F***ing With Stats: Breaking Down Pats vs. Falcons

[Nikolai]

Disclaimer: The game is played on the field. This is literally the outcome of ****ing with numbers, so take it for what it's worth.

The "TL;DR version" first (a more "detailed" explanation and a summary of the methodology below):

Ever since the Pats and Falcons punched their ticket to the Super Bowl, I've been making a mantra of my prediction; 31-23. I freely admitted that it was just a hunch. But I wanted to dig into the numbers, just to see if they undercut the media narratives following the Championship Games. The results were surprising to me.

The general idea was that I looked at the performances of each skill player on offense and how their performances varied according to the skill level of the defenses they faced. Since players on both teams demonstrated reliable adjustments according the defenses they faced, this was a somewhat easier process.

The final output is how player averages would look against a defenses of the caliber each will face in the Super Bowl (i.e. how Ryan, Jones, and company fared against comparable defenses to what the Pats have). The following are the projected stats:

Brady
26/38, 67.8%, 326 yards, 3TD, 0INT

Blount
17 carries, 153 yards, 1TD

Edelman
9rec, 109 yards, 1TD

------------------------

Ryan
21/33, 64.4%, 288 yards, 2TD, 1INT

Coleman
9 carries, 48 yards

Jones
8rec, 129 yards, 1TD

------------------------

PATRIOTS 37 | 7 10 13 7
FALCONS 17 | 3 7 7 0

------------------------

Basic Methodology

I'm not going to get into the weeds a whole lot, but the basic idea was to try and gauge how each team would do in the Super Bowl by breaking down their past performances at the lowest level possible. So instead of focusing on yards per game, or points per game, I started with the premise of yards per play per player. In the passing game, this meant yards per attempt, in the rushing game, this meant yards per carry, and in the receiving game, this meant yards per target (not reception).

After looking at a game-by-game breakdown of how each team performed, I analyzed the defense they faced while compiling those yards per play. To make a direct comparison, I looked at the yards per play for each defense they faced. I then calculated the deviation from the defense numbers; whether each player performed better or worse than what the defense usually allows.

I noticed both Brady and Ryan performed markedly better against teams they faced a second time. Because this is the first time these two teams are playing, the second game was removed from the analytic process. I limited this change to the quarterbacks only, based on the assumption that a critical factor for the better performances was a better pre-snap read of the opposing defense. Interestingly, this helped the overall stats of each quarterback, because while they improved game-over-game, those games were still below average performances.

From there, I needed to find the passing and rushing defenses that most approximated what each team will face on Sunday. I wanted 5-7 games for each player if at all possible. In some cases, such as Dion Lewis, the sample size is smaller, for which I allowed greater deviation during my adjustments. After isolating these games, I compared each player's yards per play against those defenses, took the difference (positive or negative) and then added those to what the opposing defense usually allows. I then multiplied that by the number of touches to get the stat lines for each skill player I evaluated.

Of course, both teams have role players that may take yards, catches, etc from the primary players I analyzed. During my analysis, it seemed much more likely that this would be the case with the Patriots, despite the fact that I included four WRs plus a TE in the analysis, compared to three WRs and a TE for the Falcons.

I started to whittle down these numbers further according to each team's YAC stats. However, it became quickly apparent that the yards per play on both offense and defense appeared to reflect each teams strengths and weaknesses when it comes to YAC.

After getting each player's individual totals, I reconciled the tallies of the RBs, WRs, and TEs with the QB stat line. After including the carries, I totaled the yards for each team. By this point, I already know how many TDs to expect, but I needed to compare the total yards to the yards per point allowed by each defense to account for how many field goals we might expect on top of those touchdowns.

