Search

Will the Patriots lose again in Super Bowl LIII?


It's easy to bag on the New England Patriots (pick your "-gate"), but every year they seem to be in the big game, and every year everyone NOT living in the New England area bitches about it (many of whom I hope clicked on the link of this blog simply for the title). The good news for all you Brady-Belichick-Patriots-haters is that according to my analysis, most of the stats point to a narrow L.A. Rams win.


Points scored/allowed by the L.A. Rams and N.E. Patriots over the 2018-19 NFL season.

Just based on season averages of points scored/given away (graph above), both teams have advantages in both offense and defense that balance each other out. The Rams offense tends to score more points than the Patriots (~3.8 points more on average per game), but the Patriots defense typically holds their opponents to less than what the Rams defense allows (2.5 points less on average). When you examine the ratio of points scored over points allowed, both teams produce almost the exact same number (~1.34). The Rams do have a +0.7 point advantage when predicting the final score based on the mean of the offense and the opposing team's defense season averages (see table below).


Season averages are based on all played games in the 2018-19 NFL season. The predicted score is the mean of the offense and opposing team's defense season averages.

One of the unique things about the NFL is that the final championship is played on neutral ground, which doesn't happen in most other American professional sports. Neither the Rams nor the Patriots will have a home stadium advantage, which will likely disproportionately affect the Patriots more than the Rams. Given in the graph below are the average number of points each team and side scored above/below what was expected at that point in the season based on their actual performance and what was predicted according to their season averages and their competition (see here and here for more information on my methodology). The Rams defense (on the left in blue) actually allows almost 3 more points at home (dashed line) than they do away (solid line) at this point in the season. The Patriots defense (same graph in red) tends to hold their opponent to 2 points less than expected both at home (dashed line) and away (solid line), which is about the same amount points the Rams defense allows their opposition at week 22.


Average amount of points +/- expected for New England (red) and L.A. (blue) over the 2018-19 NFL season. Data from the previous season is included in calculating averages for the current one.

There is a much starker difference in the home and away effect for the offense (on the right in the above graph); In particular, the tendency for the Patriot offense to do much better than expected when playing at home. At Gillette stadium, Brady and company score about 6 more points more than what you would expect based on their season averages and the competition they are facing for that contest. [Insert cheater comment here]. The Rams offense is well above expectations either home or away, with their unit scoring 5.2 and 4.2 more points respectively, with a p-value on this t-test no where near statistical significance (p=0.4016). To adjust for the location effect, I multiplied the average away effect for each team and side by its corresponding p-value, then I added this number to the season average (table below). Using this location adjustment, the predicted score advantage for the Rams increases from +0.7 to +2.76 points (Rams 27.87 and Patriots 25.11).

[Technically the Super Bowl isn't a typical away game because no one is playing at home, but it's definitely not a home game so I want to try to eliminate this effect from the final calculations.]

Location adjustment on season averages for Super Bowl LII. Data from the previous NFL season is included in calculating the home and away effect for all teams.

Two variables that I've found that have a significant impact on NFL games, besides location, is the number of turnovers and sacks each team gives up/takes away during the game. Looking at just the raw numbers (graph below), the Rams defense has a higher season average for both variables over the Patriots defense. For offense, both the Rams and the Patriots have the same season average for turning the ball over to the defense (~1.1 times per game), but Brady gets sacked almost half as less as the Rams Goff over the course of a game. One of the keys to success for the Patriots is protecting their biggest play-maker, and for the Rams, it will be for their defense to pressure Brady.


Density graph giving number of turnovers and sacks each team's defense gets (left) and each team's offense gives up (right) during the 2018-19 NFL season. Increasing thickness of the density graphs corresponds to more values at that level.

Since neither team is playing at home, we can just look at the average number of turnovers and sacks on away games for both teams to try to see differences. Given in the graph below are the away averages for turnovers and sacks calculated at each point in the season for both teams (I use the previous 16 games to calculate the average of each point). You can see that the Rams defense (left side of graph in blue) has consistently higher averages for turnovers and sacks when playing away than the Patriots defense (in red). Conversely, the Patriots offense (right side of graph in red) are slightly better at the Rams offense (in blue) in protecting the ball and their quarterback, but not as much of the advantage the Rams defense has over the Patriots defense. I give the advantage to the Rams in these two categories over the Patriots, but only because no team has home field advantage and the Rams defense tends to play better on the road.


Turnover and sack averages for New England (red) and L.A. (blue) at each point in the 2018-19 NFL season.

When I take everything together (season averages for points, turnovers and sacks by location), I have a final score of the Rams 34 and the Patriots 30 (table below). **[My turnover/sack adjustment tends to over-estimate the final score, but is more precise when predicting the point difference between the teams]. Obviously this is just an guess based on the available data, and this model doesn't take into account injuries that may occur during the game, or referees that screw up penalty calls (cough-cough-New Orleans). I do think that if the game is low scoring (under the 56.5 total points Vegas is giving for the over/under), then the Patriots will win. If it's over this total, I would say that the Rams should have the win.


Season average adjustments based on location, turnovers (TO) and sacks. Predicted score is the mean of the offense and opposing team's defense average and average-adjusted values.

I've been fairly accurate with my win/loss records for both these teams this past season: I'm 13-5 for N.E. and 14-4 for L.A.R. for correct predictions. I did predict with 65% confidence that the Chiefs would beat the Patriots at home two weeks ago (see below), and I'm sure the fact that Bill Belichick has played against and studied Andy Reid's teams for so many years probably nullified any advantage the Chiefs had at home. Maybe the fact that the Rams' head coach Sean McVay hasn't been coaching for very long in the league may play to his team's advantage? Who knows? I'm not a wizard.


I was in the 99th percentile last year. I swear! (I hate getting beat by the FiveThirtyEight algorithm).

Regardless of how it finally turns out, I'm hoping for a good, competitive game... and the stats seem to point to this scenario. Although I do not have any allegiance to any of teams playing in this Super Bowl (and likely not in the future given the Broncos' misfortunes), I'm usually a fan of the underdog and will be pulling for the Rams this Sunday. I will also say that as a NFL fan I frequently bag on the Patriots (who doesn't), but I do admire and respect the coaching abilities of Bill Belichick and crew, so I wouldn't be surprised if they somehow figure out a way to shut down the Rams explosive offense and come away with the win. Bet at your own risk! (Maybe I'll stick with squares this year?). :)

-Michael Edwards, Bioinfo Solutions LLC


P.S. I guess some of the Pats fans are OK... no offense, J. Flynn. :)


* Microsoft Excel for Mac 2011 and JMP Pro ver.14 were used to collect, analyze and display NFL data.


**The turnover/sack adjustment on predicted score was calculated for Sunday's game with the following equation: predicted score with location adj. + (7 - 3.4 x (turnover away ave.)) + (7 - 2.125 x (sack away ave.)) = final score.


21 views

CONNECT

Bioinfo Solutions, LLC

Parker, CO, United States

P: (720) 580-9941

E: michael.edwards@bioinfosolutions.com

ABOUT BIOINFO SOLUTIONS

Dr. Michael Edwards

Contact

STAY CONNECTED

Join the mailing list:

© 2018 Bioinfo Solutions, LLC.  All rights reserved.