Wisdom of Crowds in Review: WR, TE
by Bill Barnwell
In this final look back at our Wisdom of Crowds features from before the 2009 season, I'll be reviewing how the performance of wide receivers and tight ends matched up with what really happened this past year.
For reference, to generate the "predicted" lines for this feature, I asked my followers on Twitter to come up with the final reception, receiving yard, and receiving touchdown totals for a variety of receivers, assuming in the process that each would make it through a 16-game season. You can read the initial articles reporting the predictions here, here, here, and here.
Predicted: 70 receptions, 986 receiving yards, 8 TD
Actual: 55 receptions, 829 receiving yards, 5 TD
Whether you want to blame it on the quarterback play, the injuries up front, or T.O.'s own performance, this wasn't what the Bills paid for. I didn't think Owens would come close to the 14.1 yards per catch users predicted, seeing his issues as a vertical threat in 2008 in Dallas, but he actually hit 15 yards per reception in Buffalo. Of course, take out one 98-yard touchdown pass, and Owens falls to 13.5 yards per catch.
Predicted: 64 receptions, 922 receiving yards, 7 TD
Actual: 57 receptions, 757 receiving yards, 3 TD
The issues with Chicago's offensive line helped produce the 2009 line above, with porous pass blocking that forced Jay Cutler into scrambling or getting rid of the ball earlier than he hoped. That prevented Hester from getting downfield and took away the threat of the double move. His play clearly regressed as the season went along, which was disconcerting.
Predicted: 68 receptions, 869 receiving yards, 5 TD
Actual: 51 receptions, 722 receiving yards, 0 TD
Holt was one of the players KUBIAK stood the farthest out on heading into 2009... because, well, there was no one on the Jaguars to throw to besides him. Holt started out as the number-one receiver, but as time went along, injuries and the emergence of Mike Sims-Walker made Holt the clear number-two. The result was a disappointing season that failed to meet either KUBIAK or the crowd's expectations.
Predicted: 62 receptions, 943 receiving yards, 6 TD
Actual: 39 receptions, 600 receiving yards, 4 TD
One of the league's most erratic players on and off the field, Bryant was a total question mark heading into 2009. Using ADP, we can infer that fantasy owners saw him as equal to DeSean Jackson, Bernard Berrian, and Lee Evans. He finished last in that group, struggling with groin and knee injuries while failing to develop any sort of rapport with Josh Freeman. The yards per catch figure predicted was within .1, though, which is nice.
Predicted: 70 receptions, 1129 receiving yards, 7 TD
Actual: 79 receptions, 1248 receiving yards, 5 TD
That's just about a direct hit. Holmes technically emerged as the team's top receiver, receiving one more target than Hines Ward did. Their advanced metrics were virtually identical, finishing back-to-back in DVOA and within three spots of each other in DYAR. At this point, they really are 1 and 1A.
Predicted: 101 receptions, 1487 receiving yards, 11 TD
Actual: 101 receptions, 1569 receiving yards, 9 TD
That's good enough for government work, with the reception total spot-on and the fantasy point total within three points of the predicted figure (214 to 211). I wondered whether losing Owen Daniels affected Johnson's catch rate; it went down from 70 percent in 2007 and 68 percent in 2008 to 59 percent in 2009. Daniels went down in the first quarter of Week 8, so we'll split the data there. From Week 1 through Week 7, Johnson's catch rate was 54 percent; from Week 8 on, it was 62 percent. So the answer is, well, no. I suspected the predicted yards per catch total would be too high, but Johnson blew it away, thanks in part to improved play by the Texans' tackles in pass protection.
Predicted: 66 receptions, 927 receiving yards, 6 TD
Actual: 34 receptions, 480 receiving yards, 2 TD
I asked for Clayton predictions shortly after Derrick Mason had announced his retirement; shortly after the collection process was finished, Mason un-retired. So the prediction is mostly irrelevant.
Predicted: 71 receptions, 1070 receiving yards, 6 TD
Actual: 63 receptions, 1167 receiving yards, 9 TD
We'll have more on the historical strangeness of Jackson's season in FOA2010, I'm sure, but it sure seems like he caught more than 61 passes. In the end, his 2009 usage figures bear a remarkable resemblance to his 2008 campaign; the difference is that he averaged nearly four yards more per catch, and scored seven more touchdowns. Is his propensity for 50+ yard catches sustainable?
