Writers of Pro Football Prospectus 2008

11 Oct 2005

Just Catch the Ball

Some folks might remember an article by Kevin Pelton that we ran a few weeks ago, looking at wide receiver drops in terms of drops per pass instead of just total drops. Now, Ben Alamar takes things one step further over at the PROTRADE site, using STATS Inc. data to adjust "drop rate" to remove balls that the quarterback underthrows or overthrows. He also adjusts based on the length of each receiver's routes. The new lists aren't much different than the ones Kevin ran, but there are a few new interesting players. For example, Cedrick Wilson now shows up in best drop rates, and Darrell Jackson is back onto the list of worst drop rates.

Before you ask: No, I don't have access to this STATS Inc. data; yes, I'm looking into the possibility of getting access to it now that we're partnered with FOXSports.com; and yes, this is why we've got volunteers charting games this year, so that we have our own data in the future without worrying about the legal ramifications of such things. Thanks and tip your waitresses.

Posted by: Aaron Schatz on 11 Oct 2005

34 comments, Last at 13 Oct 2005, 10:08am by tracey

Comments

1
by MDS (not verified) :: Tue, 10/11/2005 - 4:02pm

Why do we hear so little about the Patriots' receivers? They might be the best pair in the league, but you'd never know that based on the football media.

2
by Israel (not verified) :: Tue, 10/11/2005 - 4:11pm

If Deion Branch dropped zero passes, how can his adjusted rate be anything higher than zero? They have him at 0.6% (Damn statistics!)

3
by B (not verified) :: Tue, 10/11/2005 - 4:13pm

Maybe he caught a ball he should have dropped? Or maybe it's assumed that if he has 200 catchable balls thrown his way, he'd miss one?

4
by melissa (not verified) :: Tue, 10/11/2005 - 4:22pm

Well one reason that the NE receiving corps may not have been talked about a lot is that the statistics used for this study are from 2004 when David Patten, who is on Pelton's list for highest drop rates, was on the Patriots. So the receivers as a group had a higher drop rate. Plus, usually when you do hear about the Patriots receivers in the national media it is about how "similar" and "interchangeable" they are and maybe that just isn't exciting for people to talk about, although Branch did win Superbowl MVP and that is actually a lot of recognition for a receiver.

5
by Pat (not verified) :: Tue, 10/11/2005 - 4:22pm

Well, statistically if you see "zero", and you treat it as a random process, you can estimate an upper limit on the number you could've seen as "3". (So if he gets 200 balls thrown his way, the upper limit would be 3/200, or 1.5%) Adjust that for distance and the like, and you might get 0.6%. Then again, that's properly still an upper limit, but it's better than saying "0.0%", because "0.0%" doesn't give you an idea of how strong that statement is.

6
by Richie (not verified) :: Tue, 10/11/2005 - 4:26pm

Is anybody else playing on PROTRADE? I am, but still really don't understand how the scoring works. I understand the part about DVOA being used to calculate dividends, but not how the dividends and portfolio value are calculated.

Also, how come FO.com doesn't "remember" me any more in the comments section?

7
by melissa (not verified) :: Tue, 10/11/2005 - 4:27pm

re:2
I interpretted the greater than zero adjusted drop rate to mean that Branch was not thrown as many long passes as other receivers so the adjustment makes up for that by assuming that if he was thrown more long passes he would have had a higher drop rate.

8
by Carl (not verified) :: Tue, 10/11/2005 - 4:42pm

"Darrell Jackson is back onto the list of worst drop rates."

I'm shocked! Shocked to see that he's dropping balls!

9
by Reno (not verified) :: Tue, 10/11/2005 - 4:46pm

I'm a total PROTRADE addict. An important thing to note is that the dividend system doesn't seem to be DPAR so much as PAA - Points Above Average (rather than replacement), without an adjustment for opponent strength. Don't quote me on that, but that's what it seems like.

As far as your value is concerned, you buy shares in players. For each share you own in a player, you get dividends (positive or negative). Your portfolio value is also affected by the rising and falling of his stock price. So your portfolio value reflects both the market value of your players (shares * share prices), and the dividends you've earned from them (shares * dividends). You only earn dividends, though, for what the players you own do on the field WHILE YOU OWN THEM. If you buy Mark Bulger on Tuesday, you don't get credit for what he did the previous Sunday, and if you sell him on Friday, you won't get dividends for his performance the next game.

Those are the basics, I'm not sure what exactly you're asking, so don't know if that answers it.

PROTRADE is awesome, I like it a hell of a lot more than fantasy football.

10
by safelouis (not verified) :: Tue, 10/11/2005 - 5:06pm

re #9
That sounds like an absurd farce that has about as much to do with actual football as the Amish have to do with the popularity of Paul Oakenfold.

11
by Richie (not verified) :: Tue, 10/11/2005 - 5:07pm

What does it mean when it says I can't own less than 5% of a player?

I just bought Cedric Benson, but in my portfolio it indicates that he has "last game dividends". He shouldn't since I didn't have him then.

Is there any way to see an actual breakdown of my dividends?

