Top Shutdown Cornerbacks of 2007 (putting K.C. Joyner's stats to better use)

[AdamJT13]

But on average, they're supposed to.

That's what makes it average.

If a team's "average level" is 2 points per possession, then it should score 20 points on 10 possessions, and 30 points on 15. Since that doesn't happen on average, (not even close) then that figure of 2 "points per possession" doesn't really tell us anything.

Again, you can't expect teams to perform at their average all of the time in a sample of only 16 games. It's like looking at any game from Tony Romo's 2007 season and expecting him to complete exactly 64.4 percent of his attempts for 12.6 yards per completion, with 6.9 percent touchdowns and 3.7 percent interceptions.

If he completes 64.4 percent of his passes, he should have more completions when he attempts more passes, right? Of course. But that doesn't mean he won't have a game like his 13-for-36 performance against the Eagles -- and 10 different games with more completions on fewer than 36 attempts. He had one game of 21-for-28 and another of 14-for-29. Do those discrepancies mean we should just throw out his 64.4 percent completion average, or the notion that more attempts should mean more completions? Or that more completions generally means more yards? Or that more completions generally means more touchdowns?

In those 4 scenarios, the offense/defense that performed better is the one that points per possession says performed better.

And that's exactly why points per possession is a better measure than points per game. The offense that scored fewer points and the defense that allowed more points obviously performed better. The only reason they scored fewer points or allowed more points is because of the number of possessions thay had or faced.

First, whether I think 1.85 is a big difference doesn't matter. What does matter is that "points per possession" says that, for a team that averages 2.00 per possession, it shouldn't go up 1.85, it should go up 6.00. That's off by more than just a little.

Those are averages for 10 teams with different YPP, not the numbers for one team.

And you still haven't explained why the average "should" remain the same no matter how many possessions. (Never mind the possible reasons for facing more possessions, which I've explained several times and which obviously would affect the average.) The point is that facing more possessions WILL result in more points allowed, on average. So, facing 50 more possessions than another defense OBVIOUSLY is a disadvantage, if you're being judged on per-game stats.

Second, Football Outsiders says the average number of possessions in a game is a little over 11. You provide data from 8- to 11-possession games. You're looking at the low end, which skews the results toward what you said. You know this, which is probably why you don't say what your research showed you about 12- and 13-possession games.

Like I already explained, I didn't count all of the "throwaway" possessions when the offense was kneeling on the ball, running out the clock or didn't have time for a legitimate chance to score. So there were fewer possessions that were included in each game than in Football Outsiders' numbers. The "low end" of the games I looked at was six or seven non-throwaway possessions. The high end was 12 and up. The vast majority were 8 to 11, which is why I posted those numbers.

If you want the numbers from a very small sample size, I can go back and find all of those numbers again, and we can debate why teams allowed 10 more points per game on 15 possessions than they did with 14 possessions, or why they allowed 10 more points on seven possesions than they did on six, or whatever it was. Like I said in the last thread, the numbers at the extremes, with small sample sizes, fluctuate wildly and don't offer any insight.

Just back up what you said with facts, or if you can't--ignore it.

I have backed it up. More possessions generally leads to more points. That's not ALWAYS the case, like you'd want, just as more pass attempts doesn't always mean more completions for Tony Romo.

 

[percyhoward]

Again, you can't expect teams to perform at their average all of the time in a sample of only 16 games.

Are you not able to distinguish "on average" from "always?"

And that's exactly why points per possession is a better measure than points per game. The offense that scored fewer points and the defense that allowed more points obviously performed better. The only reason they scored fewer points or allowed more points is because of the number of possessions thay had or faced.

It obviously works in your 4 tailored scenarios. Where it fails is when it's applied to real games, such as in your 160-game sample...

Those are averages for 10 teams with different YPP, not the numbers for one team.

So pick any one team (if that's a distinction you want to make) and show how the stat works. What's the difference between how one team's number projects across 160 games and how 10 teams' numbers project across 16 games each? The sample size is the same.

And you still haven't explained why the average "should" remain the same no matter how many possessions.

Adam, what's an average?