A total stat breakdown:

Brady: 26/38, 67.8%, 326 yards, 3TD, 0INT

Blount: 17 carries, 153 yards, 1TD
Lewis: 12 carries, 53 yards

Edelman: 9rec, 109 yards, 1TD
Hogan: 4rec, 84 yards
Bennett: 4rec, 41 yards, 1TD
White: 3rec, 31 yards, 1TD
Mitchell: 3rec, 21 yards
Lewis: 2rec, 23 yards
Amendola: 1rec, 17 yards

TOTAL OFFENSE: 532 yards

------------------------

Ryan: 21/33, 64.4%, 288 yards, 2TD, 1INT

Coleman: 9 carries, 48 yards
Freeman: 12 carries, 38 yards

Jones: 8rec, 129 yards, 1TD
Gabriel: 3rec, 53 yards
Sanu: 4rec, 36 yards, 1 TD
Freeman: 2rec, 34 yards
Coleman: 2rec, 11 yards
Toilolo: 2rec, 25 yards


TOTAL OFFENSE: 374 yards

 

[Nikolai]

Quick Points

1. Blount could have a huge day. A good running back can kill the Falcons defense, and Blount feasts on teams with weak run defenses. I've heard more than a few Falcons fans express concern about Blount, and in this case, I'd have to say their concerns are warranted.

2. The Falcons numbers look really rough compared to what I expected. In fact, I almost wanted to fudge the numbers to make it look better for them. The fact remains that against teams that give up what the Pats defense does per play (particularly the pass), the Falcons struggle a bit. The one exception was the San Francisco game, where Atlanta took advantage of their dismal run defense. Unfortunately for the Falcons, the Patriots run defense is pretty good.

3. Shutting down Jones might not necessarily be the path to victory. Jones has played pretty well when the Falcons face tough pass defenses. He had four 100-yard games (out of five games) against teams that had a yards per pass attempt similar to the Patriots defense. However, the Falcons were 1-4 in those games. Ironically, the game they won was his worst game of the bunch; the playoff game against Seattle. Why did they win that game? See #5.

4. The Falcons seem to be at their best when Coleman and Freeman can get going. Of the top seven pass defenses the Falcons faced, the only two times when Matt Ryan played well above what the defense typically allows is when Coleman and Freeman are either rushing or receiving for big yards. My big takeaway is that stopping these two might be the most important "key to victory" for the Pats. In the playoff game against Seattle, the two backs accounted for 204 yards from scrimmage and 2 TDs.

5. Not once did Matt Ryan ever let a defense have a better than average day against him. Every single time Matt Ryan faced a defense, he took more than the defense typically allows. Brady was almost as good, but twice, opposing defenses had better than average days against Brady.

So that's it...

Nice and brief

EDIT - Got rid of the first point, because I confused myself at 1am when I wrote it.

 

[mgcolby]

Excellent post sir. Digging in weeds pays off sometimes. also nice to see a fellow numbers cruncher. I honestly would like to see your raw data, can pm if you like and if you have it.

 

[BobDigital]

Love the effort here and looks interesting. I can't imagine the Pats D holding Atlanta to only 17 points. I know your numbers don't lie but there must be some sort of factor that could not be accounted for?

Here is my score prediction based on numbers (not my actual score prediction). For this exercise I will do something much easier and base it off "the relativity index" of CHFF with a few minor tweeks. Lets assume both Defenses improved over the course of the year and do better than their average. I will call the improvement a wash and not adjust this. However I will only count the Patriots offense with the 12 games Brady played in the regular season.

For this example I will omit all playoff information.

The Pats average O score is 30PPG and their average D score is 15.3PPG
The Falcons average O score is 33.75PPg and their average D score is 25.38PPG

The Patriots O on average scores 7PPG (estimate) more than the D they face allows on average and their Defense holds teams to 5.63PPG less then their offense scores on average.

The Falcons O on average scores 10.92PPG more than the average D they face allows and their D gives up 2.05PPG more points than the average O scores.

Lets assume then these averages against the other.

The Falcons score should be either (15.3+10.92) 26.22 or (33.75-5.63) 28.12. Lets add both together and divide by 2. (27.17)

The Pats score should be either (25.38+7) 32.38 or (30+2.05) 32.05. Lets add both together and divide by 2. (32.22)

So the score on average should be Pats 32 Falcons 27 going by those numbers.

