Saturday Feelings - Week 5
Coming off a big Week 4 and off to a decent start with Week 5’s Thursday night game, I am excited to see how the model performs this weekend.
Made a couple adjustments to the model this week that I still need to document entirely, but basically I realized I could squeeze out about 0.1% more accuracy by standardizing my football stat features, so that should slightly help the model predict ML and spread better. Unfortunately, I am still having some issues with the over/under model and need to make some serious changes. I am considering using a classification model for this as it could lead to more accurate predictions, and I will try to test that this week,
Anyway, back to Saturday Feelings.
Week 4 in Review:
What a week it was for the model. The model put on a show and went 26-12 with it’s predictions on non-optional bet predictions.
The most fun part of last Sunday (and this whole process so far) was after the first slug of 1PM EST games. The model had 6 SWEEPS in total, 5 of which came from those 1PM games - I was going crazy! All those wins in a row after such a sub-par Week 3 felt incredible. Not to mention how much it helped with marketing and visibility for the instagram page. On Sunday alone, the page gained over 300 new followers, all of whom wanted to get a piece of the predictive action the machine learning was putting out. See below for the full list of results.
I can’t believe my 49ers lost to the Eagles. It was so hard to watch that game.
Anyway, let’s take a look at Week 5!
This week I feel somewhat comfortable with the predictions. There is a chance that all these predictions could be incredibly wrong, and I am getting more comfortable with the fact that this very well could happen. I do worry that some of the followers don’t understand this as a fact and will be upset if they tail the model’s picks and don’t win because of it.
However, at the same time, I have been extremely clear that this isn’t betting advice or consulting and that all decisions people make with their own money has absolutely nothing to do with me. Hopefully people understand this. I will also be starting a series to explain why sports betting is not a worthwhile investment and how it should be used only for fun and entertainment.
Anyway, this week the model has a pretty even spread of outcomes. I do get a bit nervous when many of the games are predicted that the favorite wins outright but doesn’t cover. It feels like some kind of cop-out in a way because no matter what happens in the game, the model will get at least 1 outcome correct. We’ll see.
The most interesting thing to me this week is that the Falcons and the Texans are both 0-4. They have not won a single game this season; however, they are both the favorites in their respective games. This is something that the model doesn’t know or care about, as the model only cares about the statistics these teams have put forward. From a personal standpoint, I believe both these teams are going to claw and fight to make sure they do not drop to 0-5, but the model sees these teams by only their statistics and creates its predictions accordingly. The model has both games as losses for these 0-4 teams which is statistically likely, based on historical data. We’ll have to see how these games play out - they will either be 100% or 0% and that is a risk and potential shortcoming of the model.
The biggest surprise this week was that the model predicted the Bears over Brady and the Buccs and it happened! I have never thought to bet against Brady, but clearly the model saw something in the data that ticked it over to one side. I got a lot of great feedback from the followers and people loved seeing that the model was right.
Besides this game and the two 0-4 teams, the model has one other large surprise that I am not sure it can pull off…
Again, we have a Monday Night Football game featuring the New Orleans Saints as heavy favorites against the Los Angeles Chargers led by rookie QB Justin Herbert. After his phenomenal performance in Week 2 against the Chiefs, he has drawn recognition from fans around the country. His coaches just came out and said they fully believe in him and I’m hoping to see a great game out of him on Monday! Do they have what it takes to outright beat the Saints? My feelings say no, but the model says yes. I also got a wonderful comment from a person on instagram who said that in primetime games the public’s consensus is very typically wrong. We’ll have to see! Should be a good game either way.
Everything I do is for free because of my passion for sports and data science. I always post the predictions on instagram the day before gameday, but if you’d like the advanced analytics and predictions for the games days before everyone else, please check out the Patreon page linked here. In the same notion of this being free, every donation is appreciated if you’d like to help keep operations running, click the yellow button at the top of the page for our PayPal.
Check out our YouTube channel with our first video of the inner workings of the model.
Check out our merch shop here. And if you’d like to buy football data, check here.
And as always, good luck bettors!