Vela: A. Jones, CB Rankings and Ken Hamlin’s Best Position: K.C. Joyner Returns, Pt.1

[AbeBeta]

Look around the league. We faced the same competition that every other team faced. In fact, our Defense's DVOA was adjusted downward, which means we faced slightly better than average quarterbacks, on average.

And theogt is correct. Looking at the per-game stats is basically meaningless when the number of attempts aren't that similar.

Interesting, per game stats are meaningless? So you prefer points by play vs. points per game as a metric then?

 

[theogt]

Interesting, per game stats are meaningless? So you prefer points by play vs. points per game as a metric then?

I was (and I believe Adam was) specifically referring to per game passing yards stats. Not all per game stats are meaningless.

 

[AdamJT13]

Interesting, per game stats are meaningless? So you prefer points by play vs. points per game as a metric then?

No, points per play is stupid. Points per possession is more relevant as a metric than points per game. If it's adjusted for field position and opponent, then it's even more relevant.

For yardage, yards per play is more relevant than yards per game as a metric.

If every game included the same number of possessions, the same number of runs, the same number of passes and the same starting field position on each possession, then per-game stats would be more relevant.

 

[AbeBeta]

No, points per play is stupid. Points per possession is more relevant as a metric than points per game. If it's adjusted for field position and opponent, then it's even more relevant.

For yardage, yards per play is more relevant than yards per game as a metric.

If every game included the same number of possessions, the same number of runs, the same number of passes and the same starting field position on each possession, then per-game stats would be more relevant.

How exactly is a possession based metric more important than a game-based metric. Obviously they will be related. But you lose the forest here. Because, as I'm sure you understand, game outcomes are based on final points.

Again, you can point to yards per pass play but that fails to account for situations where those plays occurred -- just as you note the various conditions that allow you to use game stats as a metric it stands to reason that per play has the same faults. For example, what proportion of the plays were in obvious passing situations? That teams attempted 3 fewer runs per game against us last year suggests that teams were in "have to" pass situations more last year. Was that a product of a better run D? Or a stronger offensive effort?

 

[windward]

Uh, the numbers for Davis and Watkins were when Bill was running the show.... He wasn't using numbers about when Wade was running the show...... But Keith Davis sucks... I'm still undecided on Watkins...

I still remember the"Keith Davis is better than Roy" meme that circulated after Roy missed the Carolina game.

Shows the hysteria that infects the fanbase during the season.

 

[AdamJT13]

How exactly is a possession based metric more important than a game-based metric. Obviously they will be related. But you lose the forest here. Because, as I'm sure you understand, game outcomes are based on final points.

If you want to measure offenses' or defenses' abilities to score or stop scoring, a possession-based metric is better because different teams play at different paces. One team might average 12 possessions per game, while another averages nine. The team that gets more possessions might score more per game but less per possession. Say Team A averages 20 offensive points on 12 possessions per game, while Team B averages 18 offensive points on nine possessions per game. Now put them in the same game, give them each the same number of possessions and have the offense perform at exactly their average level. Who wins? Team B, of course. If they each have nine possesions, Team B wins 18-15. If they each have 12, Team B wins 24-20.

Again, you can point to yards per pass play but that fails to account for situations where those plays occurred -- just as you note the various conditions that allow you to use game stats as a metric it stands to reason that per play has the same faults.

If there were any faults, they wouldn't be the same faults.

For example, what proportion of the plays were in obvious passing situations? That teams attempted 3 fewer runs per game against us last year suggests that teams were in "have to" pass situations more last year.

Is there any evidence that teams average fewer yards per pass in "have to" pass situations? It would seem that if they "had to" pass, they'd probably be attempting deeper passes than normal, wouldn't they? Even if they completed a lower percentage, they could end up with a higher YPA because they're throwing deeper. And how many of those situations would come against a soft zone or prevent defense that doesn't really care if it gives up a 15-yard completion with a 20-point lead in the final minute?