Was Romo's average per completion affected by the number of possessions in a specific game? Or Barber's average per carry? No, because those are real averages. Yet, in your 160-game sample, the "average" points per possession kept going down with added possessions.

There were fewer possessions that were included in each game than in Football Outsiders' numbers.

About 300 fewer possessions, is what it comes out to. Considering the fact that FO doesn't count kneel down possessions either, you're really throwing out a lot of information. It's worth exploring what kinds of possessions they're counting that you aren't, to show why your methods are better than theirs (and to show that you aren't shaping the data to suit the stat.)

But again, that's not even really germane to the fact that the points per possession still decreased with added possessions.

When you say this...

"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."

...I don't take it to mean that Team B ALWAYS scores 24 points when it has 12 possessions. And I'm not saying it means Team B will NEVER score 24 points on 12 possessions either.

Look at your Romo example. Of course he didn't ALWAYS get 12.6 yards every time he completed a pass. But on AVERAGE, he did. Sometimes it was 8 yards, sometimes 17 yards, etc. The average was 12.6 per completion.

Team B doesn't ALWAYS score 24 points when it gets 12 possessions. And on AVERAGE, it doesn't either. Sometimes it scores 13, sometimes 21, etc. The average is NOT 24 points, but instead it's closer to the per game average.

That is the difference. When points per possession depends on number of possessions in the game, it's really just points per game.

 

[AdamJT13]

Are you not able to distinguish "on average" from "always?"

Of course I am.

You said, "If a team's 'average level' is 2 points per possession, then it should score 20 points on 10 possessions, and 30 points on 15."

How is that different from saying, "If a quarterback's completion percentage is 60.0, then he should complete 12 passes on 20 attempts, and 18 passes on 30 attempts"?

Neither one is always true, but the averages are still the averages.

Peyton Manning has played 160 career games (the same sample size that I used for possessions). When he has attempted fewer than 20 passes, he has averaged 8.5 completions per game and a 77.3 percent completion rate. When he has attempted 20-29 passes, he has averaged 17.7 completions per game and a 66.9 percent completion rate. When he has attempted 30-39 passes, he has averaged 22.1 completions per game and a 64.9 percent completion rate. When he has attempted 40-49 passes, he has averaged 26.9 completions per game and a 61.4 percent completion rate. And when he has attempted 50 passes or more, he has 30.5 completions per game and a 58.4 percent completion rate.

Hmm, the more attempts he has, the more completions he has, but his completions per attempt goes down. Sounds familiar, doesn't it? (Possessions, points, points per possesion.) Maybe we should be using completions per game to measure how accurate quarterbacks are, instead of completion percentage?

It obviously works in your 4 tailored scenarios. Where it fails is when it's applied to real games, such as in your 160-game sample.

First of all, those were real games from the sample. And it still works there. The better defenses at stopping scoring (without factoring in strength of schedule or field position) allow fewer points per possession. The lesser defenses allow more. And points per game still is deceiving because of the differences in the number of possessions each team has. Just like Peyton Manning's completions per game isn't a true measure of his accuracy because of the differing number of attempts in each game.

So pick any one team (if that's a distinction you want to make) and show how the stat works. What's the difference between how one team's number projects across 160 games and how 10 teams' numbers project across 16 games each? The sample size is the same.

And how do you propose that I "project" 10 seasons for a team that played one?

Was Romo's average per completion affected by the number of possessions in a specific game? Or Barber's average per carry? No, because those are real averages.

You're comparing apples to oranges. Why would Romo's average per completion or Barber's YPC be affected by possessions? That's much different from points per game being affected by possessions. That's more like saying a defense's yards allowed per pass is affected by possessions, which is about as likely to be true as Romo's YPC or Barber's YPC or anything else being affected by possessions.

Yet, in your 160-game sample, the "average" points per possession kept going down with added possessions.

Again, you're ignoring the fact that when defenses play better, they tend to face more possessions, on average, because their opponents are forced off the field more quickly. So of course the average will go down -- the defenses are performing better, on average. (Likewise, a starting quarterback who is playing poorly likely will have to attempt more passes, which is why, for example, Manning's completion percentage goes down as his attempts go up.)

