29th April – 5th May Bookings Model Update

May has started – the final month of the European football calendar. The aim this period will be to finish the season strong, to continue the good form and end on a high, but also prepare for the summer so that predictions and bets do not come to a complete standstill. Anyway- more on that later, let’s start off with the…

Weekly Results

Seemingly back on track this week, with some really positive outcomes – exactly the start to May which was required.

Lifetime Results

A Slightly Altered Strategy

I wanted to make a slight adjustment to the betting strategy this week – April ended on a bit of a sour note, and a lot of the hard work was undone. Despite that, we still ended in the green with a solid profit in excess of 10 units. 90 main lines, 7% ROI.

Going forward, I’m going to be slightly more selective with the main lines, so too with trixies. The aim is to improve ROI. Ideally, 75 bets per month would be optimal, but that becomes a bit challenging now as we approach the end of the season.

This week, the new approach paid off nicely – we got some solid returns on main lines, which was compounded by two winning trixies.

I will stick with this strategy until the end of the EU season, trying to be more selective. I think it is a solid strategy, but I will continue to periodically look into the numbers to see if any changes need to be made. It’s important to remain adaptable, and open to change and improvement at all times.

Addressing Issues

As I’ve mentioned before, doing the painstaking task of manually inputting data and results for each game can have the benefit of helping me pick out issues. One thing I noticed this week, is the code would struggle with pricing handicaps when one team has a lot to play for (whether that be a relegation battle, or fight for the title or European spots), and the other side has little to play for. In these games, the market is expecting the side with a lot to play for to account for more bookings, which makes sense. So that is another thing that has now been fixed and priced into the model.

Even though these games are quite rare, and only occur at the end of the season, it is nevertheless an important issue to address.

Brazil Serie A – This Season or Last Season?

Round 1-6 saw predictions based off data from the previous season. I found totals to be quite accurate on the whole, handicaps less so. Especially for the newly-promoted sides.

We will likely have some of these issues translate over to when the European leagues start next season, so the Brazil Serie A is important in a sense to gauge a method to deal with these outliers. I think it makes sense to weight supremacy quite high for splits, at least for the start of the campaign when predictions are based off the previous.

Round 7 predictions were a bit more wayward for the totals. Because of this, I thought it might also be helpful to do another set of predictions based off this seasons’ data, and see how they compare. Still, last season’s dataset was a lot more accurate. We’re just at a bit of an awkward part of the season, whereby last season’s data is become a bit redundant, but this season’s dataset is too small to accurately base predictions off.

I’m going to keep using both datasets for the Brazilian leagues, and try and find a turning point when we can safely use this seasons’ data.

The news is slightly better in North America…

Is the MLS Ready?

I was really pleased with the predictions this week. They looked much better, in line with bet365 on the whole but still highlighting value.

We have completed 11 gameweeks in the MLS, so there is enough of a sample size to make predictions with confidence. Going forward, I’ll be including MLS predictions for splits & totals, and begin tracking main lines.

Circling back to the Brazilian league, if round 12 is the sweet spot for using current season data, then we have a 6-week wait before predictions will be posted. It will likely be straight after the Club World Cup. I do actually have a plan for approaching that tournament, and will likely mention that in more detail soon.

Anyway, it seems like….

Round 1-6 uses last season’s data with a heavier weighting on standard supremacy values.
Round 7 – 11 is an unknown, and requires a bit more thought as to an optimal approach.
Round 12 onwards uses current season data.

I’m going to do a slightly different weekend recap going forward, focused on the unit return over the week as opposed to model performance. The aim is to be a bit more efficient with my time.

The main question to answer is around how to approach the start of the season in terms of weighting different variables, which is going to be quite a difficult task. Other than that though, it’s just really a case of tracking the games each week, collecting as much data as possible and continuously making small tweaks and improvements. These writeups will therefore only become more tedious and repetitive, so we could be coming towards the end of them.

As always, I appreciate your support on this project!