The Curious Case of the NFL MVP

The 2015 NFL regular season has three weeks left. About a month after the season winds to a close, the NFL will announce the winner of the NFL MVP: Cam Newton. Unfortunately, this choice will be the wrong one.

Let me note a few things: First, MVP is an ill-defined award. Is it the most valuable player, as in the player with the most delta-wins? Presumably--it's clearly not the best player relative to position¹. But how are delta-wins defined? If I take an average quarterback off a team with a very bad backup (or no backup at all), they might experience a greater loss than taking a great quarterback off a team that has an extremely promising backup (e.g. Aaron Rodgers or Steve Young). Does that mean the average quarterback is better than the great quarterback? Of course not, but he's technically more valuable.

We could resolve this problem (to an extent) by establishing a "replacement level" for a backup quarterback and ignoring the real-life backup situation on any given team. Thus we can expect the difference in play between the elite QB and the replacement-level QB, and the difference between the average QB and the replacement-level QB, to accurately reflect the two starters' relative talent gap.

Of course this brings in other complications: A great quarterback playing on a talented team is necessarily less valuable than a slightly worse quarterback playing on a very bad team (even assuming both backups are equally good). If both starting quarterbacks were to get injured, the talented team would have a much better chance to succeed than the bad team, simply because of the talent of the other players. This is true even if the talent in question had no impact on the QB's performance².

And yet, even though a quarterback on a great team has inherently less value than one on a bad team, the MVP award voters tend to vote much more heavily for quarterbacks on talented teams than those on worse teams. This is because the voters imagine that a very good QB can "carry" his team to the top of the league each year. Of course this is false; Drew Brees is by any measure an elite QB and his teams have fluctuated between mediocrity and eliteness over his career. If a team is at the top of the league, it is fair to say that the team has a lot of talent at multiple positions, not just QB. Similarly, it's obvious that a team can be very successful without having a good QB at all; does no one remember the Jets' back-to-back runs to the AFC championship, while being quarterbacked by the immortal Mark Sanchez?

That being said, I'll concede that the majority of very good teams in any given year probably have very good quarterbacks. That's true this year; of the top 10 QBs by Passer Rating, only one (Brees) has a losing record, and six have won at least eight games already this season. Meanwhile of the bottom ten, all but Teddy Bridgewater and Peyton Manning are playing on losing teams.

The correlation between Passer Rating and team wins is no surprise; it's been shown many times that Passer Rating Differential (i.e. team's passer rating minus opponent's passer rating) is a very good predictor of success³. Basically, if your passing game is much more efficient than your opponent's, you have a really good chance of winning. And since a quarterback has no control over his team's defensive efficiency (see the second footnote again if you're unsure about this), the best thing a QB can do for his team is pass efficiently.

The sports media has some basic understanding of this concept; of the 25 QB seasons that have won MVP since 1980, 14 of them led the league in passer rating⁴. Of course, this means that 11 didn't. Some of these are reasonable--Jim Harbaugh (yeah, that Jim Harbaugh) led the league in passer rating in '95, while MVP Brett Favre finished a very close second in PR while throwing for substantially higher volume⁵, playing on a better team, and being a much higher-profile player. Some are mistakes, but are at least understandable mistakes--there's only so many MVPs you can give to Steve Young (although this limit doesn't seem to exist for Peyton). And some are huge, obvious, how-do-you-mess-up-that-bad mistakes, like 1987 John Elway (scroll down to Part IV).

So here's where we stand: It's reasonable to focus on the quarterbacks of good teams for our MVP discussions, and the best way to evaluate a QB's contributions to his team's success is through his passer rating. So how do this year's passer stack up?

#1 is Russell Wilson, who deserves MVP consideration but won't get it because the media mistakenly thinks he was bad weeks 1-8 (he had a passer rating of 95.0 to Cam's 78.1; more on this in a bit).

#2 is Carson Palmer, who isn't getting as much MVP consideration as he deserves, and I have no idea why. The Cardinals are at least as good a team as the Panthers, and Palmer is playing far better than Newton.

#3 is Andy Dalton, whose injury will probably remove him from MVP consideration, but who also very much deserves (deserved?) to be in the conversation.

#4 is Tom Brady, who led the conversation until recently, because voters have short memories and don't realize that Brady's early-season success isn't any less valuable than Cam's late-season success (and Brady's been better across the whole season).

#5-8 will not get consideration, which is reasonable; if you're this far down the list, odds are someone above you has been playing better.

#9 is Cam Newton.

###

Q-and-A. Let's go.

Q: But running.
A: Cam is, of course, a very effective runner. But how much value does his running add to his team, and--more importantly--is it enough to make up for his lackluster passing? FootballOutsiders, as of this week, has Cam ranked 15th and 16th respectively in DVOA and DYAR⁶, their two big metrics. His rushing production is also tied for first, with Wilson, among quarterbacks.

If we add all the quarterbacks' rushing contributions to their passing numbers, Cam moves up by an appreciable amount--the value of his rushing is about 25% the value of his passing. But given that his passing wasn't all that valuable to begin with, the result is uninspiring: he still ranks only 15th in DYAR, jumping only Matt Stafford. (I can't speculate as to where his DVOA would end up, but given that his rushing DVOA is only 23rd in the league, I doubt it would move up that much.) Meanwhile Wilson moves into 4th place, jumping Roethlisberger and trailing only Palmer, Brady, and Dalton.

Now, it may be fair to say that not all the value Cam's legs bring to the table is encapsulated by DVOA/DYAR. And yes, the touchdowns are nice, although I'm somewhat unimpressed by short-yardage rushing scores (any QB can do that, but most just prefer to hand off; the fact that it's Cam running it and not Mike Tolbert doesn't actually add much value). But the idea that Cam's rushing is so much more valuable than anyone else's that it overcomes the massive gap in passing efficiency is absurd to me.


