In this week's Varsity Numbers, Bill Connelly revisits some measures and concepts: Adjusted Scores, Covariance, and momentum (or whatever you choose to call it).
09 Mar 2012
by Danny Tuccitto
As much as we love our KUBIAK fantasy projection system, there's still plenty of randomness to make the entire endeavor an inexact science. Imagine if your Week 1 starting fantasy football lineup was Michael Vick, Chris Johnson, Peyton Hillis, Santonio Holmes, Anquan Boldin, Dallas Clark, Adam Vinatieri, and Eagles defense/special teams. In early September, you were feeling reasonably comfortable with that lineup. By the time early December rolled around, you had probably been eagerly anticipating your league's Toilet Bowl for over a month.
It's for this very reason that consulting the wisdom of crowds can be so useful in the context of fantasy football. Diverse, independent estimates aggregated across members of a group are often more accurate than the estimate of any individual person. In the case of our Wisdom of Crowds feature, our Twitter following is the group providing the estimates, and I'm the aggregator.
Today, we'll review how accurate the crowd's wisdom was in 2011 with respect to quarterbacks and running backs. Next week, we'll do the same for wide receivers. For each player that was submitted to the crowd, I'll list his projected performance estimate (with error range) and his actual performance. If a player missed games, I'll also list his performance prorated to a 16-game season. (Remember that the Wisdom of Crowds projections were supposed to assume the player was healthy for the season, estimating quality and workload instead of the likelihood of injury.)
Accuracy details are to follow, but I just have to say up front that, as a Wisdom of Crowds rookie this year, I was somewhat surprised at how prescient the projections turned out to be. Of 75 total sets of player statistics, crowd estimates landed within the error range 21 times, and there were at least another 14 near misses. I'm not sure how that compares to previous seasons (since I just took over doing WoC this year), but it's definitely impressive to me.
Projected: 3,358 ± 110 passing yards, 21 ± 1 TDs, 17 ± 1 INTs
Actual: 1,955 passing yards, 9 TDs, 8 INTs
Prorated: 3,475 passing yards, 16 TDs, 14 INTs
Kolb is the first of several players we'll review who would have reached or exceeded his Wisdom of Crowds projection if not for injuries. In Kolb's case, turf toe stole four games in the middle of his season, and an early concussion during Week 14 ended it. The result is a perception that he woefully underachieved during his first (and possibly last) season in Arizona. I wouldn't blame any Kolb owners if they believed that. However, over the nine games he did play, Kolb was on pace to beat his yardage and interception projections, and fall just short of his touchdown projection. If we put aside the one-pass-and-done game against San Francisco, his prorated stat line rises to 3906-18-16. That would have made him the 17th-best fantasy quarterback rather than 30th-best.
Going forward, Kolb's future is unclear. On one hand, he improved his net yards per attempt (yards per attempt after adjusting for sacks and sack yardage) from 5.7 over four years with the Eagles to 6.1 in 2011. In addition, his interception rate fell from 4.4 percent to 3.2 percent, which is much closer to the 3.0-percent league average over the past few years. On the other hand, one of his major limitations is that he takes too many sacks (10.4-percent sack rate in 2011; 8.0 percent career), which doesn't bode well for a quarterback who has suffered two concussions in two seasons. There's also, of course, the looming possibility that the Cardinals will jettison him in favor of Peyton Manning.
Projected: 2,937 ± 141 passing yards, 17 ± 1 TDs, 18 ± 2 INTs
Actual: 3,091 passing yards, 14 TDs, 13 INTs
Prorated: 3,297 passing yards, 14 TDs, 13 INTs
Speaking of possible Manning destinations, Jackson ended up throwing for more yards and fewer interceptions than the crowd projected, although he did undershoot his touchdown target. That last outcome was somewhat surprising. It's likely the result of a revolving-door receiving corps, the Seahawks not being as bad a team as anticipated, and Marshawn Lynch exceeding expectations -- more on him later.
