## Week 1: NFL Game Scripts and Introducing Average Field Position

Wilson pulled out a close victory against Carolina.

Regular readers know all about Game Scripts, the term I’ve used to represent the average margin of lead or deficit over the course of every second of a game. Let’s use the Seahawks-Panthers game to explain how to calculate the Game Script score.

Steven Hauschka’s field goal with 9:40 left in the second quarter was the first score of the game; that means for the first 20 minutes and 20 seconds, the score was tied. Cam Newton responded with a touchdown drive, hitting Steve Smith for a three-yard score with 3:20 left in the half. So for six minutes and twenty seconds, the Panthers trailed by three. It wasn’t until 2:26 left in the third quarter that the next score occurred, courtesy of Hauschka’s second field goal of the day. This means the Panthers led by four for fifteen minutes and fifty-four seconds. Russell Wilson threw the game-winning touchdown with 10:21 remaining, the final score of the day. This means for 7:05, the Seahawks trailed by a point, and then for 10:21, Seattle led by five points (following an unsuccessful two-point conversion attempt).

As it turns out, that gives us a Game Script of exactly 0.00. In other words, on average, this game was tied. Here’s how to do the math:

TmMarginDurationMargin*Duration
Seattle020.330
Seattle36.3319
Seattle-415.90-63.6
Seattle-17.08-7.1
Seattle510.3551.8
Average60.000

By comparison, the Jacksonville-Kansas City game was much more one-sided:

• With 12:32 left in the first quarter, J.T. Thomas blocked Dustin Colquitt’s punt, which resulted in a Jaguars safety.
• That lead lasted all of three minutes and twenty-three seconds, which is how long it took for Alex Smith to find Donnie Avery for a five-yard score with 9:09 left in the first.
• Next, Smith connected with Junior Hemingway for a three-yard touchdown with 1:40 left in the first quarter.
• With 6:29 left in the second quarter, Jamaal Charles punched it in for a short touchdown, bringing the score to 21-2.
• The final score of the day was a Tamba Hali pick six of Blaine Gabbert, with 12:51 left in the fourth quarter.

Here’s how that game looked, which resulted in a Game Script of 15.6:

TmMarginDurationMargin*Duration
Kansas City02.470
Kansas City-23.38-6.8
Kansas City57.4837.4
Kansas City1210.18122.2
Kansas City1923.63449
Kansas City2612.85334.1
Average6015.6

I’ve whipped up a program to calculate the Game Scripts of every game in week one.1 To no one’s surprise, Houston’s huge comeback in the late game Monday Night gave them a win, and the Texans had the lowest game script of any winning team in week one.

Winner LoserBosxcorePFPAMarginGame Script
Kansas City Chiefs@Jacksonville JaguarsBoxscore2822615.6
Denver BroncosBaltimore RavensBoxscore4927226.3
Dallas CowboysNew York GiantsBoxscore363156
Miami Dolphins@Cleveland BrownsBoxscore2310133.6
Indianapolis ColtsOakland RaidersBoxscore211743.4
Tennessee Titans@Pittsburgh SteelersBoxscore16973
San Francisco 49ersGreen Bay PackersBoxscore342862.4
New England Patriots@Buffalo BillsBoxscore232122.3
Seattle Seahawks@Carolina PanthersBoxscore12750
Detroit LionsMinnesota VikingsBoxscore3424100
New Orleans SaintsAtlanta FalconsBoxscore23176-0.8
Chicago BearsCincinnati BengalsBoxscore24213-1.1
New York JetsTampa Bay BuccaneersBoxscore18171-2
St. Louis RamsArizona CardinalsBoxscore27243-2.6
Houston Texans@San Diego ChargersBoxscore31283-7.7

This may not have been the worst day ever for Philip Rivers — after all, he threw four touchdown passes — but it’s another disappointing loss for a Chargers team that is known for disappointing losses. In 2012, San Diego had the 11th best Game Script average and had the lead during 53% of all game seconds (the 7th best mark), but finished 7-9 after losing fourth quarter leads in four games. Only once did a team in 2012 post a Game Script of seven points or higher and lose last year: the Chargers against Mike McCoy’s Broncos.

