DadBodFootball Week 7 NFL Picks Against the Spread, Totals, and Implied Totals
DadBodFootball Week 7 NFL Picks Against the Spread, Totals, and Implied Totals
Back On My Bull****
Okay, fam - I admit that I’m embarrassed about not writing my feature length post for the past few weeks. Life gets busy some times and while I felt like my early lines preview got many of you enough of what you wanted and needed, something was probably still missing.
The Early Lines posts are great for quick hits and for those of you who like to get your bets placed before the lines and totals move closer to game day - that said, there is also value in reviewing both posts to see how things have moved/changed… most notably when teams release injury designations for Sunday’s games on Thursday.
All told - this week’s edition will still be short and sweet, I want to focus more on the content that I haven’t posted in a while and share the WAR deep dive from the early article.
Exciting News (To Me) - one of the most effective tools/visualizations for helping me determine what bets to make is the Winning Quadrants below. Now that we have 6-7 games of data on teams, we largely know who they are. Explanation is below, but this helps me identify where entirely unrelated data (how many TOTAL points are scored, on average in a game for a team; and how many net points do they win or lose by).
Chasing Zebras - Wins.Against.Replacement (W.A.R.) Ready
I teased this last week, but the WAR model effectively seeks to help us better understand how the model is returning value Week over Week. Historically, I looked at Wins and Losses as well as confidence based betting as a way to contextual if the model was winning big and missing small.
As a reminder, the model is guidance on every NFL Spread and Total - it is not intended to be accurate for every game. What we’re trying to do is hit our most confident bets - our bets that have the largest delta between the DBF Line/Spread and the Vegas Line/Spread. Here’s an example.
Team A playing Team B -
Vegas Line: Team A (-3)
DBF Line: Team A (-3.5)
DBF Confidence/Delta: 0.5 - the 0.5 difference between Vegas and the model
Team C playing Team D -
Vegas Line: Team C (-8)
DBF Line: Team C (-5)
DBF Confidence/Delta: 3 - the 3 difference between Vegas and the model
By calculating our net delta of wins minus the net delta of losses, we can determine how well (or poorly) we performed relative to the model’s ability to identify skewed spreads and totals. So..
If the model correctly predicts 3 outcomes
Game 1: DBF Confidence/Delta 5
Game 2: DBF Confidence/Delta 3
Game 3: DBF Confidence/Delta 3
And the model incorrectly predicts 5 outcomes
Game 1: DBF Confidence/Delta 1
Game 2: DBF Confidence/Delta 3
Game 3: DBF Confidence/Delta 2
Game 4: DBF Confidence/Delta 1
Game 5: DBF Confidence/Delta 0.5
We could communicate that the model went 3-5 OR we could determine if we hit more confident bets and lost less confident bets. The Correct Delta total is 11 (5+3+3) - the Incorrect Delta total is 7.5 (1+3+2+1+0.5). Our prediction delta is +3.5 -
To get to a WAR - I use the formula: WAR = 136*(Correct Delta / Incorrect Delta)-136.
Why 136? There are 272 NFL Games, meaning 272 possible spreads/totals to bet - if you were to flip a coin, you’d get 136 right (in theory…). So the model tells us how many additional wins we’d get by hitting more confident bets above the 136 bets you’d win if you just blindly picked games…
Okay - so how did we do both in record and in W.A.R.
Spreads - Correct % of all spread predictions each week
Totals - Correct % of all total predictions each week
Spread WAR - How accurate were we at predicting confident spreads (larger = better)
Totals WAR - How accurate were we at predicting confident totals (larger = better)
Some interesting notes - In week 3, we hit on 31% of spread predictions… That’s a terrible win rate (4-9), BUT! we were actually more accurately able to identify winning confident bets than the previous week when we won 58.3% of our picks.
What we’re aiming for here is continuing to improve the WAR numbers, and if we are negative, keep that as close to 0 as we can. Win Percentage is important but I model every NFL Game - the expectation is NOT that you bet every game, in fact, I generally only follow the model for larger delta spreads and lines (higher confidence). Hitting a higher % of confident bets (Higher WAR rates) - means the model is doing what it does best, it is identifying where NFL Spreads, Totals, and Implied Totals are not aligned to reality/the underlying data.
Spreads, Totals
The OVER carried 11 of 14 matchups last week, so Vegas ramped up points to 671 (715.7 in a 16 game slate). That’s why the model is leaning UNDER this week. As a reminder, I anchor the model’s expected points to Vegas’s expected points.