How the hell does some guy have a $200,000 portfolio?

I'm just not finding that site to be easy to use.

12
by Jim A (not verified) :: Tue, 10/11/2005 - 5:12pm

Some wide receivers run more deep routes where the ball is inherently harder to catch (and easier to drop).

Could you show some data to support this? It seems like it might be true, but given that poor throws and defensed passes were discarded, it's certainly not obvious. If a pass was thrown perfectly, does it matter much how far it traveled to get there? I would guess that velocity and trajectory would be just as important factors, and those are probably dependent on the route type and defensive coverage as much as simply route depth.

13
by zip (not verified) :: Tue, 10/11/2005 - 5:54pm

re: #12

I'd be a lot more likely to catch a pass thrown at any velocity from 15 yards away than I would be to catch one over the shoulder running full speed. But I'm not an NFL receiver.

It seems like there's a lot of deep balls thrown that are close enough that the receiver can get a hand on them (and thus can't be called "overthrown") but are still hard to hang on to. But maybe I'm exaggerating how long a pass has to be to be considered "deep."

14
by princeton73 (not verified) :: Tue, 10/11/2005 - 6:00pm

If Deion Branch dropped zero passes, how can his adjusted rate be anything higher than zero?

he actually caught a ball that wasn't even thrown

15
by pawnking (not verified) :: Tue, 10/11/2005 - 6:03pm

I just got a BIG red flag on this study. "..passes knocked down by an adept defensive back [are not counted against the WR]." This is a case of logic getting in the way of a true picture. Part of the game some WRs have is using their big body or a precise route to prevent an "adept defensive back" from making a play. Should a WR bear NO responsibility for allowing the DB to make such a play? This statistic assumes so. I strongly disagree, while conceeding the point that a DB will sometimes be responsible for a drop.

16
by Bob (not verified) :: Tue, 10/11/2005 - 6:35pm

re #15

It's not that the WR bears no responsibility for allowing the defender to make the play. That skill is definitely part of measuring a WR's overall production. But this study is purely looking at WR's "hands". Speed, route running ability and other factors are not included beacuse they're not trying to evaluate WR's overall abilities.

17
by Reno (not verified) :: Tue, 10/11/2005 - 6:44pm

Re #11,

I don't know how exactly the 5% thing works, I've only had it complain to me on a very few occasions, and have then just bought an extra share or two to make it work.

The guy has over 200 grand by buying low and selling high. You'll notice that some people will load up heavily on say, Alex Smith when it looks like he'll start, or Antowain Smith when McAllister is injured. They'll invest as heavily as possible in these players, and then try to sell them when their price has peaked. Thus they make a ton of money.

The idea is just like a stock market: you try to spot stocks, that, based on your information, you feel are undervalued. And this being about football, value can change based on promotions/demotions, injuries, etc. So it makes sense to buy up Kelly Holcomb at $10 if he's going to start, becuase even though Kelly Holcomb sucks, suck starting quarterbacks are worth more than $10 on the market. So you buy Holcomb, wait for his price to hit what you think he's worth, and then probably sell him for a substantial profit before he can actually take a snap and have his on-field performance affect your portfolio.

One thing I'm not crazy about PROTRADE for is that there's no realy incentive to invest for the long term, the way one (likely) would in the real stock exchange. For example, Aaron Rodgers at $15 looks like a pretty good buy right now, provided one were thinking five years down the road, but for this week/this season, there's no point in buying him unless you think Favre'll get injured or bench (which you might, and would thus make him a good buy).

18
by DavidH (not verified) :: Tue, 10/11/2005 - 8:15pm

I don't get why we can't use our dividends to buy more players...

19
by Reno (not verified) :: Tue, 10/11/2005 - 10:44pm

That's annoying, it's definitely no fun having dead money. But I think it's because of the potential for negative dividends. Since most people have all of their money invested in players, with little to no available funds, where would a net negative dividend be taken from? This actually happened to me in my tight end portfolio where I was very heavily invested in Chris Baker. He somehow got the ball on a kick-off and fumbled, and I ended up posting a -$900 for last week's games.

I suppose that they could let you carry a negative balance that would be taken out of your next sale, or automatically sell off shares in the negative players until you were in the black again, but neither of those seems like ideal solutions to me.

It would be nice if there was some kind of PROTRADE thread here, what with Aaron being involved and all.

20
by Jim A (not verified) :: Tue, 10/11/2005 - 11:05pm

Re: #12, on the other hand when running full speed to catch a deep pass over your shoulder you generally don't have to worry about changing directions or getting hit immediately after the catch. A short pass is more likely to be in traffic, and it seems like lots of drops happen when a receiver tries to avoid a hit or turn and run before catching it.

I would also surmise that short passes are more likely to be thrown accurately, which is a prerequisite for a drop under the definition used here.

I'm sure lots of factors go into determining whether a pass is dropped, and I'm willing to believe pass depth is one of them, but I wouldn't automatically assume it is so without seeing supporting data.

