1st September Bookings Model Update
August was the first losing month since we started tracking bets with the bookings model, the first in 2025. So in this article, I’m going to go into detail about what happened, what can be improved on, and some future plans.
Lifetime Results
Since tracking started in January:
- 46.84 units returned
- 9.92% ROI
Been a bit of a stationary period over the last few months, volume has been much lower and EU league are just starting up again, so that will begin to ramp up.
August Results
August was the first month of red since tracking began, not helped by a quiet first couple of weeks in August which didn’t perform to well over in the MLS & Brazilian Serie A.
Since the EU leagues resumed, there was a decent improvement, more green days than red, but ultimately not enough to get back into positive.
On the plus side, we were able to take were a larger volume, and as mentioned, a positive start to the European season.
So, instead of taking the easy option and just blaming negative variance, why did August finish as a loss, and where do we go from here to ensure that doesn’t happen again?
Sharper Lines
It is worth noting that Bet365’s accuracy has improved noticeably – it was hovering around 11% mark a couple months ago in terms of the total over-estimate on cards, which is now down to 8%.
However, the cards code has also become more accurate, responding to some of the changes I implemented shortly after the Club World Cup, which is a good sign.
With that in mind, the cards code is not to blame for the results, instead I have to look at my own process and what I might be doing incorrectly, or not as correct as I could be.
Updating the Selection Process
The first thing to do, is look into the numbers, and determine whether I am selecting the right games to stake.
Instead of using a flat difference, I’ve switched to a percentage difference, which is analogous to EV. I’m not actually sure why this idea never came to me before, but it just makes a lot more sense to use this.
From here, let’s refine a criteria to select games going forward. As always, the focus will be on unders with overs predictions proving to be a bit inconsistent.
Games with a 20%+ EV have shown to have the highest win-rate, which is ideal. Going forward, these will be our gold-highlighted games, and we will double down on these ones as the best value.
There is a slight drop in win-rate from 15-20%, which is not really what we would expect, but the sample size is relatively small so maybe there is some catching up to do in that range in terms of winning more often. Going forward, these games will be highlighted in dark grey.
I’m going to start including predictions from the 10-15% EV range, because we have a solid win-rate, and a good enough sample size to draw that conclusion with confidence. A 64% win-rate is ideal, that’s pretty 2 in every 3 bets winning in this range for EV, which is exactly what we want. This will mean an increase in volume, which might hopefully mitigate the effects of variance going forward a bit better.
Improving the Staking Plan
As mentioned, August was a relatively low-volume month given that the EU leagues started in the third week, so the first couple of weeks were quiet.
In low-volume samples, it’s not unusual to get this kind of negative variance, but I also pushed for a few more longshots this month. Previously, I had been relatively conservative, but I wanted to incorporate more longshots to raise the ceiling for potential winnings.
At the same time, putting more units down runs the risk of losing those units, so raising the ceiling also added risk, lowering the floor for the potential loss. In August, we didn’t hit as many of these longshots as required – I do want to give this strategy more time though. I would prefer to have a higher ceiling for potential returns, so I’m going to focus instead on refining the strategy as opposed to binning it after a negative month.
The ideas I have in mind are to ladder the gold games as singles, and double down on those. I will look to create trixies from the light grey games, and potentially some smaller ladders from the dark grey games.
Expanding Leagues
Bookings lines are available in the Argentinian top flight, and Liga MX. My aim is to get predictions available wherever bet365 are releasing lines. Again, a higher volume gives me a bit more freedom, and accounts for variance.
In addition, the poor run of form across the MLS and Brazilian Serie A prompted me to try and increase the volume in those league – I am trying to expand a bit more in to the North and South American leagues, so I’ll be keeping a close eye on the Argentinian and Mexican predictions.
Additionally, I’ll be trying to get some predictions out for the upcoming internationals.
As I said, where lines are available, I need to be scanning for value. It can be easy to ignore the international break, and use it as a time to do nothing, but you can find some great value during this period. I’ve not really given much thought to the cards markets in these games before, but that will change this season.
I will probably be posting predictions without any advised stake just while I take some time to figure them out.
Ultimately, a losing month is never ideal, but they can also be a good thing. Because of the results in August, I’ve realised some things can be improved, which I will work on over the next two weeks while we have the international break, and that is important. I also like to think I’ll be utilising this model for years to come, so one losing month early on in the grand scheme of things isn’t exactly the end of the world.
I’m really looking forward to continue working on improvements, and seeing how the model performs with a full dataset, which will be achieved in January 2026. Until then, my focus is on expanding out to as many leagues as possible, making the code run as efficiently in order to accommodate that expansion, and locking down a proper staking plan which brings the best out of the model.
As always, I appreciate your support on this project, and will continue grinding!
