Update on the Bookings Model
I wasn’t planning on posting an update this week – with the international break depriving us of any meaningful football, no predictions have been done for bookings. However, the last week or so has provided an opportunity to develop the code, and act upon some of the suggestions I have mentioned in previous write-ups. For one, I wanted to add in a few extra leagues which run through the summer – more on those later.
However, I had the idea to expand the code to account for additional card markets – handicaps. Lets starts there…
Handicaps Added
The model does suggest a fair card split, however I was having difficulties converting the split into fair prices. The break provided an ideal opportunity to sit down and properly look into it.
Eventually, I got it fixed. So now, the model spits out a predicted cards total and a split for each game. I can then use both to determine the odds for each team to receive the most bookings, and the tie. An example of the output is shown below…
The markets these will involve are Team 1 Most Cards, Team 1 (+1) Card Handicap etc.
Luckily, I have been tracking data for card splits right from the start of this project, so I will be able to see whether the model is accurate or not. If it seems to be pushing out similar results as the totals, I reckon it could be used relatively soon for actual bets. If not, it will be tweaked until it is.
For now, the plan is to have a look deeper into the data for card splits – determine how the model is performing and look into any potential fixes or improvements.
Once done effectively, betting volume is likely to increase. I’ll hopefully be able to do that this week, and add in an update for next weeks’ article.
Internationals
I tried to get some predictions out for the internationals recently, but ran into a multitude of problems. Without going into too much detail, the dataset for each team is incredibly small and not a very good quality. For example, for most European teams, they have played just 6 games in the Nations League so far, against just 3 teams of a somewhat equal standard. 3 home games, 3 away games – the data is just not diverse enough and I don’t really want to extend the sample back to previous seasons.
Also, bet365 are consistently poor with their offering of cards on the internationals – we only really get markets for England games and some of the marquee fixtures. Because of that, we can only really bet on about 10 games or so across the entire 2 weeks, which is no good really.
Finally, the various competitions further reduces the quality of the dataset to make predictions from. For example, a team might have played 6 Nations League games, then go into a 2026 World Cup Qualifier fixture, or a Nations League knockout – contextually, completely different fixtures. Again, it just makes it more difficult to justify the use of previous data this season.
With all that in mind, I basically have decided to just sack of expanding to internationals, for now at least.
Additional Leagues
However, there is more positive news in the Americas.
The Brazilian Serie A starts this weekend, which runs throughout summer. I’ve added those teams into the database, and will begin with predictions in the next few weeks.
MLS is also underway, 5 games in already, so that could start much sooner.
Importantly, both leagues run through the summer period, meaning there are domestic leagues to bet on throughout the calendar year. If we can replicate the accuracy for those leagues, that is absolutely ideal. But it also means the only time predictions will cease is during the international breaks, which occur four times, for 2 weeks. So 11/12 months will be packed full of cards predictions. Good news.
Domestic leagues resume soon (Barcelona kick us off with a random game on Thursday night), and it’s back to business as usual. Remember to join the Telegram Group – I’ll be posting all the predictions in there.
96 people in the group right now, and it will be capped soon! Once again, thanks for the support on this project.
Leave a Reply