Predicted: 52 receptions, 748 receiving yards, 5 TD
Actual: 48 receptions, 625 receiving yards, 2 TD
That's actually pretty close, but remember that the crowd was asked to predict Crabtree's 16-game totals. This is what he put up in 11 games as a rookie; extrapolate that to a 16-game season, and he's at 70 catches and 900 yards. And that's with joining the team in midseason. He's going to be frighteningly good next year.
Predicted: 85 receptions, 1297 receiving yards, 15 TD
Actual: 83 receptions, 1264 receiving yards, 13 TD
Yes, off by two receptions, less than three-tenths of a yard per catch, and two touchdowns. That's...um...terrifyingly accurate. I noted that Moss should exceed that average expectation if he remained healthy; Moss separated his shoulder at some point during the season, with the Patriots keeping both the exact timing and the severity under wraps.
Predicted: 86 receptions, 1072 receiving yards, 8 TD
Actual: 79 receptions, 911 receiving yards, 3 TD
I won't regale you again with the story of how the Vikings nearly stifled Sidney Rice's career by signing Houshmandzadeh, but boy, did Housh not make the impact in Seattle he was expected to. With his vertical metrics falling off a cliff and his DVOA and DYAR in decline, all that Houshmandzadeh brought to Seattle was an ability to stay healthy and what had been an average catch rate by receiver plus-minus standards. Even his catch rate fell, with his annual raw rate of 66-68 percent hitting 59 percent during his maiden voyage in Seattle.
Predicted: 97 receptions, 1391 receiving yards, 12 TD
Actual: 97 receptions, 1092 receiving yards, 13 TD
Nailing two out of three categories is impressive, but Fitzgerald came up 27 percent short on the yardage projection. It's totally out of character with his career; his yards per catch figure had steadily risen since his rookie year, starting at 13.3 yards per catch and peaking at 14.9 in 2008. No one could have expected him to fall off to 11.3. It wasn't caused by his injury, either; even before hurting his leg against the 49ers in Week 14, Fitzgerald was only averaging 11.6 yards per catch.
Predicted: 76 receptions, 918 receiving yards, 6 TD
Actual: 57 receptions, 821 receiving yards, 3 TD
Cotchery's role was usurped by Braylon Edwards, even though Edwards' advanced metrics sucked, as typical. Yes, Edwards helps create space for Cotchery because of his deep speed and leaping ability, but Edwards has played across from plenty of receivers in his career, and none of them came close to matching Cotchery's metrics this year.
Predicted: 104 receptions, 1052 receiving yards, 8 TD
Actual: 123 receptions, 1348 receiving yards, 4 TD
Remember that Welker accrued those numbers in what amounts to 13 games against an elite slate of pass defenses. Even if you just adjust the figures to a 16-game slate, he hits 150 catches if he makes it through 16 full games. 150! He's like some relic from the Dead Ball Era that pitches 350 innings. Of course, making a prediction for his 2010 performance is going to be close to impossible.
Predicted: 37 receptions, 536 receiving yards, 3 TD
Actual: 9 receptions, 124 receiving yards, 1 TD
There's no point in blaming Heyward-Bey for where he was selected. That's not his fault. Looking back at history, there have been 35 wide receivers selected in the first round that failed to clear the 10-catch mark as a rookie. Most of them are from a different era. If we only look over the past 20 years, there have been ten other players. They're a varied mix of successes-to-be, injury cases, and failures exposed as such: Ike Hilliard, Randal Hill, Desmond Howard, Michael Jenkins, Thomas Lewis, Robert Meachem, Johnnie Morton, Santana Moss, Marcus Nash, and Rashaun Woods. Considering the abysmal development rates of Oakland's skill-position players this decade, Heyward-Bey seems primed to join the Lewis/Nash/Woods side of the equation.
Predicted: 73 receptions, 913 receiving yards, 8 TD
Actual: 79 receptions, 1157 receiving yards, 8 TD
Only 20 percent of respondents figured that Gates would go above the 1,000-yard mark, but while the target and touchdown totals were about dead-on, a healthy Gates produced his best yards per catch total as a starter, elevating him back into his rightful spot as the best receiving tight end in football.
