Welcome to the 2023 Orb Season!

Welcome to the Orb and to the 2023 season! We couldn’t be more excited to have the NFL back and to use data to try and beat the odds again this season. In preparation for our second full season, we have made a lot of changes to the Orb and will walk everyone through them shortly. If you have been with us since we started the Orb in 2021, welcome back. To those of you that are new and this is your first time reading our content, we are so thankful you decided to join the community newsletter. If you have any questions or comments throughout the season or about any of our posts please feel free to reach out or comment on any of the blogs.

Before we dive into the details about what changes we have made to our algorithm, let’s quickly recap the Orb’s Superbowl picks and put a bow on last season. In what was the lowest quality blog we will ever put out, the Orb made up for our sins with a perfect game:

• Chiefs +1.5

• Chiefs moneyline

• Over 50.5 points

• Overall: 3/3, +2.78 units

You love to see it. This is the type of good vibes and data driven accuracy we hope to bring in for the start of the new season.

Since the original algorithm was written in the playoffs of 2021, the changes we made to our strategy this offseason are the biggest to date. Last season had its ups and downs from the accuracy perspective, and we learned a lot about how we can improve our code and strategy. For those of you who followed along last season you’ll remember we gave out every pick for every game: moneyline, spreads and over/unders. This is not a sustainable strategy and it was never in our long term goals to continue to try and predict every game. Probability will always drag you back down toward the 50% mark. Instead our new strategy is to give out less picks with a higher confidence of accuracy. Let’s break down how we are going to do this. For starters, we sadly are no longer going to be giving out over/unders and instead putting all of our focus on spreads and moneyline. Last season we used multi-linear regression modeling to predict the actual score of each game and then derived the picks based on our model’s output. This was a solid strategy for starting out but is no longer in our road map moving forward. Instead we are switching to a multi-model, non-linear approach. In the offseason we developed and tested three new non-linear machine learning models and based on testing, are going to use the combination of all three when giving out picks. This means we are only giving out the picks that all three models agree on, meaning each pick comes with a naturally higher confidence of three separate models. Please see below for a visual representation of our muti-model strategy. For simplicity we are going to refer to our models as Model A, Model B, and Model C: 

For testing we removed 30% of our dataset, trained each model on the remaining 70% and then tested the models against the 30% testing data. This way we can see how our models would have performed in games using real data that it wasn’t trained on. The results from our testing are very exciting to say the least. A reminder that for spreads, the magic number that we are trying to beat is 52.3% - the breakeven point of profitability on -110 picks. Remember, past performance is not a guarantee of future results, but we at the Orb sure hope they are. In the testing data that we used, the favorite ended up covering the spread in 44% of games. This means that if you happen to pick every single underdog to cover you would have been right 56% of the time. We tested all three models against this data not only to see how they performed individually, but to determine their accuracy when they agreed on the same picks. Below you can see Orb 2.0’s testing results against spreads:

Not only did each model individually perform above our target of 52.3%, when combined together (A + B + C) it gave us something to be very proud of. We of course tested the same models against the moneyline results of our testing data. Moneyline is obviously easier to predict which is baked into the odds. Predicting a favorite is going to win isn’t exciting or brave but can be still be valuable information. In this same testing period the favorite won the game outright 69.5% of the time. Meaning if you picked the favorite to win every single time, you would be right in just under 70% of games. Here are the results of our moneyline testing:

So we just through a lot of information at you guy, here are the high level bullet points of all the changes we made this offseason:

• Changed our modeling strategy

  • We went from one linear regression model to multiple non-linear models

• We are not picking every outcome of every game

  • Our vision is to give out less picks with higher confidence

  • As part of this goal, we are no longer predicting over/unders

• Our model testing was very successful

  • This does not guarantee that the Orb’s hit rate this year will be as high as our testing accuracy against a smaller sample size (70.4% for spreads, 85.7% on moneyline)

We have a lot more content coming your way this season that we hope you’re going to enjoy. Week 1 picks will be coming out next week as we get closer to kickoff. This project of the Orb and trying to build the most accurate predictive model possible continues to be a fun journey and we are so appreciative that each and every one of you is following along for the ride. With that said, I want to give a special shout and and thank you to my co-founder and our creative director here at the Orb, Andy Diamonds. We couldn’t do any of this without him. Let’s have a great season everyone!

- Team Orb Analytics

.

.

.

.

.

.

DISCLAIMER:

The information provided on this website is for informational purposes only. It is not intended to be gambling or financial advice, and should not be relied upon as such. We are not responsible for any actions or decisions taken by readers based on the information provided on this website.

The picks and predictions provided on this website are based on our own research and analysis, and are intended to be used for entertainment and informational purposes only. We do not guarantee the accuracy or completeness of the information provided, and we are not responsible for any losses or damages incurred as a result of using this information for gambling or other purposes.

By accessing and using this website, you acknowledge and agree to the terms of this disclaimer, and you assume all risks and liabilities associated with your use of the information provided on this website.