6th October Bookings Model Update

After suffering a losing month in August, the last five weeks have been a serious bounce-back, and we find ourselves +34u in the green. It definitely hasn’t been easy, and it has been a tale of two halfs, so there is plenty to address in this article – it’s a long one, but I hope a lot of it is useful for those that have been here for a while, or those who plan to be. As always, lets start off with the…

Results

Whilst October is obviously an incomplete month and we’re less than a week in, those results are definitely worth addressing, in detail. Lets break it down week-by-week.

Top red box, internationals. Send is just the first gameweek back, then second gameweek back, and then the last couple weeks are in the final box.

Starting off with the internationals, which was the first time I had posted predictions for these games. In the end, I was kicking myself for not staking more, because we posted some very decent profit in excess of 10u. I found the high-supremacy games here to be incredibly lucrative for unders, which we exploited a few times, even despite the cards offering being quite poor. Going into the October internationals, I will likely deploy a similar strategy, maybe a slight bit more aggressive.

Second box – the first gameweek back, and the first gameweek with the new strategy. Overall, the results speak for themselves this week – it simply couldn’t have gone much better. Landing a full ladder, as well as a couple double.

Third week was pretty much the same, we landed a full ladder on Friday and Saturday. In both of these gameweeks, there was exactly 6 games with 0 cards shown in the leagues we cover – the model picked out half of them.

Since then we’ve seen a pretty nasty swing, around 7u down the first weekend, followed by a poor midweek, and then another losing weekend, pretty heavy this time.

While I can understand some of the backlash in the group chat for those that are new, I would like to remind you that first of all, I am trying my best, and cannot physically stop these cards from happening. We have had a lot of pretty bad beats with late, unnecessary cards, especially on Saturday, but these things happen.

However, I back the bets myself, and losing runs affect me 10x more when I know people are dissatisfied. And because of that, I’ve got my head down and have been working on an updated strategy going forward. We’re going to persist with what has worked well, and disregard what hasn’t up until now, while also testing a couple of new things. More on that later.

I do think it’s important to not over-react to losses – the updated strategy after the international break had worked so well, would it have been fair to change it after one negative week? In hindsight, the answer was an obvious Yes, but I wasn’t to know at that time.

Those that have been in the group for some time know that we have had these losing periods before, and we always use the lessons from them to make improvements, because that’s what this project is all about. It’s about getting closer and closer to the most accurate prices, and beating the market on a consistent basis. Getting to that point wont come overnight, it’s a massive work in progress. Again, more on that later.

Up and down, but since our last article update, we have banked a total of 34 units. Here’s how I plan to increase that number over the next 5 weeks.

Working on an Overs Strategy

With the Argentinian games, referees are quite variable – you’ve got a bunch with quite a low average, and another bunch with a really high average, which led to quite the variation in predictions. However, I noticed that the model was generally quite good at predicting overs – or maybe that is just what I wanted to notice. Because of that, I’ve had a proper look through the numbers, and picked out a set of criteria.

In the 15-20% EV range, overs win-rate is 64% from a sample of 100 games.
Liga MX, MLS, Liga Profesional & Brazil Serie A have been the best performing leagues (in that order). From Europe, Scottish Premiership and Ligue 1 look decent.
I’ve also made a list of referees to keep an eye on across every single league.

I’ve mentioned a lot of times before how overs bets are not as lucrative as unders generally, but there are some exceptions. Hopefully that set of criteria above will help us spot those exceptions, and we can take advantage of it. As always with a new strategy, it’s probably best to initially be a bit cautious, but that will be tested over the international break.

Adapting the Supremacy Weighting

Over the last month or so, I’ve been backing unders for a lot of games where there is a heavy favourite. Markets adjust for these matches with a low cards line, but it’s just not as much as they should be.

However, one thing I’ve noticed from these games, is that the favourites don’t always just finish on 0 cards, even if the scoreline does get a bit out of control.

The reason for this is, the underdogs sit deep, and can’t afford to commit too many fouls when they are penned into their own box. However, when they counter and draw fouls, the favourites tend to commit cynical fouls, with a much higher likelihood of receiving a booking.

The code currently adjusts the split quite heavily towards the supremacy, so if the home team is heavy favourite, the model will output the split to reflect the supremacy much more than it does recent data. 365 do something similar, which explains why handicap bets are much less common from the cards code.

I had a look into the numbers to see if I could confirm that theory, and did actually find something interesting. I sorted the dataset by the lowest supremacy games, so fixtures where the home side are heavy favourite.

There were 22 fixtures when the supremacy was -2 or greater, and the massive underdog actually only received the most cards in 8 of these games. That seems to actually confirm the above theory, suggesting there is big value to back the favourite on the handicaps, or to just outright get the most cards. In that sample, there were 8 Ties, and the underdog only got the most cards in 6 games!

