DBF Week 2 Update
Recap of a whelming week 1, why you always bet the under, and Week 2 predictions/notes.
Week 1 really had it all - from an absolute pillow fight between the Chiefs and the Lions (two of the top offenses from 2022) to the Hindenburg-esque implosion by the Bills against the recently-Rodgers-less Jets. Let’s dive in to the model, observations, and then predications for this coming week.
Week 1 Recap
The model selected 12 lines as favorable and hit 66% (8/12). The major theme for the games it missed - two top 10 offenses from 2022, the Vikings and the Bengals, could not get out of their own way and scored well under expectation against what should have been middling defenses. For the four games that were neutral (Commanders -7, 49ers -2, Saints -3, Eagles -3.5), two had the favorite not cover - the Washington Commanders and the Saints. I don’t know what this says about their respective seasons, but in times like these, I always default to the age old saying - Good teams win, Great teams cover (and boy are the 49ers and Eagles GREAT teams)…
The model predicted 7 of 13 over/unders correctly, with all 6 of the misses guessing the OVER (Chiefs/Lions 53.5, Falcons/Panthers 39.5, Vikings/Bucs 46, Seahawks/Rams 46, Giants/Cowboys 45). We’ll dive in to this in a bit as the focus area for the week. Interestingly, the three “no bets” generated by the model all hit the OVER (Colts/Jags 46, Pats/Eagles 44.5, Chargers/Dolphins 51). Why did the model have so many overs? I assumed a flat net scoring average and not a ramped one… this has subsequently been changed.
Overall, the model returned $130 for every $100 wagered (assuming +100 odds) - it could have been more, but the Chiefs - Lions (Lions +4.5, Over 53.5), Bengals - Browns (Bengals -2, Over 47.5), and Vikings - Bucs (Vikings -5, Over 46) were all confident bets that resulted in a net loss of 37 (the only negative games of the week).
Week 1 Observation
Five things I think I think, including the focus area deep dive -
The Bears are bad, bad… And this is more true with news that the entire org is feeling the heat to win in week 2.
The Browns defense looked very good from front to back.
Offenses are trending behind defenses - which is often the case early in the season due to 1. Poor line play/lack of cohesion 2. Lack of structure in TC/Preseason and 3. Defenses having the offseason to catch up. You may read this and say “Yeah, no shit”… but it was no more apparent in week 1 with Vegas expecting 715 points across all games and teams scoring… 656
The Top 10 offenses from 2022 scored 47 total points below expectation, and if you take out defensive touchdowns, it’s 69 total points. The Top 10 defenses allowed 26 total points less than expected, meaning they held opposing offenses to fewer points than expected. The defenses who underperformed (allowed more points than expected) were the Bengals, Steelers, and Pats - two of which played top 10 offenses (49ers and Eagles)… and one is the Bengals.
Deep Dive: Always Bet the Under…
The saying “Always Bet the Under” is a favorite amongst my friend groups - which contains the full spectrum of gambling enthusiasts - from risk averse interested parties like myself to individuals who place 5 leg prop bets because they “have a good feeling about the O’s 69th game of the season…” I think there is a piece of human psychology that also drives us to the over - what person doesn’t like an exciting game with lots of points? Well - Vegas loves that we love that, and uses it against us.
So how can we use history to become smarter about betting the over. If you ask Vegas Insiders, the under hits a shade over 51% of the time, the under 47% and the remainder is a push. I wanted to get more specific to understand are there trends in expected points or other metrics we can use to identify what weeks are most under-prone. I analyzed points totals from the first 6 weeks of the last 2 seasons to see if there was a universal scoring trend. I held on including 2020 because the COVID shortened/adjusted season saw teams scoring nearly 30 points more on average throughout the season than in previous years - even when adjusted for a 17 game pace.
Quick Shoutout to Pro Football Reference and NFL Pick Watch for access to historical data - seriously, the guys running these sites are the real MVPs of DBF.
The blue lines are points scored in a given week - the red lines represent how many points Vegas expected during the week. Two things stick out -
Vegas has reduced the average number of expected points each week - from 770 per 16 game week in 2021, to 720 per 16 game week in 2022. They expected 715 in Week 1 of 2023 (656 actual points scored) and are now projecting 702 at the time of writing this for Week 2 of 2023.
Vegas leans towards giving the under in early weeks, and even at it’s peak - it doesn’t, generally, go above 9 games to the over. In an average week, 6.5 of 16 games will hit the OVER - and with most O/Us ending in 0.5… the rest can reasonable be attributed to UNDERS (there were 5 “pushes” in 2022).
Here’s 2022 showing OVER/UNDERS hit each week
When you run the numbers - 54.5% of games hit the UNDER, 44.4% hit the OVER - which, while it doesn’t sound like much, saw a 30 game delta (or ~1.66 game/week) in favor of the Under.
How can we use this? Well - I want to give the model a few more weeks to run against 2023 data before I derive any hard insights, but for this week - I’m hammering the UNDER and cautiously betting the OVER (Lions, Seahawks?). And when in doubt, I’ll be taking the UNDER.
NFL Week 3 Picks
I’m looking to build upon the successes and lessons learned from Week 1 - most notably favoring the under where I can (and reducing expected points from each team). Separately, as I called out on Twitter - I am not a fan of betting the Thursday Night games, so know that even though I have projections - I will not actually be putting money on it.
I’ll continue to adjust the lines and O/Us with a finalized version Sunday morning that will be viewable in the shared Google Sheet.
