Week 10 Newsletter and Picks

Week 9 belonged to the favorites and the Orb! The public got one back on Vegas this week as favorites covered in 10/14 games and won outright on 12/14. It was also a historic week for our models, as they officially make it four +unit weeks in a row. The Orb finished the week 5-1 against the spread and 8-1 on moneyline picks as it hit every pick that didn’t involve the Carolina Panthers. Our models had them upsetting the Colts outright, and therefore covering, so our perfect week will have to wait once again. But it was an amazing week to finish off the first half of what has been a dream season for us and our predictive models.

I typically try and highlight picks that miss more often than ones that hit, but I have to talk quickly about our favorite result so far this season. In the most fun game of the day, and maybe the season, the Orb predicted that the Texans would win the game outright but the Buccaneers would cover the three-point spread. This left us with an incredibly small window of outcomes that would result in both picks hitting. But through an injured kicker, the line that Baker toes between confident and cocky, and pure chaos, CJ Stroud led a game-winning touchdown drive in the final seconds of the game to go up two and then knelt out the extra point attempt. When the Orb makes a pick on a game's spread and moneyline, there are three possible double-winning scenarios. The first is that it picks the favorite to win and to cover, and they blow out the opponent. The second is that it picks the underdog to win outright, and therefore cover by default, and they do. The final one is what we had in the Texans/Buccaneers game where it picks the favorite to win and the underdog to cover. In honor of the performance he put up in a loss but a cover, I am going to refer to this kind of double winner as the ‘Baker’ moving forward. The second type of double winner, the underdog winning outright, should be my favorite since it results in a higher unit return, but there's something special from a modeling perspective about correctly predicting a Baker. Here is how the rest of the picks from Orb performed in week 9:

Spreads:

• Chiefs -1.5 ✅

• Buccaneers +3 ✅

• Ravens -5.5 ✅

• Panthers +2.5 ❌

• Eagles -3 ✅

• Bengals -1.5 ✅

All spreads: 5-1, 83.3%, +3.545 units

Moneyline:

• Steelers (TNF Premium) ✅

• Chiefs ✅

• Texans ✅

• Saints ✅

• Ravens ✅

• Panthers ❌

• Eagles ✅

• Bengals ✅

• Chargers (MNF Premium)  ✅

All moneyline: 8-1, 88.9%, +3.613 units

Overall a +7.158 unit week 9 from the Orb! This extends our season best run of +unit weeks to four in a row. We know -unit weeks are inevitable in this game but are thrilled to have another chance to try and keep the streak going. Since it is officially the halfway point of the regular season, we wanted to do a little deeper dive than usual in our season performance so far. Let’s see the numbers behind the first half of our season:

Season-to-date results:

• Season spread picks: 37-17-1, 68.1%, +16.633 units

• Season moneyline picks: 60-24, 71.4%, +8.267 units

Our internal emphasis on spread picks has paid off so far as they are accounting for the majority of our +24.9 unit season so far. We saw a -unit week from our moneyline picks in week 7 but since then the models have turned things around and are heading in the right direction along with our spreads. 

Next, we wanted to show you how our spread picks are performing by side via a confusion matrix. If you haven’t seen these before, it breaks down binary predicted values by actual outcomes. This gives you a sense of how our models have performed when picking favorites vs underdogs against the spread as well as shows true/false positive and negative predictions:

In the first half of the season, when the Orb predicted the favorite to cover it went 18-5 compared to when it took the underdog is picked 19.5-12.5 with one push. On the flipside, in games that the Orb made a spread pick and the underdog covered, it predicted 79.6% of them whereas when the favorite covered it was 59% accurate. 

Finally, I wanted to take a look at how our results compare to the model testing we did going into the season. Just as a quick recap for anyone who missed our welcome to the season blog, I tested our models against spread picks by removing 30% of the training data, training the models on the remaining 70%, and then testing against the 30% test data to see how they would have performed in those games. For anyone looking to get into analytics, this is a fairly common method of testing model accuracy against past data before turning the model live for real-world results. It is a safe way to see how your model theoretically would have done against the test data. The graph below is a breakdown of performance by individual model, as well as by each combination of models. As a reminder, the profit line for spread picks, and the goal of the Orb project, is picking above 52.4%. I remember this testing brought conflicting feelings of excitement and worry. The models combining for 70.4% accuracy showed real promise of what this algorithm could do theoretically but also set the bar at an unattainable height that I never in my wildest dreams thought we could replicate in practice once the season started: 

Now let's compare these testing results against how our models have performed so far halfway through the season. Below you will find a similar chart, breaking down pick accuracy by each model and combination of models:

We still have half of the regular season to go and our models are bound to miss and have bad weeks, but I’m really proud of the numbers above. Even with the notable decline in Model A’s performance from testing to real results, each individual model is picking above the 52.4% target line against a 120-game sample. As you go down each row, you also will notice that the sample size of picks diminishes as you start to combine modes which was one of our goals this offseason. The final combination of all three models, also known as the Orb, is hitting slightly below how it performed against the test data but it is doing so at double the sample size. Again, past performance is not a promise of future results but if you will allow for a 2.2% variance, I am happy to report that we were able to replicate our previously thought unattainable testing benchmark in the first half of the season. 

Thank you for allowing me to ramble on and nerd out on the performance metrics there. We dug ourselves into a hole to start week 10 as the Orb once again picked the Panthers not only to cover but to win outright Thursday Night. They at least managed to push the spread as the game ended 16-13 in favor of the Bears, but we now enter Sunday -1 unit to start our week. Now for the reason you are probably still reading, here is what the Orb predicts will happen on Sunday in week 10:

Spreads:

• Panthers +3 🟨

• 49ers -3

• Lions -3

• Giants +17.5

• Commanders +6.5

Moneyline:

• Panthers ❌

• Colts

• 49ers

• Saints

• Steelers

• Lions

• Cowboys

• Seahawks

We have loved hearing from you the last couple of weeks! A lot of work goes into the Orb behind the scenes both on the data side as well as branding. It has been red hot the last few weeks but even if the models don’t always hit, the support we have received recently in the comments here as well as our social media pages makes it all worth it. So on behalf of Andy Diamonds and myself, I just wanted to say thank you!

- 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.