About 300 fewer possessions, is what it comes out to.

It's much less than that, actually (whether you're talking about 300 fewer than the actual number or 300 fewer than Football Outsiders' number).

Considering the fact that FO doesn't count kneel down possessions either, you're really throwing out a lot of information.

My goal is to find out how defenses perform at stopping the opponent from scoring. If I want a true measure, I'm not going to count a possession when the opponent ran out the clock instead of trying to score, or when it had only a few seconds to try to drive 80 yards. I can't speak for what Football Outsiders wants to measure, so they might count possessions that I don't. But I know what I want to measure and what I don't.

That is the difference. When points per possession depends on number of possessions in the game, it's really just points per game.

Again, more possessions faced means -- on average -- the defense was performing better, forcing the defense off the field sooner. But that's not always the case, and it's not the case for every team. If you want an equitable measure of how defenses perform, you HAVE to consider the number of times they were put on the field. Otherwise, you could be rewarding a weak defense whose offense is able to sustain drives, or punishing a strong defense whose offense can't.

If that's what you want, fine, use points per game. I don't care. But most statistical analysts realize that factoring in the number of possessions is a much better way to measure defenses (and offenses).

 

[percyhoward]

The Manning example is a perfect one.

Why do you think nobody cares about "completion percentage per possession?" Because it's a meaningless number. It changes according to the number of possessions his offense has in a game. The more three-and-outs, the fewer passes he's completing, and the more possessions there are.

Just like "points per possession, "yards per possession," or anything else per possession. They all depend too much on how many possessions there are in the game.

Why would Romo's average per completion or Barber's YPC be affected by possessions? That's much different from points per game being affected by possessions.

Now you're making my point.

You said, "If a team's 'average level' is 2 points per possession, then it should score 20 points on 10 possessions, and 30 points on 15."

How is that different from saying, "If a quarterback's completion percentage is 60.0, then he should complete 12 passes on 20 attempts, and 18 passes on 30 attempts"?

There is no difference, as long as you don't use those numbers to make an argument about what that particular offense (or that particular QB) does in an average GAME.

That's your mistake.

Pulling any 10 possessions from the season, at random, the team scored 20 points on average. Pulling any 15 possessions at random, the team scored 30 points. But that doesn't mean that a team that averages 20 points in a 10-possession game, also averages 30 in a 15-possession game.

That's the way you're trying to use the stat, and it's a faulty assumption.

 

[YosemiteSam]

Clearly, though, cornerbacks have to be given credit for covering so well that opponents avoid throwing at them. One way to do this is to look at the number of yards allowed compared to the number of passes faced -- regardless of whether that cornerback was targeted.

There is a down side to this view. Imagine if Roy Williams was your CB on one side and Jacques Reeves on the other side. Reeves would look like a super star as everyone would pick on Williams. So, while teams may avoid one of a teams CBs, they are only doing so because the other CB is far easier to exploit. That in no way should credit the avoided CB as a top CB. It should only point to the fact that he is a better CB that the teams other CB.

 

[MapleLeaf]

To me it is, when people constantly refer to guys being "shutdown" cornerbacks when they don't shut down anyone. Other than maybe 2006, Champ Bailey has never been a "shutdown" type of cornerback. Asante Samuel makes plays, but he's not really a "shutdown" cornerback. Same with DeAngelo Hall (who most people finally realized hasn't been a shutdown cornerback just in time for him to have his best season).

And Deion had only two interceptions in 1995, two in 1996 and two in 1997. He wasn't much of a playmaker as a cornerback in those seasons, but he definitely was a shutdown corner.

...really powerful is if they can be used to typify an individual's play or capabilities.

That's what I see so useful here. Given Adam's analysis if certain metrics boundaries can be established to say a player is a shutdown corner as opposed to a zone corner it would give instant label to the player and their abilities.

As long as the label was derived from actual stats, then for that year the stats were accumulated that would be an accurate impression.

I wonder if YAC after the catch would be a good indicator of the corner's positioning. Typically one would think that a corner who gives up alot of yards per pass attempt is one who is either a "bad" corner or one who is often out of position at the time of the catch.