The future for Jackson doesn't appear to be as bright in Seattle as it could be for Kolb in Arizona. He did bring his interception rate down from 3.6 percent with Minnesota to a league-average 2.9 percent with Seattle. However, he also had an 8.5 percent sack rate -- up from 7.2 percent in Minnesota -- and ended up tied with Blaine Gabbert for the league lead in sack yards (243). Not surprisingly, then, he only threw for 5.7 net yards per attempt, which was identical to what he posted during his time with the Vikings. Essentially, Jackson appears to be a known (mediocre) quantity at this point, so it makes more sense for Seattle to take a flier on an unknown quantity like Manning than it does for Arizona.
Projected: 3,305 ± 130 passing yards, 20 ± 1 TDs, 14 ± 1 INTs
Actual: 1,026 passing yards, 4 TDs, 2 INTs
Prorated: 2,736 passing yards, 10 TDs, 5 INTs
On the positive side, there's McNabb's 2.0 touchdown-to-interception ratio in 2011. On the negative side, there's pretty much everything else. Not that McNabb has much of a future in the NFL at this point, but I'll note that his 5.3 net yards per attempt this season was the lowest since his first full season as a starter in 2000. As we often say with aging stars, their careers tend to end abruptly and with little warning. It's tough to predict when their production will nosedive, but, when it happens, retirement is soon to follow.
One lesson fantasy owners can take away from the McNabb saga is that "likelihood of being benched" is a factor that should never be underestimated when projecting 30-something starting quarterbacks on bad teams, especially when there's a youngster waiting in the wings. I mean, in retrospect, everyone should have seen this one coming, right? Potential drafters of Matt Hasselbeck next season should be aware of this.
Projected: 2,740 ± 131 passing yards, 14 ± 1 TDs, 19 ± 2 INTs
Actual: 3,398 passing yards, 20 TDs, 13 INTs
Whereas the Lewin Career Forecast (LCF) v2.0 was bullish on Dalton over a five-year time horizon, and KUBIAK viewed him more favorably than most projection systems in 2011, the crowd's enthusiasm was lukewarm at best. Given the problems rookie quarterbacks have historically had in their first season, it's no real surprise that the crowd didn't back Dalton. If only our countervailing evidence back then was as strong as it is now.
By all measures, Dalton exceeded expectations last season. With respect to the crowd, he easily beat his yardage, touchdown, and interception projections. KUBIAK predicted 181 fantasy points, and he ended up scoring 211. To boot, the 807 DYAR he produced in his rookie year bodes well for the 1,616 DYAR that LCF projects in years three to five. However, it must be said that both Sam Bradford and fellow-LCF-darling Colt McCoy regressed in their sophomore NFL seasons, so fantasy owners should remain somewhat cautious about Dalton heading into 2012. Of course, neither of those guys had a receiver remotely as talented as A.J. Green -- more on him next week -- running routes for them.
Projected: 3,708 ± 113 passing yards, 25 ± 1 TDs, 13 ± 1 INTs
Actual: 3,592 passing yards, 16 TDs, 22 INTs
Prorated: 3,831 passing yards, 17 TDs, 23 INTs
When you look at Freeman's 2011 season, there's a clear dichotomy between what his fantasy point total says, and what his advanced metrics say. His net yards per attempt dropped from 6.5 in 2010 to 5.9 last year, his DVOA ranking fell from 10th to 29th, and his DYAR ranking dropped from ninth to 28th. In fact, he was 901 DYAR less valuable in 2011 despite having 107 more pass plays than 2010.
On the flip side, though, his prorated stat line still ranked him 14th in fantasy points among quarterbacks after ranking seventh in 2010. Not to mention that the crowd basically nailed Freeman's yardage total.
It's no mystery what happened to Freeman: lots and lots and lots of interceptions. More precisely, his interception rate skyrocketed from an unsustainable 1.3 percent to a higher-than-league-average 4.0 percent.
