Revisiting Bend Don't Break and Time of Possession

[BradyManny]

So don't use Time of Possesion. Use number of plays. If the Giants ran more plays than the Patriots did, that is a better predictor of success than how much time each ran off the clock.

I couldn't find number of play stats for the season, or I may have used that instead. But the purpose of taking time of possession and framing it on a per drive basis was to give a more accurate picture of which unit was on the field for longer, and what was accomplished in that period of time, and what the relationship between the two is.

I agree TOP, in a vacuum, has no real value.

 

[unoriginal]

Since I've been pitching number of plays as the superior stat, I'll provide them, quick and dirty style:

Plays on Offense, 2011
Patriots        1,082
Giants          1,028
Median          1,015   
Pats avg.       68
Giants avg.     64
Median avg.     63

Plays on Defense, 2011
Patriots        1,064
Giants          1,072
Median          1,010   
Pats avg.       67
Giants avg.     67
Median avg.     63

Super Bowl XLVI   
Patriots        62
Giants          71

In the regular season, both the Giants and the Patriots had more plays their the median teams did. The Patriots had ~9 more plays per game than the median, and the Giants ~5 more plays. The difference was entirely on offense. The Giants were actually on defense more than they were on offense during the regular season.

The Super Bowl, with 133 scrimmage plays, was about as fast as each team normally plays (Giants, 131; Patriots, 135), and over league median (126 plays).

So the Giants, despite dominating Time of Possession 37:05 to 22:55, did not actually slow the game down.

 

[Deus Irae]

Since I've been pitching number of plays as the superior stat, I'll provide them, quick and dirty style:

Plays on Offense, 2011
Patriots        1,082
Giants          1,028
Median          1,015   
Pats avg.       68
Giants avg.     64
Median avg.     63

Plays on Defense, 2011
Patriots        1,064
Giants          1,072
Median          1,010   
Pats avg.       67
Giants avg.     67
Median avg.     63

Super Bowl XLVI   
Patriots        62
Giants          71

In the regular season, both the Giants and the Patriots had more plays their the median teams did. The Patriots had ~9 more plays per game than the median, and the Giants ~5 more plays. The difference was entirely on offense. The Giants were actually on defense more than they were on offense during the regular season.

The Super Bowl, with 133 scrimmage plays, was about as fast as each team normally plays (Giants, 131; Patriots, 135), and over league median (126 plays).

So the Giants, despite dominating Time of Possession 37:05 to 22:55, did not actually slow the game down.

Yes, but since that stat is ignoring how the plays were set up, it's no better than TOP. A quick strike is all it takes to make number of plays completely useless in a lot of games.

 

[everlong]

That's true, but still in that situation, the best stat is number of plays, compared both to the other team, and to team and league-wide season averages.

Right. I think we're down to semantics. If the Patriots average X and you can hold them to X-10 then the 35:25 TOP meant something. If you held them to X-1 it was meaningless.

 

[everlong]

Since I've been pitching number of plays as the superior stat, I'll provide them, quick and dirty style:

Plays on Offense, 2011
Patriots        1,082
Giants          1,028
Median          1,015   
Pats avg.       68
Giants avg.     64
Median avg.     63

Plays on Defense, 2011
Patriots        1,064
Giants          1,072
Median          1,010   
Pats avg.       67
Giants avg.     67
Median avg.     63

Super Bowl XLVI   
Patriots        62
Giants          71

In the regular season, both the Giants and the Patriots had more plays their the median teams did. The Patriots had ~9 more plays per game than the median, and the Giants ~5 more plays. The difference was entirely on offense. The Giants were actually on defense more than they were on offense during the regular season.

The Super Bowl, with 133 scrimmage plays, was about as fast as each team normally plays (Giants, 131; Patriots, 135), and over league median (126 plays).

So the Giants, despite dominating Time of Possession 37:05 to 22:55, did not actually slow the game down.

Nice breakdown.

Yes, but since that stat is ignoring how the plays were set up, it's no better than TOP. A quick strike is all it takes to make number of plays completely useless in a lot of games.

That's true and in the case of the SB the two turnovers, including the safety, certainly played a part in the Giants having a +9 play advantage and the Pats being -6 to their season average.

 

[BradyManny]

Since I've been pitching number of plays as the superior stat, I'll provide them, quick and dirty style:

Plays on Offense, 2011
Patriots        1,082
Giants          1,028
Median          1,015   
Pats avg.       68
Giants avg.     64
Median avg.     63

Plays on Defense, 2011
Patriots        1,064
Giants          1,072
Median          1,010   
Pats avg.       67
Giants avg.     67
Median avg.     63

Super Bowl XLVI   
Patriots        62
Giants          71

In the regular season, both the Giants and the Patriots had more plays their the median teams did. The Patriots had ~9 more plays per game than the median, and the Giants ~5 more plays. The difference was entirely on offense. The Giants were actually on defense more than they were on offense during the regular season.

