NFL Power Poll, Week 4

Rank	Team	Record	Score	Last Score	Difference
1	Pittsburgh Steelers	3-1	23.23	12.87	10.36
2	Philadelphia Eagles	3-0	22.97	22.97	0.00
3	Seattle Seahawks	3-1	19.48	22.39	-2.91
4	Denver Broncos	4-0	17.61	16.27	1.33
5	San Diego Chargers	1-3	16.36	29.40	-13.04
6	Green Bay Packers	2-1	16.21	16.21	0.00
7	Atlanta Falcons	3-1	14.62	0.20	14.42
8	Dallas Cowboys	3-1	12.86	19.16	-6.31
9	Minnesota Vikings	4-0	10.19	-4.71	14.90
10	Houston Texans	3-1	10.06	11.13	-1.08
11	Baltimore Ravens	3-1	9.37	9.02	0.35
12	New England Patriots	3-1	5.29	22.34	-17.06
13	Buffalo Bills	2-2	5.28	-10.27	15.55
14	Oakland Raiders	3-1	3.01	8.52	-5.51
15	Carolina Panthers	1-3	2.59	23.96	-21.37
16	Arizona Cardinals	1-3	0.34	2.81	-2.47
17	New York Jets	1-3	-0.49	2.34	-2.83
18	Detroit Lions	1-3	-2.91	0.71	-3.62
19	Cleveland Browns	0-4	-2.98	-4.52	1.54
20	New Orleans Saints	1-3	-5.90	-13.83	7.93
21	Indianapolis Colts	1-3	-6.26	-13.59	7.33
22	Cincinnati Bengals	2-2	-7.54	-13.62	6.08
23	New York Giants	2-2	-8.21	8.76	-16.98
24	San Francisco 49ers	1-3	-9.72	-13.26	3.54
25	Kansas City Chiefs	2-2	-10.54	3.81	-14.35
26	Tennessee Titans	1-3	-10.94	-11.97	1.03
27	Miami Dolphins	1-3	-12.07	-11.68	-0.39
28	Tampa Bay Buccaneers	1-3	-15.47	-13.26	-2.21
29	Los Angeles Rams	3-1	-15.67	-23.96	8.29
30	Chicago Bears	1-3	-17.23	-28.35	11.12
31	Washington Redskins	2-2	-17.75	-23.08	5.33
32	Jacksonville Jaguars	1-3	-25.50	-30.02	4.52

Biggest 1 week risers: 
1. Buffalo (+15.55)

2. Minnesota (+14.90)

3. Atlanta (+14.42)

Biggest 1 week fallers: 

1. Carolina (-21.37)

2. New England (-17.06)

3. New York Giants (-16.98)

Taking Stock:

In looking at the rankings versus records, 2 teams stand out. here I'll do a deeper dive to explain why these teams' rankings don't reflect their records. 

5) San Diego (1-3)

The Chargers were actually in the top spot last week, before falling to 5th place this week. Why are they ranked so high despite a losing record? Well, for starters, they've been leading all 4 of their games at the 2:00 warning. Not at the end of the first half, at the end of the game. Also, they are one of only 6 teams to have a net positive score in all 4 elements of the formula (the combined record of the other 5 teams? 14-5. Yes this team is 1-3, but they're really close to being 4-0. 

29) Los Angeles (3-1)

The Rams are just one of 4 teams with a negative score in all 4 elements of the formula, they arr also 3rd worst in 3rd down differential. If not for their defense forcing the 3rd most turnovers in the league through 4 games, they might be even lower on this list. 

The Forumla: 

I have very slightly tweaked my formula from last season. It remains broken down into four parts: 

Part 1: Yards per play. 

Here I take each teams yards per carry (rushing) and yards per attempt (passing) numbers and subtract from them the YPC and YPA their defense allows.  The theory being that, if Team A's offense is better per play than what their opponent's offense can muster against Team A's defense, Team A should be consistently better than their opponents over a full game's worth of plays (60 to 70 per game approximately). 

This part of the formula is unchanged from last year's model. 

Part 2: Toxic Differential
A better yards per play differential is helpful to a team's chances of winning, but just how often is an NFL team able to consistently drive down the field taking 5-8 yards at a time? You're essentially asking an NFL offense to put together 10-12 plays without more than 1-2 negative plays, be they incompletions, sacks, no-gainers, or worse: turnovers. It's doable, but it's really hard to do with any sort of consistency in a single game.

This is why coaches harp on turnovers so much. A turnover a) takes away an opponent's possession which decreases their chances of scoring more points, and b) can give your team a shorter field so you don't have to put together an 80+ yard drive to get points of your own. The problem with turnovers is you can't count on them. So much of what goes into a turnover is dependent on a) the other team and b) luck that relying on turnovers is a dangerous proposition.

So yes, turnovers are important. But there's something else that can make getting points in a drive much easier: big plays. If my offense can get 20 or 30 yards in a single play, that cuts out 4-6 plays of grinding, or 4-6 plays where something could go wrong. Now my offense only has to put 5-6 plays together on a drive where they also get a chunk play.

Brian Billick is credited with coming up with the toxic differential statistic. This adds your takeaways and big plays generated by your offense and subtracts your giveaways and the big plays given up by your defense. Again, the theory goes that teams with a better toxic differential will be better at turning drives into points and games into wins. Pete Carroll also bases his offensive and defensive identity around turnovers and big plays being the most important indicators for both sides of the ball.

Note: For this formula, a big play is considered a rushing play of 10+ yards or a passing play of 25+ yards.

I have tweaked the weight of this portion of the model again. In Year 1, this part was weighed too heavily. Last year I went too far the other way. This year I hope I have found some middle ground. 

Part 3: 3rd Down Efficiency
While turnovers and chunk plays make moving the ball down the field much easier, it is possible to crawl your way to points with long, sustained drives. However, you can't have a long, sustained drive without converting 3rd downs. If you're not hitting for explosive plays, you had better convert some 3rd downs, otherwise your drive will end in a punt, instead of points. 

This part of the formula is unchanged from last year's model. 

Part 4: Points Per Drive
What's the most important job of an NFL team? Score more points than your opponent. Rather than look simple points per game differential, I wanted to dig a little deeper and normalize the data a little further. Game-to-game the number of possessions can vary based on team tempo, weather coniditons, etc. So instead I looked at points per drive data for each team's offense and defense, and multiplied the difference by 10. Why 10? A typical NFL game has 12 possessions, but 1-2 of those come at a point where a team isn't really interested in scoring (maybe they get the ball with 12 seconds to go before halftime, or they get it with 3 minutes to go in the game up 14+ points already. 10 seemed like a good number of possessions per game where the end goal is to score points.

This part of the formula is unchanged from last year's model.