Of course, you can always look at pass DVOA, which takes into account the game situation for every play (score, field position, down and distance, opponent, etc.) and compares them to other plays in the same situation. In 2006, our pass DVOA was 8.5 percent -- an average of 8.5 percent worse than the NFL average in the same situations (which ranked 24th). Last season, our pass DVOA was minus-6.0 -- an average of 6 percent better than the NFL average in the same situations, and we ranked ninth.

So, even if our opponents "had to" pass more, and even if that made it easier for our defense (which may or may not be true), our pass defense did improve significantly last season when you compare how we did against how every other defense does in the exact same situations.

 

[Temo]

Just a couple quick notes on the statements that Joyner made about the age at which players achieve peak physical form:

Bill Jame's (and other's) work describes 1) Performance in a sport, not just physical ability and 2) baseball, not football.

This is important because performance in sport usually comes from your physical ability PLUS your knowledge and skill in the game. So while I think most people whould run faster when they're 22-25 as opposed to 28, they'll probably play better baseball because you learn a lot about the game in those years. (Again, on average... humans are not perfectly identical to each other).

Now, in a sport like baseball, deterioration in physical abilities (especially as you grow in knowledge as how to hit pitchers, pitches, take pitches, etc.) is not as important as in a more physical sport like football. In fact, most studies have shown that most football players peak earlier than the 27-28 that Joyner used. Running backs are the worst, peaking at age 25 on average. Quarterbacks are the best, actually beating baseball players in peaking at age 29 (not surprising, considering the mental work it takes to play the position, which can take time to learn).

 

[Temo]

BTW, if 25 seems young to you, consider Emmitt's statistics: http://www.pro-football-reference.com/players/S/SmitEm00.htm

You'll see him hit a peak in yards/attempt at age 24, and drop off a lot after age 26. And that's probably the most durable runner in the history of the NFL.

 

[AbeBeta]

If you want to measure offenses' or defenses' abilities to score or stop scoring, a possession-based metric is better because different teams play at different paces. One team might average 12 possessions per game, while another averages nine. The team that gets more possessions might score more per game but less per possession. Say Team A averages 20 offensive points on 12 possessions per game, while Team B averages 18 offensive points on nine possessions per game. Now put them in the same game, give them each the same number of possessions and have the offense perform at exactly their average level. Who wins? Team B, of course. If they each have nine possesions, Team B wins 18-15. If they each have 12, Team B wins 24-20.

Again, as I noted. By possession tells you one thing. But by game tells you another. The way you use those values is relevant only to comparisons wherein you control for every other variable -- that's a nice statistical value -- but it often has no practical meaning.

Is there any evidence that teams average fewer yards per pass in "have to" pass situations? It would seem that if they "had to" pass, they'd probably be attempting deeper passes than normal, wouldn't they? Even if they completed a lower percentage, they could end up with a higher YPA because they're throwing deeper. And how many of those situations would come against a soft zone or prevent defense that doesn't really care if it gives up a 15-yard completion with a 20-point lead in the final minute?

Of course, you can always look at pass DVOA, which takes into account the game situation for every play (score, field position, down and distance, opponent, etc.) and compares them to other plays in the same situation. In 2006, our pass DVOA was 8.5 percent -- an average of 8.5 percent worse than the NFL average in the same situations (which ranked 24th). Last season, our pass DVOA was minus-6.0 -- an average of 6 percent better than the NFL average in the same situations, and we ranked ninth.

So, even if our opponents "had to" pass more, and even if that made it easier for our defense (which may or may not be true), our pass defense did improve significantly last season when you compare how we did against how every other defense does in the exact same situations.

Can you break those DVOA numbers down more -- in what areas was our pass D worse and how exactly is that calculated. Without some explanation of how those values are calculated (i.e., how each is weighted) it is hard to comment on the value of the statistic. One serious concern I have about this statistic is that when you parse out the situation combinations into such fine grains you are going to be combining apples and oranges -- that is your 3rd and 14 with a 10 point lead and less than a minute left in the half against the Cardinals may be directly comparable to other teams in that same situation. But realistically, how many of those situations are you comparing? How exactly are common vs. uncommon situations weighted? That will speak considerably to how useful the estimate is.