Projected: 3,359 ± 162 passing yards, 19 ± 2 TDs, 13 ± 2 INTs
Actual: 2,733 passing yards, 14 TDs, 11 INTs
Prorated: 3,363 passing yards, 17 TDs, 13 INTs
From the afterlife, Sir Francis Galton tips his hat to everyone who participated in McCoy's projection. While putting this piece together, I wasn't expecting to see any of these eerily clairvoyant results, so I'm just too weirded out to offer anything in the way of analysis. Let's just move on to running backs.
Projected: 249 ± 22 carries, 985 ± 98 yards, 7 ± 2 TDs
Actual: 37 carries, 179 yards, 0 TDs
Prorated: 74 carries, 358 yards, 0 TDs
If we could only go back in time to that fateful August day, and tweet "Willis McGahee" instead of "Knowshon Moreno," the crowd's projection would have been McCoy-esque. Then again, maybe tempting fate like that would have just led to McGahee pulling a hamstring in Week 1 and tearing his ACL in Week 10. The Gods do have a sick sense of humor sometimes.
To put Moreno's underachievement into perspective, consider that 249 projected carries translates to 15.6 per game. Moreno didn't have a single game with that many carries, and only had double-digit carries once. Despite playing eight games, he didn't even reach the worst-case scenario projection of 150 carries, 600 yards, and 3 TDs.
Note to self: When a smallish running back has a bad speed score (96.9), an injury history, and a backfield mate likely to vulture carries and touchdowns, stay away from him in your fantasy draft.
Projected: 233 ± 14 carries, 1,005 ± 68 yards, 7 ± 1 TDs
Actual: 127 carries, 575 yards, 1 TDs
Prorated: 169 carries, 766 yards, 1 TDs
Jones' 2011 season provided further evidence busting the myth that a starter can't lose his job to injury -- not that any more was necessary, of course. With that said, mitigating circumstances make this projection less awful than it seems. First, at the time we tweeted Jones' name, DeMarco Murray was known only as a rookie with an injured hamstring. Granted, his 112.6 speed score suggested the existence of athletic talent, but there was little reason in August to think Jones would have major competition for carries in 2011. Second, after Murray broke his leg, a fully healed Jones had consecutive 100-yard rushing days, so it's not like he was a monumental fantasy bust, ala Moreno.
The Cowboys have said that Murray is their feature back going forward, and Jones will be relegated to third-down duty. As we said back in August, Jason Garrett is a fan of the workhorse, so Dallas' proclamation is probably not media shenanigans. With his performance after Murray's injury, and the fact his yards per carry actually increased in an otherwise forgettable year, Jones is well within the borders of handcuff territory for 2012.
Projected: 223 ± 20 carries, 901 ± 84 yards, 7 ± 2 TDs
Actual: 245 carries, 1,047 yards, 10 TDs
Prorated: 261 carries, 1,116 yards, 10 TDs
Across the board, the crowd was pretty darn close on this one. In August, I was skeptical because Wells was an injury-prone underachiever to that point, and the Cardinals had just drafted a potential replacement, Ryan Williams, in the second round. Naturally, the day after I wrote that, Williams tore up his knee and was lost for the entire 2011 season. With backs, opportunity counts as much as anything, and Wells took advantage of his.
There is one mounting concern about Wells' future: it resides between his right thigh and right shin. For the better part of two years, he's been running on a balky right knee that required a second surgery earlier this offseason. Therefore, I'm still not sold on Wells; even less so as a long-term option in keeper formats.
Projected: 209 ± 16 carries, 903 ± 74 yards, 8 ± 2 TDs
Actual: 122 carries, 474 yards, 5 TDs
Prorated: 195 carries, 758 yards, 8 TDs
You guys came about 70 yards short of nailing this projection too. To me, what's amazing is that the projection was this accurate despite the enigmatic nature of New Orleans' backfield rotation. Darren Sproles' touches were a given on a weekly basis, but touches for Ingram, Pierre Thomas, and Chris Ivory seemed to shuffle in and out of some unseen dimensional portal. With all four backs under contract as of right now, the uncertainty remains heading into 2012.
