DBF Week 3 Update
Recap of Week 2, was the OVERpalooza a mirage, when to bet the underdog ML, and Week 3 predictions/notes.
My fantasy teams are down bad, in large part due to early draft capital in HERO RBs who are now on IR or out for the foreseeable future. Week 2 strengthened both NFL Runningback’s position that they need better rookie wage scales and NFL Teams’ desire to not pay them…
Relatedly, watching Nick Chubb go down with a brutal knee injury and the impact it had on the Brown’s offense is a solid reminder that there are so many variables that can impact a game, and entire season, that can occur within a game.
These injuries will have a lasting impact on how Vegas evaluates scoring, spreads, and point totals in the coming weeks. I am factoring this in to the model, but the true impact of a star, “run the offense through this guy” player who isn’t a QB will be difficult to quantify. Players like Chubb, Saquon, Justin Jefferson, Jamar Chase, CMC all carry the weight of a season on their shoulders.
Week 2 Recap
The model selected 12 lines as favorable and hit 8 with 2 pushes and 2 losses (80% Win Rate with the pushes removed). Once again, the two losses saw two offensive powerhouses from 2022 fail to come anywhere near their points total (Jags/Chiefs) and the Patriots lack of offense ruin what was turning out to be a strong defensive performance against the Dolphins. The two pushes were last second scores by the underdog - in one case (Panthers +3), we’d have lost, and in the other (Eagles -6), we’d have won. Both of these were “slight” or $3/$100 bets so no harm, no foul. For the four games that were “Neutral” (Buccaneers -2.5, Cowboys -9.5, Broncos -3.5, Bengals -3.5) - two saw the road underdog come away with the WIN. Let’s dig in to that in a bit…
The model predicted 7 of 13 OVER/UNDERS correctly, with 5 of the 6 that lost picking the UNDER… This is on the heels of last week’s Nostradumbass level prognostication to “bet the UNDER” until Vegas course corrected - well.. they did that.. the OVER hit in 12/16 games, representing the highest % of games to hit the OVER in the past 2 seasons. More detail can be found in Last Week’s Newsletter - Vegas lowered the expected points total to 702 and teams responded by scoring 786… You’re bound to get blown up every now and again and, more importantly, Vegas already increased the points totals to 712 for this week. I expect we see a return to a more normal split (9 OVERS, 7 UNDERS is my guess).
Overall, the model returned $151 for every $95 wagered (assuming +100 odds) - the biggest losses of the week were the Jags - Chiefs (Jags +3.5, Over 51) and the points totals in the Bucs-Bears ( Over 41.5) and Ravens-Bengals (Over 46.5). For first time readers, the goal here isn’t to win every bet - it’s to maximize gains and minimize losses - so the model lost $33 on those games but won $89 across the other bets.
Week 2 Observations
Five things I think I think, including the focus area deep dive
Sean McVay has to bet on NFL Games…. Kicking a field goal with 2 seconds left to make it 23-30 (Rams +7.5) is wild… TYFYS!
Cleveland without Chubb and Cincinnati without Burrow are going to make for some volatile lines and point totals in Week 3 (see Early Lines publication).
It’s too early to declare the “Always Bet The UNDER” strategy as toast, and it is important to note that Vegas only moved expected points up to 712 this week (702 last week) - we could see another week that favors the over, but I wouldn’t expect another outlier. The Average Points scored in a given week in 2022 was 698. Don’t chase - stay true to the law of averages and know that statistical regression to the mean is the law.
How did we end up nearly 90 points above an average week? The top 10 offenses from 2022 (Chiefs, Bills, Eagles, Cowboys, Lions, 49ers, Bengals, Vikings, Seahawks and Jags) only overachieved by 4.5 total points (268 vs. 263 expected). The Top 10 Defenses substantially underperformed, allowing 39 points more than expected (248 vs. 209 expected). The bulk of the scoring came from games featuring middling offenses and defenses who have regressed from 2022 - Broncos/Commanders (39 Expected vs. 68 Actual) and Giants/Cardinals (40 Expected vs. 59 Actual). Those games covered the entire contribution of the top 10 offenses and defenses.