But again, my argument was that much of our improvement on D is at least partially attributable to our improvement on offense. Yes, the D did improve. The cause for that really isn't clear at this point.

 

[AdamJT13]

Again, as I noted. By possession tells you one thing. But by game tells you another. The way you use those values is relevant only to comparisons wherein you control for every other variable -- that's a nice statistical value -- but it often has no practical meaning.

Points per possession has much more practical meaning than points per game. And yards per play (and yards per rush and yards per pass) has much more practical meaning than yards per game.

You can have two defenses allow the exact same number of points and total yards but perform nowhere close to equally well.

Let's say Defense A faces Offense X in Week 1. Offense X gets eight possessions, runs 48 plays, gains 300 yards and scores 17 points.

In Week 2, Defense B faces Offense X. This time, Offense X gets 15 possessions, runs 60 plays, gains 300 yards and scores 17 points.

Assuming the starting field position was equal, Defense B clearly played better. Offense X needed almost twice as many possessions to gain the same number of yards and score the same number of points.

Can you break those DVOA numbers down more -- in what areas was our pass D worse and how exactly is that calculated. Without some explanation of how those values are calculated (i.e., how each is weighted) it is hard to comment on the value of the statistic. One serious concern I have about this statistic is that when you parse out the situation combinations into such fine grains you are going to be combining apples and oranges -- that is your 3rd and 14 with a 10 point lead and less than a minute left in the half against the Cardinals may be directly comparable to other teams in that same situation. But realistically, how many of those situations are you comparing? How exactly are common vs. uncommon situations weighted? That will speak considerably to how useful the estimate is.

You'd have to ask Football Outsiders if you want to get that specific or challenge the validity of their stats.

Once this year's book arrives (the shipping date has moved up to July 16), I could tell you how 2006 and 2007 compared overall on first down, second down, third down, in the red zone and in "late and close" situations, but they don't break it down any further than that in the book.

But again, my argument was that much of our improvement on D is at least partially attributable to our improvement on offense. Yes, the D did improve. The cause for that really isn't clear at this point.

How, exactly, did the offense's improvement help the defense reduce its yards allowed per attempt by 0.9 and its yards allowed per catch by 1.5?

An offense can help its defense's PER-GAME totals by keeping it off the field, but it can't do a whole lot to help the defense stop the opponent when the defense is on the field.

 

[AbeBeta]

Points per possession has much more practical meaning than points per game. And yards per play (and yards per rush and yards per pass) has much more practical meaning than yards per game.

You can have two defenses allow the exact same number of points and total yards but perform nowhere close to equally well.

Let's say Defense A faces Offense X in Week 1. Offense X gets eight possessions, runs 48 plays, gains 300 yards and scores 17 points.

In Week 2, Defense B faces Offense X. This time, Offense X gets 15 possessions, runs 60 plays, gains 300 yards and scores 17 points.

Assuming the starting field position was equal, Defense B clearly played better. Offense X needed almost twice as many possessions to gain the same number of yards and score the same number of points.

I understand the concept of average Adam, what you are missing is that the per possession value only had meaning because it related to the ultimate outcome - points per game. Given your two Ds you could also claim that although D clearly played better using your metric. However, there is only one metric that matters at the end of the day. And those are point. Further you might conclude that Defense B having faced so many more plays created a situation wherein the offense had the ball less.

You'd have to ask Football Outsiders if you want to get that specific or challenge the validity of their stats.

Respectfully, I would warn against placing faith in a value if you don't know its derivation

How, exactly, did the offense's improvement help the defense reduce its yards allowed per attempt by 0.9 and its yards allowed per catch by 1.5?

An offense can help its defense's PER-GAME totals by keeping it off the field, but it can't do a whole lot to help the defense stop the opponent when the defense is on the field.

Again, as I noted, an effect offense can place a team in MORE situations that are favorable to the D. That can be in terms of field position, playing from behind etc.

