Model Context: Week 17

As explained here, week 17 is chaos. As expected, the model spit out some interesting stuff. I wanted to take some time to explain the results, which can be seen here.

The essence of machine learning is to feed a model data that the machine can then use to make decisions. Data being the key word here — with bad data will come poor machine learning models. Garbage in, garbage out. When the Kansas City Chiefs take the field on Sunday, they’ll be a different team then we’ve seen all year due to resting their starters. As such, we have no data on that team and the model will do a poor job predicting their performance. The solution is to make adjustments, but with that we’re introducing a degree of human element that I’m not totally comfortable with. The key to gambling is to realize that you’re a big dumb idiot and you need a model to tell you what to do.

Even after our adjustments, the model spit out Kansas City and Pittsburgh bets with strengths of 7.65 and 6.64, respectively. These strengths are far higher than any we’ve seen this season. Both teams are expected to rest starters, and we can therefore conclude that the model is simply wrong on these two games. It’s my opinion that the market is over-correcting for the starters resting, but I don’t bet using my dumb brain, so we passed.

The model also picked Buffalo (pick ’em) over Miami. As explained in my week 17 motivation post (linked above), the Bills will either end up being the 2 or 3 seed. Since the 2 seed no longer has the benefit of a first-round bye, there isn’t as big of a difference between the two seeds as there was in years past. However, it was my opinion that the Bills were likely going to play their starters, and due to the massive strength the model gave (4.39), I was happy to make that bet. But we were too slow to the punch — from the time we posted the model to the time we recommended bets; the line had moved to Buffalo -2. Dom and I both have full time jobs, and also like to discuss the games before making official recommendations. That’s our bad, but we won’t do shady shit like recommend them as a pick em when nobody would ever have that line available. That’s for the scammers.

One other game that deserves context is Arizona +5 over the Rams. The model grabs spreads from the Vegas consensus, which you can see via Vegas Insider. At the time the model was run, the consensus did list Arizona as a 5-point dog against the Rams. At the time, we knew Jared Goff was going to be out, but the books had concern over Kyler Murray (leg) as well and weren’t listing a spread at all. I was quite confident Murray was going to play (and we now know he likely will) and would have been happy to bet the game, but since I couldn’t find Arizona +5 on any sportsbooks, I couldn’t in good faith recommend it. I honestly can’t answer why the consensus was listed as +5, I imagine it was a holdover from a previous line.

There you have it, a couple games that we had to throw out because we don’t trust our ability to adjust for starters resting and us just being good guys and not recommending bets you could never make. Happy New Year, and here’s to a better 2021.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Website Powered by WordPress.com.

Up ↑

%d bloggers like this: