7th December Bookings Model Update
Another month of tracking under the belt, and it’s a bit more of a positive review than we got last month. There haven’t been too many changes done to the model, so in this article I will recap results, and put a timeline on when I could start sharing picks again. Anyway, lets start off with the…
Results
Much, much better. The changes have brought about an improvement, which I am very grateful for. Let’s quickly recap what those changes were.
- After running analysis on the best-performing leagues, I decided to be a lot more cautious with the South American leagues, and prioritise the Premier League, La Liga and Serie A.
- Been less frivolous with longshots – a lot more selective and wary with how much stake is going into any individual game. I was more cautious on the 1u, 0.75u and 0.5u bets, but they were all profitable which is a great sign.
- I doubled down and persisted with a high volume on the 0.25u bets though, although that didn’t quite pay dividends this time around. The average odds for theses bets are probably 20/1, so you would expect 1 win every 20 bets. We have been massively exceeding that percentage this season overall, but won just 1 of 27 in November, so didn’t quite get the rub of the green on those bets. Still, a -1.4u return is nothing to panic about by any means.
- Finally, I included a bit more subjectivity with the predictions. If a game was heavily weighed as value due to a referee, but the data and supremacy points towards overs, be cautions. Similarly, if everything aligned, high supremacy, good ref, good teams, only then was I confident enough to full ladder. Naturally, this massively reduced volume, but I felt a lot more confident in these bets.
Massive spike, massive drop, but we’re closing back in on the +100u mark for the second time – hopefully we can stay above it this time! As I said last month, predictions weren’t the issue. Accuracy has been getting closer and closer to the 100% mark for both totals and handicaps, the problem was how I was selecting bets, and I really wanted to improve my own process. I’m pleased with how things have gone since the changes.
The main thing to address is, when will predictions start up again – I’ll talk about that in a sec.
South American Leagues ending
We are coming towards the end of the Argentinian, Mexican and Brazilian leagues now, which naturally will reduce volume.
While I had mostly phased these leagues out in November, I did find the odd overs bet, or handicap bet, which I would say worked much better than backing unders.
It’s a bit sad to see the Brazilian league end actually – I remember doing the predictions right from the first gameweek this season. I had two sets of predictions, one using the current season data, one using the previous season data, and was testing to find out when it would be best to only use current season data. Brazil has been a great league for learning. The Argentinian and Mexican league I started tracking much later, so I’m not too fussed about them.
I don’t think they will affect volume too much – the focus has been on the European leagues anyway. However, it will probably mean a reduction in the overs bets – I found these to be a lot more lucrative for the South American stuff.
Timeline on Posting
Now, I like the new strategy. But I also liked the September strategy, and look how that turned out in October. However, November had the biggest win-rate of the season, which I was trying to do.
I would like to start sharing bets again very soon, but there is one thing holding me back at the moment.
Before giving a definitive time, I need to give a quick overview on how the code works – I scrape card data from fbref, then I scrape referee data, and then utilise data on supremacy which I have collected to create my predictions.
Scraping data from third party sources is allowed within reason, but I’ve run into problems over the last month or two with fbref changing their cloudfare integration, which has broken my scraper. As a result, I’ve been manually inputting the data which is time-consuming, and not ideal. An average Saturday used to take 20 minutes to run through all the predictions, now it’s taking hours.
While I have probably benefitted from adding a bit of doubt and subjectivity into the bets, I don’t want to be spending so much time doing this every day. However, there is a solution.
My plan is to use my current dataset (of over 3500 games) to act as the data source going forward. So essentially, I am completely self-reliant – no need to scrape any third-party website which is unreliable. My own data recordings will become the basis of my future bets.
Nothing will change in terms of the methodology or numbers, but it will require a hefty time investment to update all the coding scripts and get everything carried over and backed up. At the moment, I’m just about able to manage all the tasks I get every week, so I’m not quite getting the free time to work on the project as much as I would like, which puts me in a slight predicament.
Improving the code will save me a lot of time going forward, and remove the stress of relying on a third-party. But at the moment I don’t have the time to make the changes. But if I made the changes, I would have time.
Regardless, I’m going to keep chipping away at it whenever I can. When I can get this implemented, I can start to think about sharing predictions again, as I was doing before. Giving a timeline on exactly when that might be is difficult, but lets set the goal of getting thing back up and running before the new year.
As always, if you’ve got this far, I massively appreciate your support. It’s been a relief to bounce back after that horrible month of October, but I’m glad it happened in a way, because it brought me back down to Earth, and I’ve used that to improve my strategy.
There are still a few key changes to implement into this project, and I’m going to keep working away slowly – this season has already been a massive grind, but I know it will pay off.
Thanks again for your support, I’ll catch you all again next month in another recap!
