for the only part that anybody cares about. PREDICTIONS. Lets keep it short and sweet.

Dec. 14th 2024 - Newcastle United vs. Leicester City - Draw or Leicester City (-136) : Our model predicted a draw, with Leicester City having a .2 expected goal advantage in the matchup. Leicester City was a heavy underdog in the matchup on the online books. The odds for the draw were not ideal so we decided to include the Leicester win.

Dec. 15th 2024 - Southampton FC vs. Tottenham Hotspur - Tottenham Hotspur (-144) : Our model predicted this to be a 1.33 goal advatange in the favor of the Hotspurs, which was on the higher end of our predictions. The odds were a steal in this matchup, as significantly higher favorites were projected lesser advantages in our models.

Dec. 16th 2024 - AFC Bournemouth vs. West Ham United - Draw (+295) : Our model predicted a draw with a .2 goal advantage to the favorites. This was our only game on Monday and with the model advtanage going towards the favorite, unlike our Leicester game, I felt that it was a better option to take the higher odds and bet on the draw.

I’ll upload a link to the model in the future so you can see the cleaning and modeling process, but for now here are the very ugly results.

https://docs.google.com/spreadsheets/d/1Yv00-YdzpNFGvOnkKmJb-WIR8qaB1VQnq-JLVlnGENE/edit?usp=sharing

Make your own picks, and check out the deposit bonuses available on MyBookie at the link below!

https://engine.mybookie.ag/refer_friend?first_page=1