Predicted: 54 receptions, 635 receiving yards, 5 TD
Actual: 76 receptions, 971 receiving yards, 8 TD
There have been a lot of people around FO, both on staff and amongst the commenters, who have been high on Brent Celek for a long time. I certainly thought he could produce solid numbers as a pro tight end, but as I wrote in last year's book, because of the nature of the Eagles' offense, it was going to be hard for him to get the target total that would be required for him to compile significant raw numbers. I pictured his upside as something close to Chad Lewis, but his figures are way better than Lewis's best year (69 catches for 735 yards and three scores).
Predicted: 74 receptions, 862 receiving yards, 8 TD
Actual: 83 receptions, 867 receiving yards, 6 TD
I had Gonzalez at 75-900-6, which is right in the same area code as both the predicted line and the actual figure. No one ever talks about it, but it's remarkable how healthy Gonzalez has stayed throughout his entire career; he's missed two games in 13 years. His yards per catch fell below 11 yards for the first time since 1998, and they're in a four-year decline, which isn't promising. That could be a sign of declining athletic ability or the nature of the offenses he's played in; there's really no one historical to compare Gonzalez to at this point.
Predicted: 68 receptions, 802 receiving yards, 7 TD
Actual: 60 receptions, 612 receiving yards, 8 TD
Much like Hester, Olsen wasn't able to use his speed because teams didn't concern themselves with the possibility of Cutler getting enough time to throw deep. Unlike Hester, Olsen was forced to spend a fair amount of time blocking, too. Even though he went from seven starts to 15, he only saw 28 more targets. With Mike Martz arriving, rumors of a trade have swirled; the Bears are probably better off dealing him now if Martz can't find a role for him in the offense. His value's only going to go down if he has another mediocre year.
Predicted: 59 receptions, 698 receiving yards, 6 TD
Actual: 51 receptions, 574 receiving yards, 7 TD
In the initial piece on tight ends, I noted the decline tight ends with 50+ catches as a rookie saw in their sophomore season. Of course, the sample size is six, so I don't think it's indicative of anything, but Carlson basically treaded water for a year.
44 comments, Last at 22 Mar 2010, 2:09pm
#1 by C (not verified) // Mar 17, 2010 - 11:37am
Nearly ever WR has = or < predicted receptions with the exception of Welker who did it in less than 16 games. BTW, who the heck predicted 37 catches for ALL ACC Honorable Mention and NFL rookie DHB? Were predicters drunk?
I think optimistic fantasy owners and readers over predicted catch rates because everybody wants to be the guy that "called it" and predicted breakouts because most of the players in this sample exercise were "good" which created a tendency to overbid catch guesses. People get pleasure from picking upsets or "calling" things that are a little bit different than conventional wisdom.
I don't like how Barnwell Chimes in with... I thought Owens would have nowhere near a 14.1 YPR... He actually DID hit 15.... ( Barnwell was wrong but) " if you took out his longest catch he would be below the mark." Yeah, and if my aunt had a $*&# she'd be my uncle. If you just took away the 28 or so passing attempts that all resulted in TD passes, then Johnny the QB wouldn't have any TD passes and his stats wouldn't look good!
#2 by are-tee // Mar 17, 2010 - 11:55am
FWIW, I had submitted the low-end prediction for Heyward-Bay - I think it was 16 receptions. So I was only off by 7.
#13 by Ben Johnson (not verified) // Mar 17, 2010 - 3:19pm
You mean you were only off by 100%
#14 by C (not verified) // Mar 17, 2010 - 4:21pm
Maybe if he took performance enhancing drugs like... Ben Johnson... he'd be correct?
Coincidence? I think NOT.
#4 by Keith (1) (not verified) // Mar 17, 2010 - 12:37pm
Do you ever say anything useful or anything that is not denigrating Bill or anybody else?
My first guess would be no, but maybe you have...once.
#6 by C (not verified) // Mar 17, 2010 - 1:35pm
Is poster drunk?
Guess again Keith (1)
#8 by sundown (not verified) // Mar 17, 2010 - 2:13pm
Things are what they are, but I did find it interesting that a single catch could drop TO's average by a yard-and-a-half for the season. That's rather amazing.
#12 by C (not verified) // Mar 17, 2010 - 2:56pm
Right, but Barnwell said he didn't think Owens would " come close" to 14.1 YPC... Not only did Owens come close, he went over 15 YPC.