 

[luuked]

So the score on average should be Pats 32 Falcons 27 going by those numbers.

If you add in the INT that @Nikolai has projected you are getting somewhat close to his score.

 

[Patsfanin Philly]

You need to tweak the numbers little bit, you have Brady 26/38 but receivers making 38 catches. And the same for Ryan 21/33 but listed 32 catches.. Otherwise, excellent

D

A total stat breakdown:

Brady: 26/38, 67.8%, 326 yards, 3TD, 0INT



Edelman: 12rec, 109 yards, 1TD
Hogan: 6rec, 84 yards
Bennett: 5rec, 41 yards, 1TD
White: 5rec, 31 yards
Mitchell: 5rec, 21 yards
Lewis: 3rec, 23 yards
Amendola: 2rec, 17 yards

That' s 38 catches

------------------------

Ryan: 21/33, 64.4%, 288 yards, 2TD, 1INT



Jones: 12rec, 129 yards, 1TD
Gabriel: 6rec, 53 yards
Sanu: 5rec, 36 yards, 1 TD
Freeman: 4rec, 34 yards
Coleman: 3rec, 11 yards
Toilolo: 2rec, 25 yards

That's 32 catches

 

[Nikolai]

You need to tweak the numbers little bit, you have Brady 26/38 but receivers making 38 catches. And the same for Ryan 21/33 but listed 32 catches.. Otherwise, excellent



That' s 38 catches

------------------------

That's 32 catches

Thanks! That's supposed to be targets. That's what happens when you do late nights.

EDIT - Fixed it

 

[Nikolai]

If you add in the INT that @Nikolai has projected you are getting somewhat close to his score.

Interestingly, I could have given the Falcons an extra field goal (because their projected point total was something like 17.8), but the reason I didn't is because I assumed them playing from behind, going for it on a fourth down and failing to convert.

 

[Nikolai]

Love the effort here and looks interesting. I can't imagine the Pats D holding Atlanta to only 17 points. I know your numbers don't lie but there must be some sort of factor that could not be accounted for?

Trust me. It was a bit of a shock to me as well. I'd expect a Falcon to see that and laugh at the idea that anyone could hold them to 17...but we've seen that before

I think the main reason it comes out so low is that the Pats are more or less decent at stopping both the run and the pass, while being able to run the ball themselves. So the projection is that the Falcons will only run 53 plays. On top of that, the Pats have shown some ability to slow down opposing teams' running backs. The one glaring exception is the Seattle game, and if the Pats can't stop Freeman and Coleman, it could be a similar outcome.

The difference is that the Falcons are not that great on defense, especially against the run. They've relied on creating turnovers to balance it out, much like some of our weaker defenses over the years. At least in this case, both teams take care of the ball quite well. Ryan's projected INT comes out of the expectation when he plays pass defenses similar (in yards per attempt) to ours; in those six games, he threw four of his seven total INTs.

But yes, hard to believe. I even went through an exercise where I massaged the numbers a bit to make it look better for the Falcons. Is it possible that they play out of their minds? Sure. Seems to me, one of the biggest factors is the Pats' tackling ability against the Falcons YAC. This is particularly true when we're talking about the running backs. If the Pats can limit their YAC, and even do half was well against their receivers, it's going to be a frustrating day for Atlanta. Conversely, the Pats make hay on YAC, and the Falcons are not so good at limiting those yards.

 

[oldrover]

Disclaimer: The game is played on the field. This is literally the outcome of ****ing with numbers, so take it for what it's worth.

The "TL;DR version" first (a more "detailed" explanation and a summary of the methodology below):

Ever since the Pats and Falcons punched their ticket to the Super Bowl, I've been making a mantra of my prediction; 31-23. I freely admitted that it was just a hunch. But I wanted to dig into the numbers, just to see if they undercut the media narratives following the Championship Games. The results were surprising to me.

The general idea was that I looked at the performances of each skill player on offense and how their performances varied according to the skill level of the defenses they faced. Since players on both teams demonstrated reliable adjustments according the defenses they faced, this was a somewhat easier process.

