Week 2 NFL Power Ratings: Projected Odds For Every Game
Chris Graythen/Getty Images. Pictured: Tom Brady
Looking for an edge?
Our model projections help power our NFL PRO Report — a tool that automatically highlights the biggest betting edges across multiple of categories, including sharp action — but we’re offering a free glimpse into that data by outlining our top NFL experts’ consensus power ratings for Week 2 of the 2020 season.
You’ll find their projected spreads for every game below, followed by their team ratings and methodologies.
Week 2 NFL Power Ratings
Compare their consensus projections to real-time spreads at various sportsbooks with our NFL odds page. Projections as of Sept. 16.
NFL Team Power Ratings
Sort this table by our experts’ consensus or individual ratings. You can learn more about each expert and their methodologies below. Ratings as of Sept. 16.
NFL Power Ratings Methodology
Sean is The Action Network’s Director of Predictive Analytics and has a 203-140-4 (59.2%) record on NFL bets he’s tracked in our app heading into the 2020 season.
My power ratings reflect how much each team is worth against the spread on a neutral field. I treat these projections as a baseline before I factor in home-field advantage, weather, matchup and any injury-related news. I also have player ratings that allow me to estimate how much a team’s rating would rise (or fall) depending on that player’s availability.
One of the primary goals I’m trying to achieve when creating my power ratings is to find discrepancies between public perception (based on a team’s win-loss record) and how a team may be benefiting from factors that can be driven by luck.
Once I have a sense of what I think the line should be, I then compare it to the market. Typically if there is a gap of 2 or more points between my projection and the point spread, I consider placing a bet.
Chris is a senior analyst and has a 249-189-15 (56.8%) record on NFL bets he’s tracked in our app heading into the 2020 season.
My power ratings represent a team’s expected point differential vs. a league-average opponent on a neutral field.
They’re standardized so that they are on the same scale as the point spread –with a mean and median of 0 for a league-average team — so that you can simply subtract one team’s rating from another to get the projected point spread on a neutral field, and then add in a home-field advantage adjustment.
“League average team on a neutral field” is also another way of saying that these ratings represent a team’s “true talent.”
The true-talent method is an approach to forecasting that can be applied not just to point differential, but to any metric you’re trying to predict. It’s a method I employ across all forms of NFL forecasting, and it’s helped me to maintain a 56.8% win rate on all NFL bets tracked in the Action App as well as place-fourth in FantasyPros‘ weekly rankings accuracy contest last year.
What this method entails is creating projections by incorporating known information and a regression to the mean component, with the known information being weighted more and more heavily as the sample size grows. This is crucial since the NFL is a small-sample sport where teams only face 12 unique opponents each regular season — luck, injuries and weather are among the variables that are difficult to predict but have a large impact on the outcome.
In this case, my rating for a team’s “true talent” as it pertains to point spread is really just a combination of true talent projections in metrics, which based on historical data, have shown to most accurately predict point differential.
For example, projected passing efficiency is the most heavily weighted factor in my model, and one of the key metrics I use is Adjusted Net Yards per Pass Attempt — it correlates strongly to point differential on a team level and stabilizes after roughly 325 dropbacks on a player level. This means if a team’s starting quarterback has 325 career dropbacks, I know I can take 50% of his career ANYA plus 50% of the league average — or some form of the league average more specific to that player, such as league average for all QBs with his level of experience, or draft pedigree, etc. — and get a relatively strong indication of his team’s “true” expected point differential.
A disproportionately large portion of the game is going to be affected by passing efficiency and home-field advantage (and luck), so while I do incorporate intangible factors, I’m always careful to have a data-driven method for doing so. If there’s a large enough sample on a coach, such as Bill Belichick or Mike Tomlin, I will factor it it. But I’m not creating coaching ratings for every team — that would just decrease the accuracy of the model.
The same is true for adjustments based on circumstances that pertain to a particular week.
The main adjustments are going to stem from injuries. I adjust for all injuries, but injuries to players who have the largest impact on projected pass efficiency — such as QBs, or elite pass-catchers or offensive linemen in some cases — will have a much greater effect than an injury to a running back or run-stopping linebacker, etc. And basically no non-QB is going to be enough to move the spread on his own. It’s when injuries really start to add up at one position (i.e. a “cluster”) that there’s an impact.
Check out our new NFL PRO Report, where we highlight factors that provide betting edges — like large wagers, historically profitable betting systems, model projections and expert picks — that when combined with sharp money can powerfully detail the smartest bets on a given slate.
Stuckey is a senior analyst and has a 328-287-9 (53.3%) record on NFL bets he’s tracked in our app heading into the 2020 season.
My personal power ratings are built on a hybrid of my team and player projection models. The ratings symbolize what Team A would be favored by against Team B on a neutral field in a vacuum. An average team is rated at 0 points, which means a team with a 7 power rating would be a 7-point favorite over an average team on a neutral field.
From there, I layer in home-field advantage to derive my base line for any game. However, that’s just a starting point.
Other factors must then be considered, some of which my colleagues mentioned. Is the team coming off of a bye? Is it a divisional game? What is the weather? Travel? Injuries?
The most important adjustment each week is then looking at each matchup. How well does Team A match-up with Team B based on their personnel and schemes? From there, I can determine my edge based on my subjective and objective inputs.
Each week throughout the season, I will also adjust my power ratings based on the results of the previous week. This is when a lot of the dirty work happens as final scores and total yardage numbers can be very misleading — you have to remove the garbage time and luck while accounting for time, situation and opponent.
Travis is The Action Network’s Data Manager and the brain behind our weekly Survivor Pool recommendations.
My power ratings indicate how many points each team would be favored by if they played a league average team on a neutral field. Because points in an NFL game are not created equal, this means that my power ratings are not linear, meaning you can’t just take the difference between two teams’ power ratings for a game line.
Other factors such as divisional games and rest will also influence the projected lines. You can see my projected win percentages for every game during the season in my weekly Survivor Pool article.