Rather than saying he was wrong and the fandom was right... we get an excuse that "if it wasn't for that one catch"... I would have been right!.
I know I predicted the Colts to beat the Saints in the super bowl, but tf you took away a select few of the best plays for the Saints in the super bowl... The Colts shoulda woulda coulda beat them!
#19 by sundown (not verified) // Mar 17, 2010 - 10:32pm
Hey, I'm as critical of him as most anybody but I'll give him slack on this one.
(BTW, it still absolutely amazes me he still hasn't corrected the stats he messed up in the QB Wisdom of Crowd article. It was pointed out numerous times in that thread and even mentioned in the RB thread where he was talking about making some edits there. Absolutely nobody else in FO who let a statistical error remain on the site--for obvious reasons. His way of proving a point of some kind? Just weird.)
#21 by Jeff Fogle // Mar 18, 2010 - 12:12am
Owens longest TD catch of the year going back to 2002...
I think there would be a general assumption from the "crowd" that there would be a long TD in the mix. 98 yards is only 7 larger than his prior peak, and 22-23 larger than a TD in the mid 70's that he's peaked at three times since 2002 (not sure if he had others that were long, just going off his peak as listed on his bio page at nfl.com.
If you pencil in mid 70's...or mid 50's...it's not that big a deal in the big picture I don't think. He wasn't going to bust any long TD's? One option would be to throw out the high and the low from the past season and see what that does to the average.
Sundown, are you the Sundown who posts some in handicapping forums (and enters contests in Vegas under that name)? Or, is this a different Sundown? Nice to see you here if it's the first, nice to make your acquaintance if its the second (lol).
Still hoping somebody will type in KUBIAK numbers. Looks like you have to order the book online to get them (will this year's be available in stores?). Nobody who's ordered the book, and who reads these comments, has felt it appropriate to type in the numbers. Maybe, as Commissioner Leaf said in the other thread, the framework is too different for a comparison. Would be nice if an FO hand stepped in to explain that, or provided an easy-to-read comparison amongst crowd/Kubiak/reality.
Happy March Madness to hoops fans. Big day Thursday...
#41 by Steve Garvin (not verified) // Mar 20, 2010 - 8:52pm
He "chimes in" with that point because 98 yard catches are extremely rare (an outlier) and if you remove that one catch then you may get a more accurate reflection of TO's season. There is not OTHER catch you can remove - from any player last year - that would have as much affect on a player's YPC than that play. So it is a legit point.
He is removing a small piece of Owens' sample - versus your intellectually-insipid example of removing 28 pieces of a sample.
If you dislike Bill's work, fine. But seriously, ranting about that the way you did makes YOU look like the idiot.
#42 by Jeff Fogle // Mar 21, 2010 - 1:16am
Let's make sure we're accurate. There's no OTHER catch you can remove from any player last year?
Stokely: 87 yd TD out of 327 total yards
Kelly: 84 yd TD out of 347
Aiken: 81 yd TD out of 326
Sproles: 81 yd TD out of 497 (as a receiver)
Crayton: 80 yd TD out of 622
Murphy: 75 yd TD out of 521
Jacobs: 74 yd TD out of 184 (as a receiver)
Jones: 74 yd TD out of 440
Henry: 73 yd TD out of 236
Probably others, but I'm running out of gas, lol.
Bill used at least 50 catches as a qualifier elsewhere in the thread. Definitely reduces the list down. But, 98 is such a big number in terms of how it affects percentages that if you use it in that manner AND make it at least 10% at 50 catches for wide receivers...then you basically self-select for Terrell.
Henderson: 9.3% at 51 catches with a 75-yard TD (just under 10%)
Miller: 10.7% at 66 catches with an 86-yard TD, (but he's a tight end not a wide receiver).
A couple of the guys above would qualify at over 10% if the threshold was 30 catches instead of 50. More bigger action guys would qualify it was 7% instead of 10% (using TD's of 70+ yards around 1,000 total).
Think the gripes are coming about the principal of the thing...from the stomps and guts site. A 98-yard TD pass is a stomp catch. Looking for consistency in how to best evaluate performance. Should be a standard philosophy most seem to think.
With or without the play included, ypc is what the crowd did BEST at with TO. Leave it in, leave it out...no conclusions are changed.
