The Big Takeaway
In week 5 last year, I investigated the relationships between preseason expectations, early-season performance, and late-season performance. I found that the correlation between early-season performance and late-season performance was smaller than the correlation between preseason expectations and late-season performance; in other words, I found that players who underperformed or overperformed in the early going tended to regress towards our initial expectations.
If you would like to read the full article, it can be found here. If you’d rather not, here is, in my opinion, the key takeaway:
All told, I looked at 95 players from last season (the top 24 QBs, 36 RBs, 48 WRs, and 24 TEs by preseason ADP, minus everyone who missed more than two games). Of those 95 players, 59 finished games 5-16 ranked closer to their preseason ADP than they did to their game 1-4 ranking. An additional 5 players finished games 5-16 ranked equally distant from both their preseason ADP and their game 1-4 performance. That means just 31 players ranked more similarly in games 5-16 to games 1-4 than to preseason ADP. Where a player's performance in games 1-4 diverged from his preseason ADP, that player was almost twice as likely to finish closer to his preseason ADP than he was to finish closer to how he started.
You’ve probably seen me refer to this article in recent weeks, because I think it’s one of the most important facts for any fantasy owner to remember. It sounds so simple in theory, but when we try to apply it in practice we find so many reasons not to.
In an effort to hopefully reinforce the power of this simple realization, I wanted to devote this week to reflecting back on what happened since I wrote that piece last year.
To walk through the process once again… first, I compiled a list of the top 24 quarterbacks, 36 running backs, 48 wide receivers, and 24 tight ends according to 2013 preseason ADP. These are the players that were generally expected to be fantasy relevant before the season.
After I had those lists, I removed the name of any player who missed more than one of his first four games, or more than two of his final twelves games. The reason for this was simple: if a highly ranked player missed two games early in the season, his early-season production would be terribly decreased, and then his late-season return would be perceived as a “rebound” when in reality it was just him getting healthy.
Likewise, if a player missed more than two games later in the season, his late-season production would be decreased, and his performance would appear as a slump when in reality it was probably just an injury.
Since I believe that we largely cannot predict injuries, I’ve omitted these players from the sample so that we can get a truer sense of what’s happening with the guys who are staying healthy.
Equipped with my player lists, I calculated how many points they scored through four games using PPR scoring. Please note that by “four games”, I do not mean four weeks; for any player that had an early bye, I have added their week five performance so that all players are being compared based on four games worth of production.
Finally, I calculated how much each player scored over the final twelve games of the season, and where that production ranked him at his position.
Last year wound up being much more predictable than years past, whether you used production from the first four games or preseason ADP. In fact, early-season production wound up holding an ever-so-slight edge, with a correlation of 0.655 vs. one of 0.649 for preseason ADP. Both correlations were actually shockingly strong, and they were by far the two best I’ve yet seen.
In total, after accounting for injuries, there were 87 players left in the sample. 9 of the players had a rest-of-year ranking equidistant from both their preseason ranking and their early-season ranking. 38 players finished the year closer to their preseason ADP, while 40 players finished the year closer to their early-season ranking.
What about players who dramatically outperformed or underperformed expectations over the first four weeks? 42 of the 87 players had a ranking over the first four games that was either ten places higher or ten places lower than preseason expectations.
Over that 42-player sample, once again, correlations were very close. This time, preseason ADP held a slight edge, with a correlation of 0.618 vs. a correlation of 0.614 for early-season performance. 23 of the 42 players finished closer to their preseason ADP, while 19 of the 42 finished closer to their early-season performance.
Breaking it down even further… 15 players outperformed their ADP by at least 10 spots over the first four games. For these “overperformers”, their early-season performance actually was a sign of things to come; the correlation between their early performance and their late performance was a very robust 0.740, while the correlation between their ADP and their late performance was a much weaker 0.582. So they clearly performed closer to their early returns
On the other hand, 27 players underperformed their ADP by at least 10 spots over the first four games. For these “underperformers”, it turns out our initial expectations were largely correct; the correlation between their preseason ADP and their rest-of-year production was 0.751— the highest we’ve seen yet— while the correlation between early-season performance and rest-of-year performance was 0.651.
Breaking it Down
If your eyes are starting to gloss over from all of the numbers I’m throwing out, let me boil it down some.
At the very least, preseason ADP does as much to predict performance from week 5 onward as early-season performance does. In 2013, both measures were roughly equally predictive. In past years, preseason performance has actually held more predictive power.
In the case of players who dramatically outperform our expectations over the first four weeks, roughly two thirds continued to outperform expectations going forward, though roughly half of that group still took a decent step back from their hot start.
