11th June Bookings Model Update
It has been a fairly long hiatus on the bookings model updates, but predictions have been ticking along. With the EU domestic season now fully concluded, time has been freed up to have a proper delve into the analytics, and try and work on a few improvements.
Before going into that, and the plans for the next few weeks, let’s first of all see how the model has done up until this point…
Results To-Date
After making some changes towards the end of April, May was our most successful month so far in terms of return, with a ROI just shy of 16% as well – it was a very positive month overall, especially with the end of the season being quite difficult to navigate in general.
Volume naturally was a bit lower as the fixtures dwindled, but that is nothing too unexpected.
Lifetime results since Tracking begun in February:
- 42.93 Units returned
- 13.2% ROI
Updates to Selection
While the results have been positive up until now, there are always ways to improve. I had a look into the analytics to see if anything needed to change. The model is still quite inconsistent with predicting overs, but I did notice something interesting with the unders…
The table above shows the win-rate for each range – the left column is the difference between the model prediction and the market line. As you can see, the win-rate for unders in general is high.
However, we need to be more selective to improve the win rate. Currently, I’ve been highlighting games with a 6+ difference in grey, and a 10+ difference in gold. But I’ve decided to change that.
The sweet spot is in the range of 8 to 14, with the results being:
- 195 total games
- 133 winners
- 68% win-rate for unders in this range. Off a very good sample size.
These are the new gold games.
Above the 14 range, we have a much smaller sample size, with the results being:
- 44 total games
- 26 winners
- 59% win-rate. Although this is decent, it’s not as high as we really want – it’s not high enough to justify itself as a gold game (highest-value). I think a bit more data is needed before doing that, so I’ve decided to change these to be grey highlighted. Which is slightly counter-intuitive – bigger difference = less value somehow, but there is probably a factor that I’m not accounting for.
Switching over now to below the 8, we have a much larger sample of data, so can hopefully draw more accurate conclusions. The 62% win rate in the 4-5.99 range appeals to me, and I think that justifies being a grey game especially based off the 237-game sample size. Looking into the 4-7.99 range, we have:
- 389 games
- 239 winners
- 61.4% win-rate. Again, very solid.
So the overall range for greys has been increased, which should mean a higher percentage of games are highlighted as value with the aim of increasing volume while also improving win-rate.
Club World Cup Plans
With the EU seasons ending, MLS and Brazilian Serie A predictions are in full effect, although volume is a lot less than usual. Hopefully the changes mentioned above will help get volume back up though.
Those leagues do have a break coming up as well, to make way for the Club World Cup. I mentioned this in my previous post a while back, and since then I’ve managed to get a model setup, tested and am happy with the progress and predictions from it.
It’s difficult to know whether teams take it seriously – I think it is likely that sides will play their strongest eleven, but with it being a post-season tournament for the European sides, I would imagine intensity isn’t quite as high, and we see some experimentation and rotation.
It is worth waiting until after the first gameweek to identify which teams will properly go for the title, and then re-assess from there. As always though, I’ll continue to share predictions. I have also been active in the analysis of outright markets, and have taken some nice angles for the tournament. I’ve been wanting to share them, and will probably do so on this website as a separate article, or on Twitter, so stay tuned there.
As always, I appreciate the support on this project and will continue it’s development and improvement!