 

[AdamJT13]

There is a down side to this view. Imagine if Roy Williams was your CB on one side and Jacques Reeves on the other side. Reeves would look like a super star as everyone would pick on Williams. So, while teams may avoid one of a teams CBs, they are only doing so because the other CB is far easier to exploit. That in no way should credit the avoided CB as a top CB. It should only point to the fact that he is a better CB that the teams other CB.

Each cornerback's target percentage is surprisingly independent of his teammate's target percentage. The average target percentage for the 94 cornerbacks that Joyner tracked is 16.67. The highest target percentage for any starting cornerback was 21.8. The extra 5.13 percent (or, in most cases, 2 or 3 percent for highly targeted cornerbacks) doesn't all come off the other cornerback's target percentage, it comes off the other 83.33 percent of targets for the entire defense -- the other cornerback, the safeties, the nickel and dime backs, the linebackers, screen passes, dumpoff passes, throwaways, etc. Or it could even come off the sack total, as better coverage by that cornerback could have resulted in a few more sacks. The difference in the number of targets for the rest of the defense individually -- as a result of playing with a big target -- could have less of an impact on their final stats than if they had allowed one 20-yard catch allowed in a prevent defense at the end of a 30-point blowout. As I said in the original post, the numbers shouldn't be considered precise, so a handful of targets, catches or yards over the course of a season aren't that important.

I went through all 32 starting cornerback tandems, and there are tandems who were both targeted less than the average and tandems who were both targeted more than the average. Most teams have one who was targeted about average and one who was targeted more or less. If we somehow were stuck with Roy Williams and Jacques Reeves at cornerback, then it's likely that both of them would have high target percentages.

 

[AdamJT13]

I wonder if YAC after the catch would be a good indicator of the corner's positioning. Typically one would think that a corner who gives up alot of yards per pass attempt is one who is either a "bad" corner or one who is often out of position at the time of the catch.

Joyner used to measure how close the cornerback was to the receiver each time he was targeted, but he doesn't do that anymore. One problem with using YAC for that is that the cornerback might have tight coverage but miss a tackle and allow a bunch of YAC, or he might give up a long catch-and-run that inflates his YAC allowed. But it would be nice to have YAC allowed just for the sake of having the air/after-catch breakdown of the yardage allowed.

 

[AdamJT13]

The Manning example is a perfect one.

Why do you think nobody cares about "completion percentage per possession?" Because it's a meaningless number.

Again, you're completely comparing apples to oranges. Nobody cares about "completion percentage per possession" because nobody cares about "completion percentage per game."

As I said, Manning's attempts are like an offense's possessions. People obviously care about completions per attempt. They don't care about completions per game because they know that more attempts lead to more completions (ON AVERAGE). What's important is how accurate he is on each attempt, no matter how many he has, which is why completions per attempt is a better measure of his accuracy. It's the SAME with points per game vs. points per possession. Points per possession is a better measure of effectiveness at scoring or stopping scoring, even though the average goes down with more possessions (for the reasons repeatedly stated, and just like Manning's completion percentage goes down with more attempts).

If you want to think of it in even more of a direct comparison, think of a possession as a "pass attempt" and a touchdown as a "completion." Which is more effective at completing its "attempts," an offense with a 33 percent "completion rate" or an offense with a 40 percent completion rate? What if the 33 percent offense averaged 12 "attempts" per game and the 40 percent offense averaged nine? If you think the 33 percent offense is more effective because it has more "completions" per game, it would be the SAME THING as saying a quarterback who completes 60 percent of 35 attempts per game is more accurate than a quarterback who completes 67 percent of 31 attempts per game because he has more completions per game.

Just like "points per possession, "yards per possession," or anything else per possession. They all depend too much on how many possessions there are in the game.

No, they AFFECT how many there are in the game.

Now you're making my point.

I'm pointing out that your examples are unrelated to the issue.

There is no difference, as long as you don't use those numbers to make an argument about what that particular offense (or that particular QB) does in an average GAME.

In case you haven't noticed, I'm not doing that at all. I'm arguing that what the offense or QB does in a game IS NOT what is important. It's what happens on each possession/attempt that is the better measure of effectiveness.