#43 by Steve Garvin (not verified) // Mar 21, 2010 - 11:07pm
If the discussion was about all of those players you list above the SAME argument could be made - their YPC was somewhat inflated by their huge play.
To argue that one "self selects" for TO because of the 50 catch caveat is kinda unfair - the point in evaluating him is to ask the questions most relevant to him: Is he the same gamebreaker he used to be? Since his YPC is a robust 15 per, that answer seems to be yes. Take out the one stomp and it lessens - more than it would for other similarly elite gamebreakers.
#44 by Jeff Fogle // Mar 22, 2010 - 2:09pm
I think "somewhat" understates the case. For that list, the big TD had a high percentage impact. That's the point. We agree that the same argument could be made. I was disagreeing with your assertion that there was no "OTHER catch you can remove - from any player last year - that would have as much affect on a player's YPC than that play."
*If you're wondering if a guy is the same gamebreaker he used to be, why would you TAKE OUT his biggest gamebreaking play?
*Why would anyone be wondering if he's the same gamebreaker he used to be when Bill just posted that he only had 5 TD's last year in a full season? The guy turned 36 in December, had his lowest TD output of the decade, and readers are wondering if he's still the same gamebreaker he used to be? Wasn't getting the sense that it was an issue up for debate. The "crowd" had projected his lowest per-game TD total of the decade (If you double the 6 TD's he had in only 7 games because of injury in the 2005 season, you get 13-16-13-9-14-12-13-15-10 this decade). The projection of 8 seems reasonable based on his age, and the weather potential in the latter stages of the season. Bill showed that he underachieved even the lowered expectations.
Predicted: 70 receptions, 986 receiving yards, 8 TD
Actual: 55 receptions, 829 receiving yards, 5 TD
I'm guessing that, out of habit, Bill does yards-per-catch stuff in his head when he sees something like that, and wanted to clarify a point about TO's average. The people who weren't doing averages in their head pretty much got the point I'd think...that TO was well below expectations. Avid fans who knew his career numbers knew that the "crowd" had predicted a decline from past years, and that TO declined even further than expected.
*The "principal of the thing" issue arises again with how you phrased your question. You may not have been here during the Patriots discussions, don't recall seeing your name. Hate to do this exercise again, but it makes the point. FO was arguing in favor of counting the outlier with Pats/Titans, and others were basically making your argument regarding TO:
"The point in evaluating THEM is to ask the questions most relevant: Are they the same LEAGUE POWER they used to be? Since their DVOA is robust, that answer seems to be yes. Take out the one stomp and it lessens - more than it would for other similarly elite LEAGUE POWERS."
That's the point some are making...that FO argued to keep the outlier with the Pats, but then Bill took it out with his brief comment about TO.
And, that's why some stat-types argue for medians. Would be cool to evaluate receivers on their median catch...and on how often they land in the range of that median catch. May become more important as the game continues to evolve. Might be part of why the drive stats here at FO smile on the Colts, but "Expected Wins" has undershot them 8 years in a row turning their final stats into final wins.
#3 by Bobman // Mar 17, 2010 - 12:15pm
Bill, are you cherry-picking?
The WR/TE predictions seem to be so much closer than the QB/RB ones. Freakishly close....
Are they easier to predict? Since receivers depends a lot on QBs and OLs, as well as their WR counterparts, I'd think there would be more variance than RBs (who, of course, also depend on OLs). Looks like I'd be wrong.
#5 by JonFromTX // Mar 17, 2010 - 1:24pm
He's not cherry picking. They didn't do predictions for every WR/TE in the league before the season, I think this is the entire list. The predictors just called it really well.
#10 by sundown (not verified) // Mar 17, 2010 - 2:27pm
How did they decide who to list and who not to? I was curious to see what people thought Owen Daniels would do for the year--he was on a great pace when he got hurt halfway through--but it appears he wasn't included.
#7 by jsdarkman (not verified) // Mar 17, 2010 - 1:53pm
Fitzgerald caught too many short slants because of AZ's poor OL. With a top QB AND a top OL, Fitzgerald could be a 15-18 YPC guy.
#9 by drobviousso // Mar 17, 2010 - 2:22pm
Of course, take out one 98-yard touchdown pass, and Owens falls to 13.5 yards per catch.