The Super Bowl, with 133 scrimmage plays, was about as fast as each team normally plays (Giants, 131; Patriots, 135), and over league median (126 plays).

So the Giants, despite dominating Time of Possession 37:05 to 22:55, did not actually slow the game down.

Perhaps I should clarify what I mean by slowing the game down - I'm not talking about pace, I'm referring to the number of possessions, and its resulting effect on the perceived "length" of the game. Just two weeks prior, the NFCCG which even before OT approached twice as many total possessions (excluding kneeldowns) than the Super Bowl. I would call that game "longer."

If the average number of possessions in an NFL game is 22, I call anything under that "shorter" and anything more "longer." Maybe that's poorly worded on my part, but for the sake of this thread and my OP, that's what I'm going with for now.

The number of plays stat is telling to the pace of the game, but I don't think number of plays tells a better story in terms of indicating success. Even putting aside the factor that Deus mentions, there are other issues. Number of plays doesn't take into account field position. Taking SB46, consider the Patriots had a 10 play, 60 yard drive and an 11 play 48 yard drive that combined for only three points. Field position factored into the latter drive, as they started from their 8. They also racked up 14 plays on one drive when they were forced to go 98 yards for a score to end the half.

That's 35 plays in just three drives, and 10 points to show for it. (Side note: we can see how costly settling for a field goal on that first drive was, with possessions at such a premium in this game.)

(See Ivanvamp's posts for a nice breakdown of how field position factored in heavily.)

You can only get seven points on a possession, so the number of possessions in the game is extremely important - it determines the amount of points you can score.

No matter how many plays the Giants ran, they did their damage in 8 possessions, and only gave the Patriots nine cracks at it, one of which they wasted with a safety, and another which was largely pure desperation at the end of the game. The Pats didn't have many shots at it. And both the offense and defense is to blame for it, but it is - to me - the biggest reason they lost.

Curious where you got the Plays stat - that's good stuff. Thanks for bringing that to the table, I was looking for that (couldn't find it on footballoutsiders or the other usual suspects). If you can give me the link, I'd be curious to see all teams and compare that to the number of drives, one could derive pace of play for each team.

 

[ausbacker]

No need to be a snobby little putz about it. I'm sure it will go over most people's heads. That's what happens when you're working with the internet instead of having a conversation in real life. Perhaps next time you might want to include some smilies...

:snob:

 

[unoriginal]

Curious where you got the Plays stat - that's good stuff. Thanks for bringing that to the table, I was looking for that (couldn't find it on footballoutsiders or the other usual suspects). If you can give me the link, I'd be curious to see all teams and compare that to the number of drives, one could derive pace of play for each team.

NFL.com > Scores & Stats > Stats > Team Stats > Select "2011" and "Regular Season" > Select any "Complete List" link > Toggle "Offense" and "Defense" > Sort by "Scrm Plys"

NFL Stats: by Team Category

NFL Stats: by Team Category

 

[hovis]

Perhaps I should clarify what I mean by slowing the game down - I'm not talking about pace, I'm referring to the number of possessions, and its resulting effect on the perceived "length" of the game. Just two weeks prior, the NFCCG which even before OT approached twice as many total possessions (excluding kneeldowns) than the Super Bowl. I would call that game "longer."

If the average number of possessions in an NFL game is 22, I call anything under that "shorter" and anything more "longer." Maybe that's poorly worded on my part, but for the sake of this thread and my OP, that's what I'm going with for now.

The number of plays stat is telling to the pace of the game, but I don't think number of plays tells a better story in terms of indicating success. Even putting aside the factor that Deus mentions, there are other issues. Number of plays doesn't take into account field position. Taking SB46, consider the Patriots had a 10 play, 60 yard drive and an 11 play 48 yard drive that combined for only three points. Field position factored into the latter drive, as they started from their 8. They also racked up 14 plays on one drive when they were forced to go 98 yards for a score to end the half.

That's 35 plays in just three drives, and 10 points to show for it. (Side note: we can see how costly settling for a field goal on that first drive was, with possessions at such a premium in this game.)

(See Ivanvamp's posts for a nice breakdown of how field position factored in heavily.)

You can only get seven points on a possession, so the number of possessions in the game is extremely important - it determines the amount of points you can score.

