8th – 14th April Bookings Model Update

Another week of business-as-usual tracking under the belt of the Bookings Model – we are approaching 1000 individual games tracked this season which would be a real milestone. This write-up will go into the weekend & lifetime results (as usual), with a brief mention on how the MLS & Brazilian Serie A are getting along, and also some additional challenges that are beginning to pop up as we approach the end of the season.

Anyways, lets start with the…

Weekend Results

114 games tracked this week from Tuesday 8th until Monday 14th!

View Results as a PDF here – Bookings Model Tracking – Week Recap (2)

This week the graph looks a tad bit different – an extra trio of columns have been added in order to evaluate the success of handicap predictions – more on those later.

Overall, main lines returned a profit of 0.5u, so pretty much a break even week. It was another poor weekend, which took away from some really nice midweek results. Nothing out of the ordinary though, some weeks that will happen, and there are always lessons to be learned and improvements to be made, which is the point of these write-ups. As it stands, April is +12.32 units on main lines.

Lifetime Results

Tracking Handicaps

First of all, I’ll explain how this works in regards to the tracking. When the cards code predicts the home side to account for more booking points that the market is offering, games are highlighted in red.

Alternatively, when the cards code suggests the away side will account for more bookings, these games will be highlighted in green. So for these green games, the model is suggesting to take the away side on the card handicap. The exact value is 7% – when there is a difference of 7 or more between the model and market, games will be highlighted either in red or green.

Now, looking into the lifetime results, the model has been incredibly effective at predicting the total split count – to a 97.9% accuracy. The first new column, Most Cards, is used to determine whether the model correctly determines which side will be awarded the most cards in a game. So if the market is at a 42 split, and the model predicts a 40 split, it suggests there is value on backing the away handicap. If the away side then receives the most cards, this will register as a win, and if the end result is a tie, or the home side gets the most cards, it will track as a loss. Simple enough.

Too simple, though, because I don’t bet on team to get most cards markets – they’re too short. I’m betting on the Asian Handicap markets, so need a way to track how the code is performing in regards to these ones. In order to do that, I had to relate the split to a handicap. For example, in the case of a 35 split, the home side is given a +1 card handicap, and visa versa. This is what the Cards Hcap Total is showing.

The third new column, Hcap Res (stands for Handicap Result), takes into account this handicap value before determining if the bet will win or not. So the Hcap Res column is important. Now, lot’s of words there, apologies, but let’s shift over to looking at the numbers.

The Model predicts which team will get the most cards approximately 41% of the time from a 829-game sample.

Because these markets are 3-way (as opposed to 2-way for totals), that number probably sounds worse than it is. However, a 40% win-rate is about what we would expect to break even. Obviously, this is not ideal, so there needs to be a way to filter or screen games in order to select the best value – just as we do with totals. Here comes the fun part…

Green bets (where the away handicap is value) are slightly less common, and have a win rate of 41.9% – not too far off the general total. However, red bets are more promising. They are more common, and have a 50% win-rate, which is great for a 3-way market. Intuitively, this makes sense. The market overall is more likely to back away sides for cards, and home teams tend to be favourites as well, which can shift the balance.

Sample size are still slightly too low to draw any concrete conclusions, but what lessons can we take from the above results?

Green Bets for a difference of 10 or greater
Red Bets for a difference of 7 or greater

This will be the criteria going forward!

MLS & Brazilian Serie A Update

We are 8 gameweeks into the MLS, and predictions are still a bit wayward. Still, a larger sample size is required before they are shared.

Brazil Serie A looks better though – I’m going more off how it compares to the market / how predictions look as opposed to results. Predictions involving the newly-promoted sides are still a bit off, but that is expected given that it is based off Serie B data.

Also, overs seem to be a lot more common in Brazil, with card makeups very high compared to the European leagues. More data is required, and I’ll check back in with these leagues over the next few weeks.

Challenges

European games have proven to be a challenge – especially the second-legs. It has been difficult to account for first-leg results in regards to gamestate, which has led to the handicaps being off. Also, I’ve found lines to be bumped up by about 25% for the second-legs when ties are close. Similarly, when ties are a blowout, lines are deflated by quite a bit more. Because of that, totals have also been a bit off – particularly for Europa and Conference League games.

I do think there needs to be more work on these European matches, additional factors need to be considered, which is something to look into for the upcoming rounds.

Domestic predictions have remained solid, although we are at the point in the season when tensions and emotions are beginning to be higher – games are important now in regards to relegation scraps, title deciders and battles for European spots. Some lines have been inflated because of these, although I do have a way for account for those.

Similarly, some teams have wrapped up the league title, or are very close to doing so. PSG & Liverpool are two examples. Going forward, it will be interesting to see how their card makeups change, although they would be expected to drop given that fixtures are deadwood.

Still work to be done here, and improvements to be made. As always, I appreciate the support on this project!