The final output is how player averages would look against a defenses of the caliber each will face in the Super Bowl (i.e. how Ryan, Jones, and company fared against comparable defenses to what the Pats have). The following are the projected stats:

Brady
26/38, 67.8%, 326 yards, 3TD, 0INT

Blount
17 carries, 153 yards, 1TD

Edelman
9rec, 109 yards, 1TD

------------------------

Ryan
21/33, 64.4%, 288 yards, 2TD, 1INT

Coleman
9 carries, 48 yards

Jones
8rec, 129 yards, 1TD

------------------------

PATRIOTS 37 | 7 10 13 7
FALCONS 17 | 3 7 7 0

------------------------

Basic Methodology

I'm not going to get into the weeds a whole lot, but the basic idea was to try and gauge how each team would do in the Super Bowl by breaking down their past performances at the lowest level possible. So instead of focusing on yards per game, or points per game, I started with the premise of yards per play per player. In the passing game, this meant yards per attempt, in the rushing game, this meant yards per carry, and in the receiving game, this meant yards per target (not reception).

After looking at a game-by-game breakdown of how each team performed, I analyzed the defense they faced while compiling those yards per play. To make a direct comparison, I looked at the yards per play for each defense they faced. I then calculated the deviation from the defense numbers; whether each player performed better or worse than what the defense usually allows.

I noticed both Brady and Ryan performed markedly better against teams they faced a second time. Because this is the first time these two teams are playing, the second game was removed from the analytic process. I limited this change to the quarterbacks only, based on the assumption that a critical factor for the better performances was a better pre-snap read of the opposing defense. Interestingly, this helped the overall stats of each quarterback, because while they improved game-over-game, those games were still below average performances.

From there, I needed to find the passing and rushing defenses that most approximated what each team will face on Sunday. I wanted 5-7 games for each player if at all possible. In some cases, such as Dion Lewis, the sample size is smaller, for which I allowed greater deviation during my adjustments. After isolating these games, I compared each player's yards per play against those defenses, took the difference (positive or negative) and then added those to what the opposing defense usually allows. I then multiplied that by the number of touches to get the stat lines for each skill player I evaluated.

Of course, both teams have role players that may take yards, catches, etc from the primary players I analyzed. During my analysis, it seemed much more likely that this would be the case with the Patriots, despite the fact that I included four WRs plus a TE in the analysis, compared to three WRs and a TE for the Falcons.

I started to whittle down these numbers further according to each team's YAC stats. However, it became quickly apparent that the yards per play on both offense and defense appeared to reflect each teams strengths and weaknesses when it comes to YAC.

After getting each player's individual totals, I reconciled the tallies of the RBs, WRs, and TEs with the QB stat line. After including the carries, I totaled the yards for each team. By this point, I already know how many TDs to expect, but I needed to compare the total yards to the yards per point allowed by each defense to account for how many field goals we might expect on top of those touchdowns.

A total stat breakdown:

Brady: 26/38, 67.8%, 326 yards, 3TD, 0INT

Blount: 17 carries, 153 yards, 1TD
Lewis: 12 carries, 53 yards

Edelman: 9rec, 109 yards, 1TD
Hogan: 4rec, 84 yards
Bennett: 4rec, 41 yards, 1TD
White: 3rec, 31 yards
Mitchell: 3rec, 21 yards
Lewis: 2rec, 23 yards
Amendola: 1rec, 17 yards

TOTAL OFFENSE: 532 yards

------------------------

Ryan: 21/33, 64.4%, 288 yards, 2TD, 1INT

Coleman: 9 carries, 48 yards
Freeman: 12 carries, 38 yards

Jones: 8rec, 129 yards, 1TD
Gabriel: 3rec, 53 yards
Sanu: 4rec, 36 yards, 1 TD
Freeman: 2rec, 34 yards
Coleman: 2rec, 11 yards
Toilolo: 2rec, 25 yards


TOTAL OFFENSE: 374 yards

"You had me at 37-17."