No matter how many plays the Giants ran, they did their damage in 8 possessions, and only gave the Patriots nine cracks at it, one of which they wasted with a safety, and another which was largely pure desperation at the end of the game. The Pats didn't have many shots at it. And both the offense and defense is to blame for it, but it is - to me - the biggest reason they lost.

Curious where you got the Plays stat - that's good stuff. Thanks for bringing that to the table, I was looking for that (couldn't find it on footballoutsiders or the other usual suspects). If you can give me the link, I'd be curious to see all teams and compare that to the number of drives, one could derive pace of play for each team.

Your posts have me thinking about some possible metrics to measure the effect of a teams time-possession strategy on their ability to win games. Just watching the Pats last year, I'd say one weakness they had at times was putting together long clock-killing drives, and preventing other teams from doing so on defense.

The trouble with all this statistical analysis is that we never know intent, we can only infer it. In my opinion, most drive stats don't do enough to deduce the likely intent of the offense.

What I'd like to see is a drive pace classification metric, coupled along with average yards per drive pace category and average points per drive pace category. First, someone would need to measure the game time between each snap for each drive, and compute an average. Next, you'd need to compare that to every drive for every game in the league, and compute the average play length per drive percentiles. Then you'd categorize the drive by tempo (Top Quartile - Clock Killing, 2nd Quartile - Moderately Slow, 3rd Quartile - Moderately Fast, 4th Quartile - No huddle / 2- Minute warning speed), and use it to compute each team's success rates in those categories.

For instance, if the Pats consistently scored TD's and gained eighty yards per drive on drives with an average per snap length of ten to fifteen seconds, the numbers would show they're a great team when working at a fast pace. Conversely, if they consistently went three and out on drives where the average play length was 35-40 seconds, the numbers would show low average yards and points on slow drives; in other words, you'd have numbers to show they suck at clock killing.

There are a few obvious problems with the proposed metric. Mainly, the influence of defense could make it difficult to infer the intent of the offenses strategy. For instance, if a passing team tries to dink and dunk down the field to kill the clock, but gets all of the passes batted away, then the drive would be improperly categorized as fast tempo. Same thing with interceptions, timeouts, forceouts, and injuries (real and fake). Perhaps counting the play clock length between plays could fix that.

Still, with a large enough sample, I'd think it'd be a useful stat for gauging an offense's ability to control tempo.

 

[unoriginal]

The number of plays stat is telling to the pace of the game, but I don't think number of plays tells a better story in terms of indicating success. Even putting aside the factor that Deus mentions, there are other issues. Number of plays doesn't take into account field position. Taking SB46, consider the Patriots had a 10 play, 60 yard drive and an 11 play 48 yard drive that combined for only three points. Field position factored into the latter drive, as they started from their 8. They also racked up 14 plays on one drive when they were forced to go 98 yards for a score to end the half.

That's 35 plays in just three drives, and 10 points to show for it. (Side note: we can see how costly settling for a field goal on that first drive was, with possessions at such a premium in this game.)

(See Ivanvamp's posts for a nice breakdown of how field position factored in heavily.)

I note nothing you said or referenced in that post about the problems with play count isn't also a problem with ToP.

You (and Deus) are arguing a larger thing than I am. You are trying to put together a complete statistical picture, and Deus is saying that play count is, like time of posession, not a complele statistical picture. I am pointing out that, in constructing that complete picture, you should use plays and possessions over time of possession. I quote my first paragraph in this thread:

Time of Possssion is not a good stat. It is a conveinent stat, I understand, as every box score notes it. But it doesn't measure anything important. Perhaps, over a large enough set of data, it could point to which team is most likely to execute a clock-killing drive in the 4th quarter, which is situationally useful. Everywhere else it is just a lazy equivilent of the things that do matter: number of possessions, and number of plays.

So I think me and you complete agree. I have since concentrated mainly on number of plays because I figured I could get those stats fairly easily. (see post above) I don't know where to get number of possessions easily. Maybe add # punts, # field goal attempts, # touchdowns, and # turnovers together? You'd lose some end-of-half stuff, but if it ended without any of the above it probably was just a kneel-down sequence anyways. Still it would be laborious to compute.

You can only get seven points on a possession, so the number of possessions in the game is extremely important - it determines the amount of points you can score.

No matter how many plays the Giants ran, they did their damage in 8 possessions, and only gave the Patriots nine cracks at it, one of which they wasted with a safety, and another which was largely pure desperation at the end of the game. The Pats didn't have many shots at it. And both the offense and defense is to blame for it, but it is - to me - the biggest reason they lost.

