Hello!
A quick review of July.
I was really curious to see how this time of the year would go, with a lower betting volume and matches where motivation is often questionable (friendlies, preseason games, etc.).
I chose not to exclude some competitions (including friendlies) because I wanted the algorithm to gain experience – even if that meant taking losses this summer. At least, I’ll learn which situations to avoid in the future.
In the end, for 9 out of 14 strategies, July turned out to be the month with the highest number of bets placed.
Very mixed results depending on the strategy
The overall P/L performance varies a lot from one strategy to another.
- Strategy 1 crashed badly, but that’s not a surprise: it’s the least restrictive.
I’m still curious to see how it performs over a full year. In a year’s time I’ll decide whether it’s worth keeping track or dropping. - The new A and B strategies introduced last month performed much better.
Strategy 1a limited the damage, and the others had very good results!
As for my usual service, only 2 bets were shared: 1 win and 1 void.
I don’t really have the focus for that right now, so I prefer to put it completely on hold for the time being and focus 100% on tracking and developing my new strategies.
What has been done on the algorithm side
I managed in July to accomplish almost everything I had planned, and even more:
- Fixed several small bugs here and there
- Added new parameters to track in the match history for every analyzed game
- This should allow me to fine-tune and optimize my strategies even further over time
I also introduced the calculation of a Model Score 2 and therefore a new Strategy 2, which basically shows whether a bet would still be interesting with a slightly less aggressive handicap.
Automated reporting I now receive every day
I really enjoy the daily P/L recap I get every morning:
Throughout the day, I also receive the list of bets for today and tomorrow, displayed like this:
This gives me a quick overview of which strategies are currently performing best based on the number of bets placed.
I initially thought about adding columns for each month of the year, but it would just become unreadable.
Next step: Automating bet placement on Pinnacle
Given the number of bets and the P/L of each strategy, which one would you go for?
Personally, I’d be tempted to go with Strategy 2b, which currently combines the best profit with a steadily growing volume of bets.
Maybe Strategy 5 as well, which is slightly different from 2b but has a very high success rate with a growing number of bets too.
(To simplify: Strategy 5 focuses on model scores ≤ -2, while 2b targets model scores ≤ 0.)
I need to be very careful though: my first attempt at automatic betting (done way too quickly with ChatGPT) got one of my PS3838 broker accounts banned!
So for now, this won’t be my main objective for August. I’m aiming to have it ready by the end of the year, and if it comes sooner, even better 🙂
Mastering the PS3838 API will also let me fetch live match lists and odds in real time, which is not possible today.
Recent progress with the Pinnacle API
Lately, I’ve made good progress:
- Added safeguards to avoid getting banned from the API
- Managed to fetch Pinnacle match lists and link them to football stats data
- I can now run my usual analysis with real-time Pinnacle odds and spreads (instead of data updated only every 3 hours)
The next step is straightforward: trigger automated bets whenever a match meets specific strategy criteria.
Upcoming projects
- Restarting an analysis I had left aside for a long time: favorite teams tracking
- Studying how results would change if I used one step lower handicap lines
(e.g. instead of +1.25 or +1.5 or +1.75, simulate what would happen with +1)
This could mean higher odds and potentially bigger profits.
That’s it for July!
I’m super satisfied: so many positive changes in just a few months, and I can’t wait to see how each strategy evolves over time.