The Commanders Total and Spread both saw substantial movement Tuesday to Friday
Spread: Commanders -7.5 —> -9.5
Total: 49.5 —> 51.5
The model still feels good about the Commanders and the OVER, but I have reservations - mostly that the Panthers can score enough points to break a 51.5 point spread.
Implied Totals
For those new to implied totals (or wondering what the heck I’m talking about) - an Implied Total is the expected points for ONE team in a matchup (with a TOTAL being both teams).
DBF Data points - Interpreting Implied Totals: In the DBF Tracker below, I include the Implied Total and delta to two other important data points
Average Points Scored: Informs how many points a team scores on average against a neutral opponent
DBF Points Expected: The Model’s output for expected points including adjustments for opponent defense, weather, home/road performance, injuries
This allows us to calculate the DOUBLE DELTA - how far off both average points and DBF Expected Points an implied total is. The more negative or positive, the more… theoretically… likely it is to be OVER or UNDER.
Winning Quadrants - Do Teams Win And How?
This is a mainstay of the weekly post - visit Week 10 of the 2023 season for a longer description Winning Quadrants graphic from Week 10. I’ll be adjusting to 2024 data right around Week 6.
Y-AXIS Net Points: The “higher” a team is on the Y Axis - the more team “Wins” by, and the “lower” - the more a team loses by.
X-AXIS Average Total Points (net 2023 NFL Scoring average of ~43.8 PPG) -The left-most teams participate in games with the lowest TOTAL points - the right-most teams participate in the highest scoring games. I netted (subtracted) season scoring average to show if they are above or below the average game.
How can you use this? Well - it’s an interesting way of aligning our “feel” about certain teams - the Dolphins tend to win and score a lot of points - and identify how they generally align against their spreads/totals.
For instance, the Buccaneers and Ravens are participants in some of the highest total games this season. I would hammer the over - it was 48.5 on Tuesday, and is 49.5 when I’m writing this article on Thursday.
NFL Week 7 - 3 Pack
This new section is where I go a bit deeper on how the data and analytics in this post can be applied to a given week’s matchups. I wouldn’t say that these are picks - more how I’m feeling about a week’s set of lines, totals, and implied totals.
*** NOTE *** - I generally write these on Thursday/Friday before game day, so the lines may NOT be exactly as they are below, but the notes won’t change substantially.
Miami Dolphins Implied Total 20.25 - UNDER
The Dolphins average 12.0 points per game, and though the colts defense is middle of the road (20th of 32 in the model), the Dolphins have little to no passing offense. The DBF model expects them to score 14 points - giving us a double delta (see above for more context) of -14.8.
I would also consider taking the Colts -3 as a bonus.. and wouldn’t touch the Over/Under.
Tampa Bay (+160) vs. Baltimore Ravens - Over 49.5
Much of this prediction hinges on Mike Evans playing, but the Bucs are a well rounded offense and Baker has familiarity with Baltimore’s defense (although I will argue that these aren’t your father’s Ravens). The model expects the Bucs to win at home (which is without any home bonus in the model) and it LOVES the Over… a lot. The Ravens and Bucs are the #1 and #3 scoring offenses in the NFL. A Total less than a touchdown above the NFL scoring average for the 2024 NFL season seems… low.
Cincinnati Bengals (-5.5) vs. Cleveland Browns
The Browns are in Deshaun Watson contract hell, they just traded their best offensive player, and the model loves the Bengals to win. Bear in mind that the model only adjusts for historical performance, but this Browns team sucks by every statistical measure in the model… I’d also hammer UNDER 18 points for the Browns Implied Total.
LA Chargers vs. Arizona Cardinals UNDER 44 points
The chargers are winning in games that average nearly 15 points less than the season long NFL scoring average of 44 points (29 PPG TOTAL). The Total this week for their primetime matchup with the Cardinals is… 44… Take the under.
Model Performance
Reminder - the model GUIDANCE and not PICKS. I run every game through the model to identify opportunities where the underlying scoring and efficiency data does not align with the Vegas Spread and Total. I’ll continue to include WAR score here and a link to the post explaining why / how we can use it to determine if we’re hitting our most confident bets.
On Spreads, we went 9-4 with a WAR of 122.6. This means that not only did the model accurately predict a majority of spreads, but it correctly predicted the highest confidence/largest delta spreads too.
On Totals, we went 6-7-1 with a WAR of -34.2. This is the worst week to date in terms of predicting totals confidently, but is also more of a byproduct that NFL scoring was well over expectations - with 617 total Vegas points to 689 actual points scored.