Take for example the games where Dallas didn't produce on O in 2006. Both Philly games, NO, Det. 136-1282, for a whopping 9.4 YPA. We didn't have those lows in 2007 -- aside from possibly the Philly game where they ran over us and the Washington game where we didn't play starters. But you can often see big advantages for the D in games where the O is dominating and big disadvantages when the O stinks.

 

[percyhoward]

Adam, is there any evidence that teams score more points when they have more possessions?

Obviously, as the game goes on, a team is more likely to have scored more points by the 3rd quarter than it had in the first. But I mean, is there any data that shows that, in a game in which there are more total possessions, there are also more total points (on average)?

 

[AdamJT13]

I understand the concept of average Adam, what you are missing is that the per possession value only had meaning because it related to the ultimate outcome - points per game.

Points per possession has meaning because it more accurately measures the defense's performance than points per game does.

Further you might conclude that Defense B having faced so many more plays created a situation wherein the offense had the ball less.

That's not necessarily true. If the defense faced 15 possessions, the offense also would get roughly 15 possessions, compared to roughly eight for the other offense in the other game. It's up to the offense what it does with those extra seven possessions. It could have run more plays and scored more points than the opponent, while the offense that had eight possessions could have run fewer plays and scored fewer points than its opponent.

And by the way, allowing the opponent to run 4.0 plays per possession isn't exactly "so many" plays, considering that the NFL average is typically about 5.6 plays per possession.

Again, as I noted, an effect offense can place a team in MORE situations that are favorable to the D. That can be in terms of field position, playing from behind etc.

Generally, the more your offense produces, the more the opposing offense produces because it opens up the gameplan, passes more and throws deeper, while your defense softens up a little because it has a big lead.

When your offense isn't producing, the opposing offense becomes more conservative, runs more, throws shorter passes and tries to keep the clock running.

Remember when all of the naysayers said our defense was so good in 2003 only because our offense was so bad and our opponents didn't have to do much? Now you're saying our defense was so good in 2007 only because our offense was so good that it made our defense good?

At some point, you have to realize that the defense is pretty much on its own when it's on the field.

Take for example the games where Dallas didn't produce on O in 2006. Both Philly games, NO, Det. 136-1282, for a whopping 9.4 YPA.

Why would you single out those games for the offense "not producing"? The offense produced more in several of those games than it did in some other games.

And again, you'll have top explain how the offense was responsible for the defense allowing long passes in those games (or whatever other games you pick).

We didn't have those lows in 2007 -- aside from possibly the Philly game where they ran over us and the Washington game where we didn't play starters. But you can often see big advantages for the D in games where the O is dominating and big disadvantages when the O stinks.

Not particularly. I can see sack percentages and maybe interception percentages going up (the defense can focus on rushing the passer, and the offense takes more risks), but I can't see how it helps your defense limit the opponent to a lower YPA or yards per catch. You'd expect both of those stats to be higher when the opponent is forced to play catch-up.

 

[theogt]

I understand the concept of average Adam, what you are missing is that the per possession value only had meaning because it related to the ultimate outcome - points per game.

This last part isnt true at all. PPG may be a better indicator of defense NET of the offense's effect, but when youre comparing defense to defense you should adjust for the effect of the offense. One easy way to do this is look at points per possession.

 

[percyhoward]

Where is the evidence that suggests that a team that scores 14 points in 10 possessions will score 21 points in 15 possessions?

 

[Temo]

Where is the evidence that suggests that a team that scores 14 points in 10 possessions will score 21 points in 15 possessions?

To take your statement to another level, where is the evidence that a batter who gets 30 hits in 100 ABs will get 60 in 200 ABs? There is no definite evidence. But should we toss batting average out the window and instead judge a batter on hits per game? So we say that a guy who goes 1 for 3 had a worse day than a guy who went 2 for 7 in extra innings? Should we toss out completion percent as a statistic and just use completion totals from now on?

Rate statistics are a measure of effectiveness given a base amount of units... they enable you capture performance by excluding the effect of a large number of attempts.