For a positive handicap – when the away side is the heavy favourite, there were only 13 games in the sample with a +2 handicap or greater, and the results were less promising. The underdog got the most cards in 2, 4 games were tied, and 7 games had the underdog get the most cards. The away side actually got exactly 0 cards in 8 of those games, with some incredibly low card makeups.

Based on that, target handicaps when the hosts are heavy favourites, and unders when the away side are.

Going forward I definitely want to trial out using the data a bit more, and test whether there is in fact an inefficiency there to exploit. This would involve tinkering with the actual code, so the output for split will change (totals will remain the same), so it may require a bit of time to get that sorted. Again, I do want to add a few more handicap bets into the mix for a bit of variation, but that will only happen if the value is there to exploit.

Refining the Unders Strategy

I’ve mentioned in those previous two sections ways in which I would like to increase volume, so I’m going to try and refine the unders strategy hopefully in an attempt to increase that win-rate.

In addition to tracking overall units, return, ROI, I also set up an additional system to track longshots, so I know what is performing best.

On the whole, and last month in particular, trixies were a real let-down. Considering these are composed of games between 10-15% EV (generally), it suggests we got the least success with those, so should be dropped.

I’ve again looked into the data and come up with an updated criteria:

Target the EV range 17.5% and above.
Target leagues: Serie A, Premier League, La Liga, Cup Competitions (in that order) primarily.
Be wary with Ligue 1, Bundesliga and Liga Profesional.
Again, I’ve also made a list of referees to target.

This will reduce the volume on unders, and hopefully work to improve the win-rate. We’ve stuck with the selections that have brought us the most success over the last 5 weeks, and binned off the parts which have let us down.

I don’t want to compromise on the overall volume too much, it might reduce slightly, but I want it to be higher-quality, hopefully this new strategy can bring that.

Answering Your Questions

The Cards Code group has become quite a big bigger recently, which has brought ups and downs – there is quite a lot I want to address here. First of all, prices have generally held well after posting, but I have noticed a bit more movement than usual, especially in the last week or so. That could just be an anomaly, but I am going to push the group a bit less in the future.

I’ve had a lot more messages recently with people asking questions, which has been useful for letting me know where I need to improve, or how my system could be refined. For example, I’ve had several people ask me for a P/L. All bets are tracked on the spreadsheet I share, I just crop out the tracking bit. But it’s all there. At the end of each month, or around quieter periods like international breaks, I collate all the data for the month, and then share it in one of these articles, as I have done so above.

So, going forward, I’m going to automate the tracking process a bit more, so that I will always have an updated P/L at hand, so when people ask for it, I can share it immediately, as opposed to saying ‘wait until the end of the month’ – that’s obviously not the right thing to say.

Another thing, I’ve had people ask whether it is fine to only take bets on 365. I am going to be posting PaddyPower bets a bit more, just because straight reds count as one card so we do get more value there compared to 365. I get that might not be good news for those that tail from aborad, who only have access to 365, but by showing the fair odds with all my posts, it will give people more of an idea of whether the bet is good value or not. Additionally, for PaddyPower bets, I will explicitly mention whether it is good to take with 365 as well, which I hope will help those people who only have access.

I’ve touched on the backlash briefly, but would just like to say that I’m going to continue working hard on this project. I’ve only been running the model since January, not even a full season, so there are plenty more lessons to be learned, and plenty more of these bad runs to experience. On my tracking page, I’ve set up a table for the 26/27 season, and the 27/28 season. It’s a long-term project, and we’re only getting started.

I did also get a couple of messages on that Saturday which gave me a lot of confidence, hence why this article is so long, with so many potential improvements listed. We had a losing month in August, and picked ourselves back up for a 50 unit month in profit for September. We’ve started October in a rut, but best believe it’s not going to stay that way.

I do also appreciate the people that suggest improvements – the message about the referee change was an incredibly useful one, as was the messages picking up on the fact I accidentally used the fair odds for a U3.5 line instead of U2.5. Things like that, very useful, so thanks.

As I said before, I’m gonna keep pushing out predictions (even if I do get the poo emojis on every post aha), keep improving the model, and keep trusting the numbers.

The increased volume has made posting stressful at times – increasing the stake we’re putting down puts us more at risk of some heavy losing days. In September for example we had a couple terrible days of 6u losses, and the 10u loss on Saturday, which is never going to be easy to take. But we also had winning days of +19u, +22u. I think it’s important not to get too flustered by bad days, they will happen, but also not to get too excited and over-confident after winning days.

Overall though I can’t complain, this project has been really enjoyable throughout, the constant changes and upgrades, random things breaking and requiring a fix, and never-ending list of ideas on ways to improve, and trying to find the time to implement them. But it’s come a long way, and hopefully will go a long way further.

As always, I massively appreciate your support on the project, especially if you have made it this far into the article!