My takeaway, bet the over in games featuring top or middling offenses against low/middling or injured defenses. OVERS favor competitive games as teams are less likely to get blown out and take their foot off the gas… unless they’re the Cardinals (AGAIN, DANG IT). The four highest point totals of Week 2 saw an average “net win” of 4.25 points
Vikings 28 - Eagles 34 (62 total points, 6 point win)
Seahawks 37 - Lions 31 (68 total points, 6 point win)
Giants 31 - Cardinals 28 (59 total points, 3 point win)
Commanders 35 - Broncos 33 (68 total points, 2 point win)
Deep Dive - “Never Bet an Underdog to Cover, bet the ML”
While watching the Lions lose in typical Lions fashion, my friend made the ludicrous statement above - saying that only 16% of underdogs COVER in a LOSS. In other words, 84% of the time an underdog covers the spread - they do so in a win… So why not just bet the ML where you have better odds?
Well - for starters, it’s not 84%, but the actual numbers are more significant than I expected. Since 2022, there have been 171 games where the underdog covered or there was a push (56.25% of games). Of those 171, 109 saw the underdog win as well (63.75% of underdog covers - 35.9% of all games). That doesn’t seem like a lot, but over 1/3 of NFL Underdogs win - over 5 per week. So how can we use data to help us predict which dogs to bet?
First, we need to use these statistics to identify a break-even point. At what average odds does it make sense for us to bet the ML (or even parlay some ML bets)? Using the statistics above - if we bet $10 on every underdog to COVER in 2022, we’d win $1520 (152 covers * $10 at +100). If we bet $10 on every underdog to WIN, we’d need slightly north of +160 odds on average to hit the $1520 won betting straight covers. That doesn’t seem difficult, but we now need to evaluate WHAT kind of games are most likely to see the underdog WIN.
Data from Pro Football Reference and NFL Pick Watch
I’m choosing only to focus on 2022 for the purpose of identifying characteristics as some of the variables are skewed due to 2023 being a short season…
Question 1 - Is there a discrepancy between home dogs and road dogs?
Very clearly - yes, a team is more than 1.5 times as likely to win an “upset” on the road.
Question 2 - Are there particular match-ups (familiarity vs. lack of familiarity) that impacts this?
The “March Madness” philosophy at its finest - is there a correlation between underdogs covering and a team’s familiarity with its opponent. From the chart above, Divisional Games are the least likely to have the underdog win while Conference and Non Conference games share a meaningful gap. In other words, bet the underdog ML in Conference and Non Conference Games.
Question 3 - Is it actually possible/advisable to get to +160 odds? At what point do we see shrinking returns on odds?
Yes - of course…. over half of the underdogs who won had less than +150 odds at the time of kick-off. To me, the bigger takeaway here is that anything between +150 and +200 is viable. Those are where you can take some risk for upside - I wouldn’t go above +200 as your odds drop significantly (13 games in all of 2022 saw a team +200 win, and 5 of those were in week 1).
Deep Dive Summary -
Let’s run what we know now through our Week 1 NFL Model - in essence, I’m going to bet any DBF Picked Underdog on the road, in a Conference or Non-Conference Match-up, with better than +200 odds.
Week 1 - there were TWO DBF favored teams who fit this profile - Detroit (+170) vs. KC and Miami (+140) vs. LA Chargers. If we add in Divisional Matchups (there were many in Week 1 and Week 2) - that adds in LA Rams (+180) vs. Seattle and Bengals (+100) vs. Browns. The Bengals did NOT cover (21 point loss) - so that would have been a wash. The added $$ on Miami (+140), LA Rams (+180) and Detroit (+170) would have been $49 vs. the $30 we won going ATS.
My biggest takeaways here:
This is a viable strategy
Look for ROAD matchups where the away team is giving points but the model feels they shouldn’t be (larger delta, guidance towards road team).
Include Divisional for now as we’ve had 8 of the total 15 cover/wins come from these matchups in 2023 (granted almost every team has faced at least one divisional opponent, skewing prevalence).
Avoid chasing odds above +200 unless there’s something TRULY moving in the model.
NFL Week 3 Picks
Let’s keep up the momentum and employ a bit of the ATS opportunities we have above. I’m not favoring the OVER and have baked a similar adjustment to Vegas in expecting a total of 711 points to be scored this week (Vegas 711.5).
Two UNDERDOGS the model likes (and within our “rules”) to both COVER and WIN are
New Orleans (+2, +105)
Steelers (+2.5, +120) **DBF Edit - I mistyped this line as -2.5 initially
For the first time readers, these are picks ATS - the model likes/dislikes certain lines more than others, so the higher the delta between the DBF Line and the Actual Line, the “better” the bet.
Share this post, leave a comment, or reach out to me on Twitter @dbfinsights