What if Passer Ratings Were Used for WR's?

[percyhoward]

I'd wager that almost any knowledgeable football observer would conclude that interceptions have a lot more to do with the quarterback than the receiver. The question, then, is why would you weight interceptions equally as heavily for a receiver as you would for a quarterback?

As long as you don't try to conclude that Miles Austin's 127.8 rating as WR means he had a better year than Tony Romo with his 97.6 rating as a QB, what's the difference?

If we're trying to decide how much to count an interception as, I still vote "1."

 

[NinePointOh]

As long as you don't try to conclude that Miles Austin's 127.8 rating as WR means he had a better year than Tony Romo with his 97.6 rating as a QB, what's the difference?

You are saying that a receiver who puts up big numbers but whose quarterback throws interceptions while aiming for him should have his production discounted. That penalty will hurt some receivers more than others, and for the most part it will have very little to do with the receiver's own play. So even if you're not comparing WRs to QBs, you're unfairly penalizing guys like Reggie Wayne and unfairly rewarding guys like Mario Manningham and Mike Sims-Walker, who were vastly inferior to Wayne in every single other statistical category.

Technically, it really wouldn't matter whether the weight was or wasn't exactly equal to that in the QB rating formula -- the main point is that it's far, far too powerful a factor compared to its actual usefulness in comparing receivers. It just happens to be too powerful because you're trying to fit a formula that's typically applied to players for whom interceptions are actually extremely important and who are actually directly responsible for almost every interception.

If we're trying to decide how much to count an interception as, I still vote "1."

For receivers, I vote "0" unless you actually sit down and differentiate when it's their fault, which is rare.

 

[percyhoward]

9.0 I appreciate the passion you put into your argument.

What if you were to substitute the "INT" part of it with, say, "drops?" Now a drop counts as much as an INT. That seems too much, since every INT is a turnover while many drops don't even hurt team involved.

You could use "3rd down drops," but that still leaves out drops on 1st or 2nd down that would have led to big plays. You could assign a scorer to keep track of drops, tipped passes that result in INT's, rounded-off or wrong patterns, lack of effort, any other receiver error you can think of, and you see where this is going. It's turning into baseball. A lot of people like baseball.

It all depends on what you're trying to get out of the statistic. If all you're trying to get out of it is "which receivers helped their teams most with the least negative plays (whomever is blamed)", I think this receiver rating is a pretty decent stat.

One person sees Wayne's low ranking, and hears the list telling him Wayne isn't as good as he is cracked up to be, and that person disagrees with the list. Somebody else looks at Wayne's ranking and hears the list telling him nothing more than that Wayne's production didn't come without a cost to his team.

 

[percyhoward]

I predict that the top 5 all fall below #5 in the ranking next year, although all will stay in the top 15.

That's pretty close to what happened--the whole top 5 from 08 stayed in the top 15 in 09 (athough Bolden and Jackson didn't fall that far).

My list was actually more static than the yardage list, where 2 of the top 5 from 08 fell out of the top 15 in 09.

2008/player/2009
1 Bolden 6
2 Fitzgerald 11
3 VJackson 4
4 RMoss 15
5 Jennings 14

6 Smith CAR 27
7 Moore NO --
8 Ward 12
9 Welker 5
10 Muhammad --

11 SMoss 24
12 Breaston --
13 Mason 20
14 Wayne 22
15 Royal --

 

[THUMPER]

Again, I don't dispute that WRs can have some impact on the number of interceptions. However, the percentage of interceptions where the receiver is at fault in any noticeable way is so low that it may not be worth including in the formula in the first place. At the very least, you'd need to either (a) actually watch the play and judge whether the receiver had anything to do with it, or (b) weight interceptions far less heavily than you do in the QB rating formula.

That would make the most sense. Good idea.

 

[NinePointOh]

9.0 I appreciate the passion you put into your argument.

What if you were to substitute the "INT" part of it with, say, "drops?" Now a drop counts as much as an INT. That seems too much, since every INT is a turnover while many drops don't even hurt team involved.

Why bother forcing receiving stats into a formula made for QBs in the first place? If there's no receiving statistic comparable to INTs for QBs, then it would probably make more sense to just use a different formula.

You could use "3rd down drops," but that still leaves out drops on 1st or 2nd down that would have led to big plays. You could assign a scorer to keep track of drops, tipped passes that result in INT's, rounded-off or wrong patterns, lack of effort, any other receiver error you can think of, and you see where this is going. It's turning into baseball. A lot of people like baseball.

So if it's too difficult to keep track of plays where receivers make mistakes, we should discount their production by the mistakes other players on their team made? Should we also divide the stat by the number of times the running back fumbled?

It all depends on what you're trying to get out of the statistic. If all you're trying to get out of it is "which receivers helped their teams most with the least negative plays (whomever is blamed)", I think this receiver rating is a pretty decent stat.

Except that combining the positive plays of a single player with the negative plays by "whomever" leads to a very convoluted statistic without a straightforward use.