I want to introduce one other stat today, and solicit input from the crowd. We all have heard the stat on where each team’s average drive started, but I went a step further and calculated the average yardline of each team in each play in week one. In other words, how far from their own endzone was each team on its average snap in week one? A higher number means a team operated more frequently in enemy territory — having the ball on your opponent’s one-yard line gets recorded as a “99” while having the ball at your own one is simply a “1.” The numbers below have some errors, like all play-by-play data, but I don’t think that will be too significant.2

Team# playsAvg Yardline
OAK6255.4
TEN6354.5
DET7553.5
PHI7553.3
SFO7553
NWE8851.7
HOU7351.5
ATL5450.8
NYG5948.9
CHI6048.7
DAL7346.5
CLE7146
STL6245.9
KAN6245.8
NOR6645.2
SDG5144.8
DEN6744
SEA5943.5
GNB5743.2
WAS6842.6
MIA6542.4
PIT5342.3
CAR5042.1
NYJ7341.6
IND5340.7
ARI6940.3
CIN5540.2
BAL8638.6
TAM5836.6
BUF6034.7
MIN5232.9
JAX7031.1

This means that on average, the Raiders average snap came inside the Colts 45-yard line, while the average Jacksonville offensive snap was at their own 31. I have my thoughts on how to use this information, but I don’t want to prejudice the reader. To entice you to provide your thoughts in the comments, here are the results from the 2012 season:

Team# PlaysAvg Yardline
NYG96551.4
NWE118850.9
SFO96049.6
DEN108849.5
ATL101849.4
CIN101348.6
GNB103848.5
MIN99848.4
SEA96847.8
PHI106947.6
DET115647.4
WAS99047.2
HOU108747.1
TAM100447
NYJ102746.8
PIT102046.6
IND110746.6
CAR98546.4
SDG98446.3
CHI99746
DAL104545.9
BAL103545.8
JAX98945.3
NOR106545.1
MIA97745
OAK102344.9
KAN101144.9
BUF98044.9
ARI100944.8
CLE99544.6
STL99544
TEN95342.7
1. Here’s some fine print. I would like to run Game Script numbers pretty frequently during the year. Normally, it takes several hours to calculate Game Scripts at the end of the season, and it takes no less time to calculate them for one week than for seventeen. However, I found a way to cut a couple of corners and get Game Scripts calculated pretty quickly. The downside is it’s possible the Game Scripts scores I present will be inaccurate by a couple of tenths of a point from time to time. I don’t consider that a big deal, but I wanted to alert you to that possibility. []
2. For example, I’ve got the Eagles at 75 plays, not the 77 they officially ran. My play-by-play missed the final kneel down at the end of the game and the following third down attempt in the third quarter: Michael Vick pass incomplete deep left intended for Riley Cooper. Penalty on Brent Celek: Unsportsmanlike Conduct, 15 yards, Penalty on Brent Celek: Offensive Holding (Declined). In the interest of full disclosure, I always want to let you guys know when I’m working with non-exact/nonofficial data. But if I want to produce current data with a short turnaround, I’m going to have to cut some corners. This is my way of preempting anyone in the comments say “what are you talking about!! The Eagles ran 77 plays, not 75! This post is a huge waste of my time!” Which is ironic, since I now know that is a lock to happen. []
• chazmcd

good stuff; like all your work.
my interest would be any predictive value, either from game to game or season to season. that would be helpful for fantasy and ‘entertainment’ purposes. at a quick glance, both tables seem to have a higher concentration of winners at the top/losers at the bottom. is it just a correlation or is there any causal relationship? and which way does the causality flow? do teams win bc of field position, or is field position a result of winning?