 

[AbeBeta]

Points per possession has meaning because it more accurately measures the defense's performance than points per game does.

And points per game more accurately reflects who actually wins the game. So you can focus on whatever you want

That's not necessarily true. If the defense faced 15 possessions, the offense also would get roughly 15 possessions, compared to roughly eight for the other offense in the other game. It's up to the offense what it does with those extra seven possessions. It could have run more plays and scored more points than the opponent, while the offense that had eight possessions could have run fewer plays and scored fewer points than its opponent.

Not necessarily true? It also isn't necessarily false.

And by the way, allowing the opponent to run 4.0 plays per possession isn't exactly "so many" plays, considering that the NFL average is typically about 5.6 plays per possession.

It was a comparative statement
[/QUOTE]

Generally, the more your offense produces, the more the opposing offense produces because it opens up the gameplan, passes more and throws deeper, while your defense softens up a little because it has a big lead.

When your offense isn't producing, the opposing offense becomes more conservative, runs more, throws shorter passes and tries to keep the clock running.

Remember when all of the naysayers said our defense was so good in 2003 only because our offense was so bad and our opponents didn't have to do much? Now you're saying our defense was so good in 2007 only because our offense was so good that it made our defense good?

When did I say that in 2003?

Take your DVO stats from 2007 -- the correlation between offense and defensive numbers is -.26 (higher offense means better lower defense means better). Using that metric there is a relationship. Not that I am endorsing the metric

At some point, you have to realize that the defense is pretty much on its own when it's on the field.

7% of the variability in defense can be predicted by offense. In a game where it is hard to predict variability in performance, that is a pretty substantial number. We don't know if those #s are chickens or eggs but we do know that they are related.

Why would you single out those games for the offense "not producing"? The offense produced more in several of those games than it did in some other games.

I just picked out the handful of games where i remembered the worst performances. Just an example.

And again, you'll have top explain how the offense was responsible for the defense allowing long passes in those games (or whatever other games you pick).

For one, when the other team's O is not producing you can run your O as intended - that gives you more options and allows the opponent to dictate the plays as opposed to situations wherein a good offensive performance limits the types of plays that the opponent is going to run (i.e., puts them in a mostly passing mode).

Not particularly. I can see sack percentages and maybe interception percentages going up (the defense can focus on rushing the passer, and the offense takes more risks), but I can't see how it helps your defense limit the opponent to a lower YPA or yards per catch. You'd expect both of those stats to be higher when the opponent is forced to play catch-up.

If sacks go up it stands to reason that general pressure on the QB goes up -- that can lead to more throw aways, rushed throws that drop incomplete, and even a reliance on short quick passing. There's one explanation for you

 

[AbeBeta]

This last part isnt true at all. PPG may be a better indicator of defense NET of the offense's effect, but when youre comparing defense to defense you should adjust for the effect of the offense. One easy way to do this is look at points per possession.

so you propose a simple univariate comparison for what is admittedly a complex, multifaceted relationship?

 

[percyhoward]

To take your statement to another level, where is the evidence that a batter who gets 30 hits in 100 ABs will get 60 in 200 ABs?

On average, that's exactly what happens, though.

Obviously, there's no guarantee that it will happen, but we're not talking about certainties, anyway. It's about the assumption that, in games with more possessions, more points will be scored--ON AVERAGE.

If there's no evidence that scoring goes up in proportion to possessions, that it doesn't make a whole lot of sense to factor number of possessions into a scoring stat.

 

[Temo]

On average, that's exactly what happens, though.

Obviously, there's no guarantee that it will happen, but we're not talking about certainties, anyway. It's about the assumption that, in games with more possessions, more points will be scored--ON AVERAGE.

If there's no evidence that scoring goes up in proportion to possessions, that it doesn't make a whole lot of sense to factor number of possessions into a scoring stat.

Intuitively speaking, why shouldn't an increase in the number of attempts to score lead to more scoring? That's what most people think when they say "So and so has a .300 batting average, so given 100 more ABs, it'd be safe to say he accumulates 30 more hits".