Pulling any 10 possessions from the season, at random, the team scored 20 points on average. Pulling any 15 possessions at random, the team scored 30 points. But that doesn't mean that a team that averages 20 points in a 10-possession game, also averages 30 in a 15-possession game.

For the umpteenth time, that's usually because the team's performance determines how many possessions there are.

Back to Manning. Pulling any 50 attempts from his career at random, he would complete 64 percent of them. But that doesn't mean he averages 64 percent completions in a 50-attempt game or completes 64 percent in a 20-attempt game. It's the same thing.

That's the way you're trying to use the stat

No, I'm not. I've repeatedly stated that teams don't perform at their average level in every game. How many more times does it need to be said?

 

[percyhoward]

I've repeatedly stated that teams don't perform at their average level in every game. How many more times does it need to be said?

I've repeatedly said that you run into problems when you take a number that only works for X number of possessions, and try to apply it to >X.

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.

Since teams don't perform at their average level in every game, and their point total does not go up in proportion to their number of possessions, you have to conclude one of the follwing about that 12-possession game...

1) Team B played much better than its average level, or

2) You got team B's point total wrong.

You see, if they'd played at their average level for a 12-possession game, your own research shows that they would have scored 20 points, not 24.

You can either go by what your 160-game sample showed to be average, or what points per possession says is average. But you have to consider that when points per game collides with points per possession, the latter gets knocked backwards quite a ways while the former hardly budges.

 

[AdamJT13]

I've repeatedly said that you run into problems when you take a number that only works for X number of possessions, and try to apply it to >X.

But nobody is doing that.

What you're doing is like throwing out Peyton Manning's completion percentage because it isn't the same with 15 attempts as it is for 50, then using completions per game to measure how accurate he is.

Since teams don't perform at their average level in every game, and their point total does not go up in proportion to their number of possessions, you have to conclude one of the follwing about that 12-possession game...

1) Team B played much better than its average level, or

No, they played at their average level and got more than their average number of possessions. When they play at their average level, their points total DOES go up in proportion to their number of possessions.

You can either go by what your 160-game sample showed to be average, or what points per possession says is average. But you have to consider that when points per game collides with points per possession, the latter gets knocked backwards quite a ways while the former hardly budges.

I have no idea what you're trying to say.

Here's a question -- other than a raw game total (total points, total yards, etc.), what statistic do you consider to be a good measure of how an offense performed?

 

[percyhoward]

No, they played at their average level and got more than their average number of possessions. When they play at their average level, their points total DOES go up in proportion to their number of possessions.

Think about that.

"They played at their average level and got more than their average number of possessions."

There are only three ways they could have more possessions:

1) Their possessions were shorter
2) Their opponent's possessions were shorter
3) Both teams' possessions were shorter

Only in #2 do they play at their average level and end up maintining their average points per possession. In #1 or #3 playing at their average level would have left them short of the goal line, or they would have had to score quicker. They could not have played at their average level in #1 or #3 and maintained their average points per possession.

Here's a question -- other than a raw game total (total points, total yards, etc.), what statistic do you consider to be a good measure of how an offense performed?

When you put it to the test, offensive points per game isn't such a bad one after all.

 

[percyhoward]

To boil it down to just one question...

Given that a higher number of possessions usually means the team isn't performing up to its average level, why would a team that scores 18 points on 9 possessions, score 24 on 12?

 

[AdamJT13]

Think about that.

"They played at their average level and got more than their average number of possessions."

There are only three ways they could have more possessions:

1) Their possessions were shorter
2) Their opponent's possessions were shorter
3) Both teams' possessions were shorter

Only in #2 do they play at their average level and end up maintining their average points per possession. In #1 or #3 playing at their average level would have left them short of the goal line, or they would have had to score quicker. They could not have played at their average level in #1 or #3 and maintained their average points per possession.

That's not true. They can have quicker possessions and score at the same average per possession -- they can get good field position a few times, get a few big plays, have some quick turnovers or three-and-outs, etc. They'd have to score on only four of 12 possessions (three TDs and a field goal) to maintain their average points per possession. All of the nonscoring possessions could be shorter than normal without affecting their points per possession at all.