 

[Nikolai]

Excellent post sir. Digging in weeds pays off sometimes. also nice to see a fellow numbers cruncher. I honestly would like to see your raw data, can pm if you like and if you have it.

Just saw this. I have an excel file, but it is the closest approximation you can get to chicken scratch on a computer screen.

 

[BelichickFan]

1. The Patriots have not faced a defense that gives up as many yards per pass attempt as the Falcons. It's not even close. Cleveland was the worst team the Pats faced at 7.24 yards per pass attempt. The Falcons came in at 7.73 yards per pass attempt. Thus, the numbers for the Patriots offense are actually somewhat lower than they would be projected because they are based off an average of around 7 yards per pass attempt.

Haven't made it through all the way yet but is this correct ?

ESPN has the Falcons 11th at 6.9 YPA on defense.

2016 NFL Team Passing Stats - National Football League - ESPN

 

[Ring 6]

Haven't made it through all the way yet but is this correct ?

ESPN has the Falcons 11th at 6.9 YPA on defense.

2016 NFL Team Passing Stats - National Football League - ESPN

That would throw a monkey wrench in to it.

 

[Nikolai]

Haven't made it through all the way yet but is this correct ?

ESPN has the Falcons 11th at 6.9 YPA on defense.

2016 NFL Team Passing Stats - National Football League - ESPN

In raw numbers, Atlanta is actually 6.5 (I don't know where BSPN is getting 6.9; 4267/655=6.51), but I had adjusted the YPA numbers in the final analysis to reflect performances against similarly ranked offenses. Atlanta takes advantage of crap offenses playing desperate from behind.

EDIT - Probably, I may have to get even more into the weeds on the methodology, but...ehhhhh

 

[BelichickFan]

In raw numbers, Atlanta is actually 6.5 (I don't know where BSPN is getting 6.9; 4267/655=6.51)

That is weird. They have the Patriots at 6.8 on defense but if you use their Att/Yds numbers is comes out to 6.39.

 

[Nikolai]

That is weird. They have the Patriots at 6.8 on defense but if you use their Att/Yds numbers is comes out to 6.39.

Yeah, I honestly don't know what they're basing their numbers on. I took raw numbers from Pro Football Reference and worked from there. What's interesting in looking at BSPN's numbers is that, at first glance, there is a fairly uniform 0.4 yards added to each team's total.

 

[IllegalContact]

don't be surprised to see Floyd active for this game.....between him and Hogan, the Pats can field 2 WR's as big as half of the Falcon defenders. A couple of willing blockers like that at WR could repeatedly have much smaller guys isolated in the open field against Blount and seeing as they're the worst tackling team in the league, would appear to be something that is not solveable by game planning......which in turn will produce huge play action ooportunities

the Falcons will overpursue big time

 

[Ring 6]

That is weird. They have the Patriots at 6.8 on defense but if you use their Att/Yds numbers is comes out to 6.39.

It is probably NET yards per attempt meaning they deduct sacks.

 

[Ring 6]

In raw numbers, Atlanta is actually 6.5 (I don't know where BSPN is getting 6.9; 4267/655=6.51), but I had adjusted the YPA numbers in the final analysis to reflect performances against similarly ranked offenses. Atlanta takes advantage of crap offenses playing desperate from behind.

EDIT - Probably, I may have to get even more into the weeds on the methodology, but...ehhhhh

How do you determine similar offenses and defenses if you only include games you consider similar?

 

[luuked]

don't be surprised to see Floyd active for this game.....between him and Hogan, the Pats can field 2 WR's as big as half of the Falcon defenders. A couple of willing blockers like that at WR could repeatedly have much smaller guys isolated in the open field against Blount and seeing as they're the worst tackling team in the league, would appear to be something that is not solveable by game planning......which in turn will produce huge play action ooportunities

the Falcons will overpursue big time

I was thinking the same thing but just can't see who could be inactive for Floyd. Difficult decision..