A New Bookings Model
Moving into February, I’ve been working on a new model to predict the total number of booking points across various leagues. It’s a project I have dedicated a lot of time towards, and am excited to share.
To hold myself accountable and remain consistent, I’ll be logging the results weekly in articles posted on this website, along with a bit of qualitative analysis / my thoughts and opinions on the results and potential improvements.
The model itself is incredibly efficient, able to generate predictions based on just a few inputs taking no longer than 30 seconds. If results can match it’s efficiency, it has an incredibly high potential.
The final goal of the model is still unclear, I might release it on the website for people to use whenever they please (although there are issues with this method), or potentially share the picks to a specific group. That’s still to be decided.
For now though, I’ll be posting the results of the model as I continue to test it. Here is what we got from Day 1:
Friday 24th Results
Apologies for the absolutely tiny image – I’m not too sure how to fix that for now but I’ll find a workaround going forward.
To summarise though, there were 6 games on Friday across various European leagues, and the model generally was in agreement with the market line (which mostly refers to bet365 pricing). For example, it was just 0.3 booking points off in Las Palmas vs Osasuna and 1.5 in Samsunspor vs Gaziantep. The model suggested to play the under in these games, which would have proven to be a losing selection – fairs.
However, in the other four games, the model suggestion would have generated a winner. That is particularly appealing for Torino vs Cagliari and Wolfsburg vs Holstein Kiel.
Starting with the former, the model predicted 46.5 booking points, with the market line set at 56, which is a massive difference – almost an entire line in favour of backing the under. In the end, just 10 booking points were awarded, which is a significant win for the model.
Next, the model predicted 56.3 booking points in the Wolfsburg game, compared to a 51 market line, suggesting to take the over. In the end, 60 booking points were shown for a narrow model win. Again, the model prevails when the difference is large, which is important.
If the difference between the model prediction and market line is small, there isn’t much value to be had, meaning it’s probably best to ignore these games, or play with a lower total stake. However, where the difference is largest, it’s worth playing an increased stake, and potentially considering a ladder.
6 games is an extremely sample size, but it’s a seriously promising start. As mentioned, the code is fantastically efficient, meaning if these kind of results can persist over a massive sample size, we could be on to a winner here. Lots to look forward to. I’ll check back in on Monday or Tuesday, after the domestic gameweek finishes in the top European leagues!