The biggest reason to me: Giants scored points on defense, and had a +2 turnover differential, safety and interception.

 

[BradyManny]

NFL.com > Scores & Stats > Stats > Team Stats > Select "2011" and "Regular Season" > Select any "Complete List" link > Toggle "Offense" and "Defense" > Sort by "Scrm Plys"

NFL Stats: by Team Category

NFL Stats: by Team Category

Nice, thanks for posting.

 

[BradyManny]

I note nothing you said or referenced in that post about the problems with play count isn't also a problem with ToP.

You (and Deus) are arguing a larger thing than I am. You are trying to put together a complete statistical picture, and Deus is saying that play count is, like time of posession, not a complele statistical picture. I am pointing out that, in constructing that complete picture, you should use plays and possessions over time of possession.

Fair enough, understood there.

So I think me and you complete agree. I have since concentrated mainly on number of plays because I figured I could get those stats fairly easily. (see post above) I don't know where to get number of possessions easily.

FootballOutsiders keeps track of possession counts, excluding kneel downs - which I used in the formula for some of these other stats. I couldn't find the play stats anywhere, now that you posted those, I might be able to combine them - then I could clearly determine time per play, which would tells the story on the team's pace on offense, which helps tell the story re: TOP, # of possessions, it's all linked together.

The biggest reason to me: Giants scored points on defense, and had a +2 turnover differential, safety and interception.

No arguments here - the safety was truly a killer. Especially given the Patriots success from poor field position later in the game, it's hard to measure what they truly gave up there - but it's bad. (And I still question the call. Or maybe, rather, the rule in that specific instance...but I don't want to sound like a Raiders fan, so I'll stop there...)

But here's the thing, and why I always take it back to possession total: In SB38, Brady threw a pick in the end zone when they could've basically gone for the jugular. Situationally, I think that is unquestionably the worst mistake he's made in his playoff career. Yet, that game is largely considered his best SB performance, and earned him his second Super Bowl MVP. And no one is going to argue that he deserved it, he did, because at the time, he put up some of the most productive numbers the SB had seen. But that Super Bowl had 27 possessions. It gave the offense enough time to make up for these mistakes.

If any of the two balls the Giants put on the ground bounce towards a Patriot, the safety, the Welker drop, the Blackburn INT never even enter our mind, like the EndZone safety in SB38. Mistakes happen in football games. No one plays a perfect game. The Giants had some luck/randomness on their side. The "longer" the game, the more possessions, the less luck will factor in.

 

[BradyManny]

What I'd like to see is a drive pace classification metric, coupled along with average yards per drive pace category and average points per drive pace category. First, someone would need to measure the game time between each snap for each drive, and compute an average. Next, you'd need to compare that to every drive for every game in the league, and compute the average play length per drive percentiles. Then you'd categorize the drive by tempo (Top Quartile - Clock Killing, 2nd Quartile - Moderately Slow, 3rd Quartile - Moderately Fast, 4th Quartile - No huddle / 2- Minute warning speed), and use it to compute each team's success rates in those categories.

For instance, if the Pats consistently scored TD's and gained eighty yards per drive on drives with an average per snap length of ten to fifteen seconds, the numbers would show they're a great team when working at a fast pace. Conversely, if they consistently went three and out on drives where the average play length was 35-40 seconds, the numbers would show low average yards and points on slow drives; in other words, you'd have numbers to show they suck at clock killing.

There are a few obvious problems with the proposed metric. Mainly, the influence of defense could make it difficult to infer the intent of the offenses strategy. For instance, if a passing team tries to dink and dunk down the field to kill the clock, but gets all of the passes batted away, then the drive would be improperly categorized as fast tempo. Same thing with interceptions, timeouts, forceouts, and injuries (real and fake). Perhaps counting the play clock length between plays could fix that.

Still, with a large enough sample, I'd think it'd be a useful stat for gauging an offense's ability to control tempo.

Interesting take. You propose an interesting metric, though I think you point out some of the issues in arriving at it. I really like the idea of a drive pace metric. Very interesting. All of these things probably would need a lot of work by hand.

In terms of controlling tempo and the Pats offense, one thing to point out - ultimately, as you allude to, we know its all situational. And here's where we see the downfall of looking at raw numbers: statistically speaking, looking at it in a vacuum - I'm assuming the Patriots offense is better off scoring faster, which they do. Again, if my whole premise is more possessions is better for them, the faster they score, the better. However, in reality, that doesn't bear out when we consider game situations. There's no real incentive to score faster most of the time, and if anything, as you point out, there is often real incentive to score while chewing up clock.