One person sees Wayne's low ranking, and hears the list telling him Wayne isn't as good as he is cracked up to be, and that person disagrees with the list. Somebody else looks at Wayne's ranking and hears the list telling him nothing more than that Wayne's production didn't come without a cost to his team.

I don't care whether any specific player is ranked high or low as long as they're ranked according to criteria that make sense. I could make up an arbitrary formula and produce an ordered ranking of receivers, but I'd still have to justify why I included the stats I included and why I combined them using the specific formula I used.

I haven't seen a single justification as to why interceptions should discount a receiver's production so powerfully, let alone why they should be included in the formula at all if they usually have little to do with the receiver's play.

 

[percyhoward]

That would make the most sense. Good idea.

I'll make you guys a deal. If you can come to an agreement about how much to count interceptions, I'll make a new list.

 

[theogt]

The passer rating formula doesn't attempt to "weigh" any of the variables more than the other. It just attempts to have each variable have the same variance. So what you would do is just run a variance test on each variable group until you got where you needed to be.

 

[percyhoward]

The passer rating formula doesn't attempt to "weigh" any of the variables more than the other. It just attempts to have each variable have the same variance. So what you would do is just run a variance test on each variable group until you got where you needed to be.

Does that apply to the "interception problem?"

 

[theogt]

Does that apply to the "interception problem?"

Not sure what you're asking. I haven't followed the conversation in the thread.

 

[percyhoward]

Not sure what you're asking. I haven't followed the conversation in the thread.

You were responding to something, weren't you?

 

[theogt]

You were responding to something, weren't you?

If the "interception problem" is determining how much to "weigh" interceptions in a formula, then yes it applies to the interception problem.

 

[percyhoward]

If the "interception problem" is determining how much to "weigh" interceptions in a formula, then yes it applies to the interception problem.

So how do you "figure out" how much interceptions should count?

 

[theogt]

So how do you "figure out" how much interceptions should count?

http://cowboyszone.com/forums/showpost.php?p=3428426&postcount=48

 

[percyhoward]

Could you possibly be a little more vague, theo?

 

[theogt]

Could you possibly be a little more vague, theo?

Take a group of statistics (e.g., every variable of the passer rating formula for every wide receiver in a given year). Run a variance test and see what the results are for each statistic. You can do this using Excel. Then adjust the divisor on the variable until you reach the same variance for each statistic group.

For example, here are the four variables for the passer rating formula:

a = (((Comp/Att) * 100) -30) / 20
b = ((TDs/Att) * 100) / 5
c = (9.5 - ((Int/Att) * 100)) / 4
d = ((Yards/Att) - 3) / 4

If you apply each variable formula individually for every QB in a given year and then run a variance test on the results, you'll have a very similar variance for each group of statistics. In other words, the variance for "a" for every QB in a given year will be substantially similar to the variance for "b," "c," and "d" for every QB in a given year.

The idea is to achieve this same thing for a group of WRs in a given year. What you would do is set up an excel sheet and have the results for a, b, c, and d for every WR. You would adjust the divisor (i.e., 20, 5, 4, 4) for each variable formula until the resulting variance for each statistic group is the same. If there's too much variance, you raise the divisor and vice versa. Some people would say this is adjusting the "weight" given to each statistic. The idea is just to have the same variance among each variable such that they're comparable enough to then add them together into a single rating.

 

[percyhoward]

I'm convinced that it would yield something objective, without a doubt. But if it involved any changes in the formula itself (as opposed to modifying the number you'd put into the already existing formula), then that lets me out, because I didn't do any calculations myself. I just used the QB rating calculator online.

Interesting to know how they came up with the rating, nonetheless.

 

[NinePointOh]

The passer rating formula doesn't attempt to "weigh" any of the variables more than the other. It just attempts to have each variable have the same variance. So what you would do is just run a variance test on each variable group until you got where you needed to be.

Trying to make it work exactly like the passer rating is the problem in the first place. For a more meaningful statistic, interceptions really shouldn't be included at all for receivers.

Unfortunately, creating meaningful statistics does tend to involve a little calculation.

 

[percyhoward]

Trying to make it work exactly like the passer rating is the problem in the first place. For a more meaningful statistic, interceptions really shouldn't be included at all for receivers.

Unfortunately, creating meaningful statistics does tend to involve a little calculation.

Not counting interceptions at all would make the calculations easier and the results less meaningful.

To affect the order of the list, you either have to:

a) exclude interceptions from the stat altogether and accept the less complete picture that it gives you,
b) score each individual interception and assign blame (as done with drops),
c) count each player's interceptions less and less the more they have, so that the 5th interception might be worth 1/2 INT, the 6th worth 1/4, and so on.

Like I said, if you just want interceptions to count less, I can lower the value of interceptions across the board. But unless you plan to use the results to compare a WR's rating to a QB's, it wouldn't make any difference.

 

[TonyRomo09]

not to get off topic.. but looking at that stat site you used... D-Ware led the league in pressures.... BY 20!!! 56 when the next best person had 36