One little aside before moving on: Has anyone else noticed that Saints running backs over the past few seasons have had a rash of severe lower-body injuries, all of which were suffered in the Superdome? In 2007, Deuce McAllister tore his ACL at home against Tennessee. In 2010, Ivory suffered a Lisfranc injury at home against Tampa Bay, and Thomas missed 10 games due to a high ankle sprain suffered at home against Atlanta. Then, this past season, Ingram injured his heel at home against Indianapolis, and then was lost for the final five games because of turf toe that flared up in a home game against Detroit. Anyone have any theories?
Projected: 220 ± 28 carries, 901 ± 110 yards, 7 ± 2 TDs
Actual: 222 carries, 1,091 yards, 6 TDs
Prorated: 253 carries, 1,246 yards, 6 TDs
OK, so you guys are obviously pretty good at this. Yardage and touchdown projections were spot on, with a slight underestimation of Mathews' yardage total. Therefore, special congratulations go out to the lone Twitter follower who predicted 1,250 yards, which started out as the best-case scenario projection, and ended up being the most accurate.
Based on what we've seen of Mathews so far, there are a few things we know with reasonable certainty going forward. No. 1: His 111.2 speed score at the Combine and 4.9 yards per carry in 2011 -- up from 4.3 in 2010 -- suggest he's the real deal. No. 2: He has elite fantasy potential if touchdown vulture Mike Tolbert ends up leaving San Diego in free agency as expected. No. 3: Disregard No. 1 and No. 2 if he continues to lack durability.
Projected: 245 ± 26 carries, 1,010 ± 100 yards, 9 ± 2 TDs
Actual: 84 carries, 321 yards, 1 TDs
Prorated: 268 carries, 1,027 yards, 3 TDs
Hey! Whaddyaknow, another bullseye for the dart throwers! If Hightower doesn't wreck his knee in midseason, his season ends up on-target for projected carries and yardage.
Speaking of that knee, NFL.com fantasy guy, Michael Fabiano, posted some research recently, which showed that running backs return to pre-ACL-tear production faster if they're younger and suffer the injury earlier in the season. Hightower's injury came at 25 years old, and occurred in Week 7 of 2011, so he has a good chance to be productive in 2012. To me, that's a buy-low fantasy opportunity.
It's also potentially a buy-low for NFL teams because Hightower will become an unrestricted free agent on March 13th. I know, I know. Football Outsiders is an industry leader in disdain for free agent running back signings. However, what we're actually against is signing older free agent backs to inflated contracts. That doesn't apply here, especially given the likelihood of a bear market for Hightower post-injury.
Projected: 246 ± 36 carries, 900 ± 118 yards, 7 ± 2 TDs
Actual: 285 carries, 1,204 yards, 12 TDs
Prorated: 304 carries, 1,284 yards, 12 TDs
Speaking of that running-back-free-agent caveat, Lynch just signed a well-deserved, long-term deal with Seattle after a 2011 season in which he exceeded everyone's expectations, including yours. After consecutive seasons of sub-4.0 performance, Lynch ran for 4.2 yards per carry this past season, up over half-a-yard from 2010. What's even more impressive is that he accomplished this despite playing behind an offensive line that ranked 28th in ALY, 29th in power success rate, and 32nd in stuffed rate. Seattle also ranked 11th in open-field yards, which suggests that their running game was a one-man gang. If the Seahawks do anything to improve their offensive line, Lynch, who will only be 26 years old in 2012, could have big things ahead of him.
That's all for this week. Next week, we'll review the Wisdom of Crowds projections for wide receivers.
32 comments, Last at 15 Mar 2012, 2:25pm by Noah of Arkadia