• Chase Stuart

Thanks chazmcd.

Yes, I plan to dig into this stuff in a future post.

• Brodie

Very interesting:

I’d be interested to see the correlation between Game Script vs. W/L% at the end of the year. Same story with AFP vs. W/L%. I’d wanna see the marginal value of starting X yds closer to the endzone and how that affects your chance of winning.

Keep up the good work.

• Chase Stuart

Thanks for stopping by, Brodie. You can check out some of my old posts on Game Scripts to see the correlation, which is very high (unsurprisingly; I wouldn’t use a high correlation as a reason to use Game Scripts, which is simply a descriptive stat).

• Richie

I’ll have to think about this to decide if the average field position has any value. I’m thinking of an extreme example. After 30 seconds into the game, Minnesota had an average field position of 22, yet held a 7-0 lead.

Seeing Oakland with the top number, makes me think that maybe Indy was letting Oakland move on their own side of the field, but when they crossed the 50 the defense was playing tougher and forcing Oakland to take a lot more snaps on the Indy side of the field.

• Chase Stuart

The Colts defense was pretty bad, I think, against the Raiders. They forced just two punts, and allowed five 10+ play drives out of just eight drives (and one of those other three was a two-minute drill at the end of the first half).

• Dave

Obviously, average field position included a lot of luck elements like fumbles, interceptions,fgs,but is also the net effect of the offense/defense/special teams

It would be fun to see who performed better or worse than you might expect given their average FP.

It might also be another way to do a power ranking if team adjustments are included.Though it would include a fair amount of non-predictive luck data.

• JeremyDe

Small edit. Think the ‘3:20 left in the game’ in paragraph 2, sentence 2 should read ‘3:20 left in the half (or 2nd quarter if you prefer).

I am half-awake and had to read that sentence 4 times before I got it.

• Chase Stuart

Fixed!

• Ty

ANS Just posted an article about how starting field position has a significant impact on winning and losing. From the list posed about average offensive snap, it is clear that the closer you are to the opponents end zone, the more you are likely to score and win, and the further you are from the opponent’s end zone, the harder it is to score, and you are more likely to lose. This information wouldn’t be reliable, though if it could be predicted from season to season, but I have a belief that teams with good offenses and good defenses usually have better field position than teams with worse offenses and defenses.

• Ty

couldn’t* be predicted.

I probably shouldn’t have said reliable, although I can’t find the word I am looking for.

• Richie

What is ANS?

• Richie

I calculate a .65 correlation between AvgYrdLine and Wins in 2012. The correlation is .73 between AvgYrdLine and Pythagorean Wins in 2012. So, pretty strong relationship.

• Red

The correlation between field position and wins feels rather circular to me. Offense, defense, and special teams cause field position, not the other way around. For example, the 49ers have had the best field position differential the last two years. But that’s not the reason they win. The reason they win is because they have a good offense, defense, and special teams, and their good field position is a byproduct of that. It’s akin to saying that time of possession is important in winning games, when in reality, it’s the traits of winning teams that often result in high TOP.

ANS = Advanced NFL Stats (Brian Burke’s website)

• Josh

This is fantastic stuff. It enriches watching american football so much I think. Great job.

Are you familiar with Rugby Union as a sport. Kicking (ie punting) for territory is staple tactic but it isn’t something that I’ve hear many pundits talk about in the NFL. In rugby union Team A might voluntarily kick away possession in the hope that when Team B kicks in returns the ball won’t go as far as thus netting Team A an overall gain. With that in mind should this colour the 4th down discussion, ie punting on 4th down as a tactic to gain starting field position if you have a good special teams unit and defence that can force 3 and outs?

Once again great website.

• Chase Stuart

Thanks Josh. I’m not familiar with Rugby Union but I appreciate the example!

What you’re describing is the way coaches have thought for ages in the NFL, but is far less applicable now that teams are moving up and down the field with relative ease. In 2013, possession is king.