While this is a (marginally) useful tool for describing the play by play productivity of a player with a single big play, it is of no use here. You go from comparing apples:apple to apples:apple-flavored-food-product.
#11 by Whatev (not verified) // Mar 17, 2010 - 2:35pm
You know these are the Buffalo Bills we're talking about here, right?
#15 by TheDeuceisLoose (not verified) // Mar 17, 2010 - 6:43pm
Just a quick note on Carlson - he might have had much better stats if Hasselbeck hadn't been injured so early in the season (week 2 if I remember correctly) and the O-line had not been so terrible all year (resulting in Carlson often being kept in to block). In week 1 Carlson was awesome, I had him on my fantasy team so I know this, and the next week Hass was out for 3-4 games (I think?) and Carlson's #'s fell WAY down. Just thought that might be worth mentioning in his defense.
#16 by Still Alive (not verified) // Mar 17, 2010 - 6:58pm
"Of course, take out one 98-yard touchdown pass, and Owens falls to 13.5 yards per catch. "
Ok Peter King/Greg Easterbrook/Fox NFL Production Assistant # 63...
That type of analysis just means absolutely nothing.
#17 by John G (not verified) // Mar 17, 2010 - 8:54pm
There is some point to it: TO got about 12% of his yardage total and 1 of his 5 TDs from that one play.
#18 by countertorque // Mar 17, 2010 - 10:18pm
How many plays did it take him to get the other 4 TD's?
#25 by jebmak // Mar 18, 2010 - 10:29am
#22 by Still Alive (not verified) // Mar 18, 2010 - 12:34am
So are we going to subtract each players best play from his totals? How about his worst play? How about plays vs good or bad players. Why not reduce the tiny sample even more!
#24 by Bobman // Mar 18, 2010 - 2:29am
Now you're thinking like an agent!
#26 by John G (not verified) // Mar 18, 2010 - 10:58am
The point about being "1 of his 5 TDs" was more to emphasize his low total TDs for the season.
And as far as taking away each player's best play....FO has already done a similar thought process when looking at boom-and-bust players. Chris Johnson this year was so good to be an exception, but there have been other RBs analyzed by FO that showed they were actually not very good outside of one or two big plays.
AKA "Success Rate"
#20 by are-tee // Mar 17, 2010 - 10:45pm
Cotchery's dip in production had nothing to do with Braylon Edwards. He was injured, missing two games and playing hurt in another two. That and the fact that the Jets didn't throw the ball nearly as much as they did in previous years.
#23 by Misfit74 // Mar 18, 2010 - 2:19am
That and Sanchez was horrific as a passer. But we wouldn't want to tick off the NY fan-base, now would we?
#27 by Bill Barnwell // Mar 18, 2010 - 11:09am
Terrell Owens' longest catch accounts for 11.8% of his total yardage on the season. That's a disproportionately high figure; in fact, for receivers over 50 catches, only one other player was above ten percent, San Francisco's Josh Morgan (11.6%).
My point was not "I would have been right if only for that one darn catch!". My point was "Terrell Owens had one fluky long play that inflates his yards per catch to a figure that doesn't accurately portray his level of performance."
#28 by bravehoptoad // Mar 18, 2010 - 11:37am
Who says it's fluky? Check post #21. I mean, all good WRs have occasional long plays.
#29 by Joseph // Mar 18, 2010 - 12:51pm
bravehoptoad, here's the fluky part--it came from the TWO yd. line. Sure, many receivers have a long bomb or two per year (or a long catch & run). But if his one long bomb would have come from his 40 (giving him a 60 yd TD vs. a 98 yd TD), his ypc drops 0.7 ypc.
#37 by DeltaWhiskey // Mar 19, 2010 - 5:02am
"his ypc drops 0.7 ypc," which means 14.3 ypc, which is pretty damn close to the Crowd's predicted 14.1 ypc.
#30 by Eastside Rocky (not verified) // Mar 18, 2010 - 1:13pm
Hester didn't play 16 games. His play declined as the year went on because he basically didn't play the last 4 games or was extremely limited.
#31 by Noahrk // Mar 18, 2010 - 1:47pm
I'm with Bill here. Knowing about that play tells us T.O. wasn't as effective as his yards per catch might lead you to believe. Interpreting that data as an attempt on Barnwell's part to appear not to be as wrong is purely subjective: a reader's choice. I, for one, enjoyed learning that fact.