When you put it to the test, offensive points per game isn't such a bad one after all.

That doesn't answer the question. And points per game is not as good of a measure as points per possession, which you've already admitted in at least some cases.

To boil it down to just one question...

Given that a higher number of possessions usually means the team isn't performing up to its average level, why would a team that scores 18 points on 9 possessions, score 24 on 12?

If it wasn't performing at its usual level, it wouldn't score 18 points on nine possessions, either.

Back to the question you didn't answer -- other than a raw game total (total points, total yards, etc.), what statistic do you consider to be a good measure of how an offense performed?

If all you can think of is total points, you might want to think back to this example --

Offense A's possessions
5 plays, 14 yards, PUNT
3 plays, 20 yards, TOUCHDOWN
3 plays, 8 yards, PUNT
3 plays, 15 yards (plus 13 in penalties), TOUCHDOWN
3 plays, 36 yards, TOUCHDOWN
3 plays, minus-4 yards, PUNT
6 plays, 50 yards, TOUCHDOWN
5 plays, 44 yards, TOUCHDOWN
3 plays, 6 yards, PUNT
7 plays, 52 yards, FIELD GOAL
3 plays, 4 yards, PUNT
3 plays, minus-1 yards, PUNT
6 plays, 14 yards, INTERCEPTION
6 plays, 25 yards, PUNT


Offense B's possessions
4 plays, 96 yards, TOUCHDOWN
13 plays, 74 yards, FIELD GOAL
5 plays, 65 yards, TOUCHDOWN
10 plays, 78 yards, TOUCHDOWN
10 plays, 69 yards, TOUCHDOWN
3 plays, minus-7 yards, PUNT
13 plays, 69 yards, FIELD GOAL


Offense A scored 38 points. Offense B scored 34 points. But you already said Offense B performed better, so total points can't be the best measure.

 

[J-DOG]

To me it is, when people constantly refer to guys being "shutdown" cornerbacks when they don't shut down anyone. Other than maybe 2006, Champ Bailey has never been a "shutdown" type of cornerback. Asante Samuel makes plays, but he's not really a "shutdown" cornerback. Same with DeAngelo Hall (who most people finally realized hasn't been a shutdown cornerback just in time for him to have his best season).

And Deion had only two interceptions in 1995, two in 1996 and two in 1997. He wasn't much of a playmaker as a cornerback in those seasons, but he definitely was a shutdown corner.

I agree with that whole-heartedly.
The days of a shutdown cb are over.
The only true shutdown corners I can remember off the top of my head are Deion and Mike Haynes(New England and Oakland).

 

[percyhoward]

Adam

When you say that a team that scores 18 points on 9 possessions will score 24 on 12...it's assumed that you are talking about what happens with that team on average in a 12-possession game, with all else being equal.

What you're doing now is like if I said, Romo averages 250 yards per game, so in his next game he will get 400 yards. Why? Because there might be some turnovers by the other team, big plays, etc.

Here is what you said:

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.

If I misunderstood, then what were you trying to say?

 

[AdamJT13]

Adam

When you say that a team that scores 18 points on 9 possessions will score 24 on 12...it's assumed that you are talking about what happens with that team on average in a 12-possession game, with all else being equal.

And if it performs at its usual level.

A team that averages 18 points and nine possessions per game scores 2.0 points per possession. If it performs at exactly its usual level scoring-wise, it will score 2.0 points per possession no matter how many possessions it gets.

What you're doing now is like if I said, Romo averages 250 yards per game, so in his next game he will get 400 yards. Why? Because there might be some turnovers by the other team, big plays, etc.

Huh? Now you're not just mixing apples and oranges, you're throwing in bananas, pineapples and kiwi as well.

If Romo averages 10 yards per attempt and 25 attempts per game, then got 40 attempts in the next game, how many yards would he have if he performs at his usual level?

 

[percyhoward]

And if it performs at its usual level.

If a team doesn't usually perform at 2.0 when it has 12 possessions as opposed to 9, then how is 2.0 its usual level for a 12-possession game?