In the case of players who dramatically underperform our expectations over the first four weeks, roughly two thirds rebounded from their slow start, though roughly half of that group still underperformed preseason expectations going forward, albeit by a smaller amount than they initially had.
Or, to put it another way: based on last year's data, if a player starts hot, there's roughly a 1/3rd chance that he maintains that pace, a 1/3rd chance he regresses a little, and a 1/3rd chance that it proves to be an illusion entirely. If a player starts cold, there's roughly a 1/3rd chance that he rebounds all the way back to preseason expectations, a 1/3rd chance he rebounds part of the way back to preseason expectations, and a 1/3rd chance that he continues struggling just as much for the rest of the season.
Andrew Luck is undeniably talented, but I’ve always held one reservation about his fantasy value. In recent years, the trend has been that the league’s most talented quarterbacks are also its most prolific; historically speaking, this trend is something of an aberration. In the past, the correlation between passing efficiency and number of pass attempts has been much weaker. The possibility has always existed that Andrew Luck would not be featured enough to put up the truly bonkers fantasy seasons we’ve seen from Peyton Manning, Drew Brees, Aaron Rodgers, and Tom Brady in recent years, (note: "bonkers" is the scientific term for the production level of these quarterbacks). Just look at last year, when Indianapolis insisted on featuring a terrible running game and Andrew Luck did not rank among the top 10 in pass attempts.
After a pair of monster games, Andrew Luck seems to be laying those minor concerns to rest in a big way. Through four weeks, Luck ranks 1st in pass attempts and 1st in fantasy points. This will not be the last time in his career where that is the case.
On the topic of quarterbacks, one game is hardly dispositive, but it was nice to see Teddy Bridgewater playing so well in his debut. This next bit shouldn’t be taken as anything other than some fun trivia, but if you play with the numbers, you can make Bridgewater’s debut look impressive, indeed. For instance, if we ignore players who played in other leagues before joining the NFL, then among true rookie quarterbacks who made their debut in the first quarter of the season, only Robert Griffin III, Cam Newton, Jim Zorn, and Fran Tarkenton scored more fantasy points in their first start than Teddy Bridgewater. If we include guys who played in other leagues first, we can add Jim Kelly to that list, too. In NFL history, Bridgewater joins Matt Ryan as the only true rookie quarterback with 30 pass attempts, 10+ yards per attempt, and 0 INTs, (Jeff Garcia joins the duo if we include players who played in other leagues first).
Just to be clear, these are just fun tidbits and examples of how positively you can spin anything if you’re motivated to do so. If I’d set that minimum at 25 pass attempts instead of 30, I would have produced a much less impressive list. But regardless of how I spin it, Bridgewater’s debut cannot be interpreted as anything other than a good thing.
Since the beginning of the 2012 season, here is a complete list of all backs who have averaged more points per game than DeMarco Murray in PPR scoring: Jamaal Charles, Adrian Peterson. Among the top 12 fantasy backs, here is a complete list of all who are younger than Demarco Murray: Le’Veon Bell, Giovani Bernard, Lamar Miller, Knile Davis.
Given that Miller and Davis are only performing like they are because of an injury ahead of them on the depth chart, I think a pretty compelling case could be made that DeMarco Murray is currently the #3 running back in dynasty.
At the same time, I could just as easily make a case for having Murray outside of the top five. He does have a history of nagging injuries and he’s a free agent after the season, after all. LeSean McCoy is the same age, and Jamaal Charles is just one year older. Eddie Lacy is also much younger and has a history of production. Le’Veon Bell and Giovani Bernard seem like a solid choice for the top two backs, but that second tier spanning from third to seventh could really be ordered in any way.
Through four weeks, Antonio Brown is second in the NFL in touchdowns with five. Never bet against regression.
Speaking of Brown, there’s a common sentiment that a receiver needs to be tall to dominate in goal-to-go situations. It’s true that receivers under 200 pounds or 6 feet tall will often put up much higher ratios of yards gained to touchdowns scored. At the same time… from 2009 to 2013, the top five receivers in terms of touchdowns scored per target at the goal line were Dez Bryant, Greg Jennings, Wes Welker, Jeremy Maclin, and Lance Moore. Bryant is 6’2”, but Jennings stands just 5’11”, Welker is 5’9”, Maclin is 6’0”, and Moore is 5’9”. If all else is equal, it’s better for a receiver to be tall, but those short guys can often hold their own surprisingly well even on a short field. Especially when they happen to be playing with a quality quarterback.
Surprising, but true: Martellus Bennett is younger than Jimmy Graham. Rob Gronkowski is younger than Larry Donnell. Make of this information what you will.