#32 by Eddo // Mar 18, 2010 - 3:14pm
I also find it funny that one of the biggest criticisms leveled at FO during the season was that they had the Patriots rated too highly, and that they really should throw out the 59-0 Titans win, because it was such an outlier. Now, Bill's being criticized for the exact opposite.
#33 by Jeff Fogle // Mar 18, 2010 - 6:53pm
Eddo, so we could change the sentence around a little to read like this:
---My point was "The New England Patriots had one fluky huge DVOA game (about 130% eyeballing it in the chart that ran with the AFC Wildcard previews) that inflates their DVOA to a figure that doesn't accurately portray their level of performance."---
And, that's the point many were making about the Pats through the season. True?
Think we all agree that T.O. didn't live up to most expectations, or justify what the Bills were paying as Bill pointed out. Don't think ypc, whether it's 13.5 of 15.0, would influence those perceptions. The "crowd" was in the neighborhood with ypc, but undershot catches and TD's by a good bit.
(PS: always using medians helps avoid these kinds of discussions, lol).
#34 by Eddo // Mar 18, 2010 - 8:07pm
Fair enough. As long as you have both be true; if you throw out the Patriots' outlier, you have to throw out Owens's, and if you keep Owens's outlier, you have to accept the Patriots' as well.
Personally, I fall on the side of removing the two outliers, but I have no statistical reasoning for that.
You're right, I wonder if median yards-per-[blank] would be a better measure than mean, in almost all cases.
#35 by C (not verified) // Mar 18, 2010 - 9:04pm
TO had a longer LONG Catch this year, but he was playing for the... Buffalo Bills... a team that sucked on offense... He probably had more chances to have 90 yard catches as the Bills were most likely pinned back deeper than the Cowboys, Eagles, and decent 49ers teams Owens played for. Part of having a 98 yard touchdown catch is having a CHANCE to have a 98 yard touchdown catch and having a team that will get the ball at the 2 yard line, and throw the ball from the 2 yard line.
If you take a step back, I think the Hilarious thing is that Bill Barnwell is now the one saying that the "numbers don't tell the whole story".
Isn't it usually the FO saying "trust the numbers", and I'm the one saying... look beyond the numbers, Byron Leftwichs has poor fundamentals... or the New England Patriots just ran their DVOA up against a crappy Kerry Collins Titans team that quit in a blowout win....
Now Barnwell is trying to argue with the absolute-ness of the numbers while other people defend the final numbers.
#36 by Eddo // Mar 18, 2010 - 11:00pm
"Isn't it usually the FO saying 'trust the numbers'"
No. In fact, your history of disdain for Barnwell can be traced back to a prediction (that the Giants would have the worst record in the league in 2007) that was made as part of an article in which the FO writers point out how they disagree with the numbers.
#38 by DeltaWhiskey // Mar 19, 2010 - 5:21am
Why do both have to be true? I would think each should be carefully considered and the merits of each be weighed.
If you're going to throw out TO's 98 yd reception, then what about the two receptions on October 18, 2009, that were for a total of 2 yds?
The problem is, Bill wanted to throw out the 98 yd. reception, but it appears TO had at least 2 receptions that pulled his avg down.
I said it before on the QB article, it's really too bad that on a "stats" based website, there is no true statistical analyis of these results that allow for a true quantification of the Crowd's success/failure. At a minimum, reporting the standard deviations of the Crowd's estimates would have given us a yardstick to measure with. That is, we could say that TO was within -0.2 SD for YPC of the crowds prediction while Welker exceeded the Crowd's prediction by 0.6 SD for catches.
#39 by Eddo // Mar 19, 2010 - 10:42am
You're not wrong. In fact, when I say to throw out the outliers, I also mean low-end outliers. So yes, if you take away Owens's 98-yard catch, you have to also remove his catch for the least yardage.
I like the standard deviation reporting idea, by the way. It would be nice to see that.
#40 by roguerouge // Mar 19, 2010 - 3:53pm
The point of this article was to try out a neat toy and see what happened. It was supposed to be fun. When did this stop being fun?
The only question I have is how many people are in this crowd. Generally, crowdsourcing requires a goodly number of people, so I'm curious if we have thousands of people doing this, hundreds or dozens.