21
by Aaron (not verified) :: Tue, 10/11/2005 - 11:42pm

Hmmm. Well, I certainly have no problem with people who are trying out PROTRADE making comments here, and I've told the PT people to come take a look. Meanwhile, people can talk about the drop article also.

22
by Jeff Ma (not verified) :: Wed, 10/12/2005 - 1:03am

Thanks to all of you guys who have left comments on Ben's article and on PROTRADE... I will encourage Ben to post some comments about your criticisms of his work as I'm sure it can be a great dialogue. As for thoughts on PROTRADE, this product was made for you guys, the type of people that read FO and BP and in order for us to create a really great product we need your feedback (no one ever gets anything right on the first try)... I encourage you guys to send any feedback or thoughts to support@protrade.com and hope that we can make a great product together.

Thanks and thanks to Aaron and FO for helping us build our company.

23
by Jamie T. (not verified) :: Wed, 10/12/2005 - 11:50am

Another thing with the Protrade magic $200k portfolio; alot of those guys started buying players 1 and a half minute after protrade opened for trading. For instance: I bought Manning on Sept 19 before alot of people started trading on Protrade. His price was $219 then. Since that time the amount of people participating in the Protrade Open Challenge has trippled. More investors means a higher price for all of the players. Manning's value now is $267. I suspect that this year we will see these same fortunate guys who have these unbeleivably large portfolios continue to dominate the rest of us latecommers. Next year will be a different matter..unless Protrade explodes in popularity again after the first week or two of trading.

24
by Mark (not verified) :: Wed, 10/12/2005 - 1:20pm

Re: 23

The challenges are judged on total percentage appreciation, not dollar appreciation. So a trader that starts a challenge with $200,000 should have the no advantage over a trader that starts with $25,000.

25
by Ben A (not verified) :: Wed, 10/12/2005 - 3:14pm

Thanks again for the interst in my article. As for the evidence requested: the adjustments to the drop rate are done based on an estimated regression equation using wrs drop rates as the dependent variable and the % of their catches that are longer than 15 yards before YAC, % of their catches that are between 7 and 15 yards before YAC, and the % of the teams attempts that were thrown their way as the independent variables. The coefficients were all statistically significant with the long ball and medium ball coefficients both being positive (indicating that the higher percentage of long passes is associated with a higher drop rate).

26
by Jim A (not verified) :: Wed, 10/12/2005 - 6:00pm

Ben, thanks for checking in here. Any chance you could post the actual drop rates (and totals) by pass depth?

27
by Dan (not verified) :: Wed, 10/12/2005 - 6:55pm

Regression equations can be useful, but may not fit the data very well. I'd like to know what the adjusted R squared and tolerance values are and also what the distribution of the residuals looked like, for starters.

28
by Ben A (not verified) :: Wed, 10/12/2005 - 7:27pm

R2 = 0.556 and residuals are normal (no heteroskedasticity or serial correlation)

29
by Dan (not verified) :: Wed, 10/12/2005 - 7:32pm

Re: 28

Impressive.

30
by DavidH (not verified) :: Thu, 10/13/2005 - 12:04am

heteroskedasticity

You had to have just made this word up.

It sounds like the capital of some fake Scandinavian country.

31
by Jim A (not verified) :: Thu, 10/13/2005 - 12:13am

Here's a quick and dirty way to look at drops by distance using numbers from Scientific Football. KC Joyner lists attempts and drops for each qualifying WR in 2004 by short, medium and deep routes. Maybe someone want to type (or scan) them all in and compute totals, but I looked at just the median number. The median number of short attempts was 47 and median number of short drops was 2, so that's a 4.3% drop rate. At medium depth the medians were 18 attempts and 1 drop (5.6%) and deep route medians were 22 attempts and 1 drop (4.5%). Combining medium and deep routes gives you 5%. Mean averages would probably be higher and perhaps more significant because a few drop-prone receivers tend to skew the results. But this does seem to suggest that deeper passes are slightly more likely to be dropped. This is obviously very rough, but if it holds true on further scrutiny, it implies that a WR's typical route depth likely wouldn't account for more than one additional (or one fewer) drop over the course of an entire season.

32
by Jeff F (not verified) :: Thu, 10/13/2005 - 3:25am

Are the stats from last year or this year? I know, probably last year, but Branch has ONE dropped pass this year! That's unacceptable!

33
by Michael (not verified) :: Thu, 10/13/2005 - 4:04am

Does the "accurately thrown" requirement cause great catches to be excluded? These would be passes that would have definitely been excluded had they been dropped.

34
by tracey (not verified) :: Thu, 10/13/2005 - 10:08am

RE: Drop rate and length of the pass

For those debating whether longer balls are more difficult to catch, I say both sides are right.´ Pass length is a factor, but it certainly isn't the only factor.

The problem is that reality is more complex than just short, medium and long. As one post mentioned, a short pass in zone coverage, between linebackers, is harder to catch then a quick hitch on the outside.

But is still may be helpful to simplify things as short, med and long, to prevent the analysis itself from becoming too complex. As long as this simplification still does a "good" job of describing reality.