Q: Cam's playing with bad, drop-happy receivers. Ted Ginn Jr. is his #1, for crying out loud. And didn't you see his numbers throwing to all his receivers besides Ginn and Olsen? Clearly, he's being held back by his receivers.
A: It's true that the Panthers have a worse-than average drop rate, but it's not as bad as Panthers fans make it out to be (and notably better than the Patriots'). Their rate of 4.8% is about 1.1% worse than the average of 3.7%. That's a difference of about four passes of Cam's that got dropped over the course of the entire season. Four passes. Is that really enough to make up the difference in efficiency?⁷

It's also worth noting that the Panthers' high drop rate is almost exclusively due to Ginn, who is tied for second in the league in drops and is 12th in the league in drop percentage (among receivers with 10+ receptions). The Panthers' second most drop-happy receiver is Greg Olson, who is tied for 38th in the league and has dropped only 3 of the 108 balls thrown at him. That's about as much consistency as you can ask for.

As for the stat about Cam's numbers when not throwing to Ginn or Olsen: Newton overwhelmingly targets those two receivers. They have 80 and 108 targets, respectively; no one else has more than 50. In fact, nearly half of Newton's attempts, completions, passing yardage, and touchdowns come from those two players. It shouldn't be surprising that opposing defenses hone in on them. Besides, if Ginn and Olsen are so terrible and his other receivers are so much better, why in the world doesn't Cam simply throw to them? Answer: Because Ginn and Olsen are very good receivers. Ginn is a huge deep threat, and his 17.4 yards per reception ranks 6th in the league. And Olsen's 65 receptions are t-3rd in the league for tight ends, and his 969 receiving yards are second among TEs and 13th overall.


Q: The Panthers are winning games. Cam's the quarterback of the best team in the NFL. Why is this so hard to understand?
A: First of all, there is some disagreement on the point of the Panthers being the best team in the league. But I won't bother arguing that here. Second, and more to the point, we've already established that a team can be successful despite having a non-elite quarterback, or mediocre despite having an elite one. Cam doesn't get the award just for playing on the best team. Yes, the only time we saw a 16-0 season it was (partially) because Tom Brady was having an all-time great year. That doesn't necessarily mean that every 16-0 season must feature a QB having an all-time great year (obviously; Cam isn't, and the Panthers definitely might go 16-0).

Third, every quarterback I've talked about so far is playing for a Super Bowl contender. The fact that the Panthers have a couple more wins (or, in the case of the Seahawks, quite a few more wins) than the rest of the teams isn't necessarily attributable to Cam's performance. Which is also obvious, because Cam is barely a top-10 passer this year⁸ (and maybe not even that, if you trust FootballOutsiders, which you should). If he's not carrying the Panthers with his arm, and he's not doing it with his running, then what exactly is he doing so much better than every other quarterback to allow his team to go undefeated?


Q: Cam's been more consistent than Wilson, or anyone else, this year. That's why the Panthers have gone undefeated. And that's why he should win MVP.
A: Good point. Except it's completely wrong. Over the first eight weeks of the season, Cam put up a passer rating of 78.1, to go with his 1523 passing yards, 11 touchdowns, and 8 interceptions. The only impressive thing here is that the Panthers were able to go undefeated over that stretch despite Cam's mediocre play. (Notice how much better the Seahawks started playing when Wilson improved his game. That's value.)

Wilson, on the other hand, put up a 95.0 passer rating over the first eight weeks of the season. He threw for 1878 yards, 9 TDs, and 6 interceptions. And although Cam had more rushing touchdowns⁹, Wilson had more rushing yards and rush yards per attempt. But really, if this doesn't convince you that Cam didn't carry his team over the first eight weeks of the season, nothing will.

###

If at this stage you're still on the Cam-wagon, there's probably nothing I can say to convince you otherwise. And as I said to begin the article, unless something dramatic happens, Cam Newton will win the MVP. He just won't deserve it.

---

1) Ideally, MVP should be the best player relative to their position. An MVP which can only be won by quarterbacks is a stupid award (unless it's like the Cy Young and there's also a separate MVP award. Some people say that's what OPOY does in the NFL, but OPOY is also massively biased towards QBs and RBs, going to a player at any other position twice [both times it was Jerry Rice]). But if we're going to be literal, MVP should have gone to a QB every year of the modern era. Which, of course, it also hasn't; Alan Page, Lawrence Taylor, Mark Moseley, and 18 RBs have won the award. And while Page and Taylor are maybe the two greatest defensive players ever, they still weren't as valuable as a great quarterback (although to be perfectly honest Page might have been). (I'm pretty sure Moseley won as a joke.) As for the RBs: the value of an elite RB is massively overstated in the league today. First of all, the vast majority of RBs, including and especially elite ones, are a product of their offensive lines; look at Priest Holmes, Larry Johnson, Shaun Alexander, DeMarco Murray, Emmitt Smith, and every other running back not named Barry Sanders or Marshawn Lynch. And second of all, an elite running back on a bad team can't actually carry the team (despite what the 2012 MVP voters thought). Of the seven players who have rushed for 2000 yards in a season, exactly one of them was on a team that won more than 10 games (Terrell Davis, on the '98 Broncos). Meanwhile Tom Brady and Peyton Manning have a combined three seasons of single-digit wins (out of 29). That's the value of a QB versus a running back.

2) A quarterback's performance is tied, to some extent, to the quality of his receiving talent, and to the quality of his offensive line. All available data shows that a quarterback's performance is not related to the performance of his running game or defense, despite common belief to the contrary.

3) 36% of NFL champions (since 1940) finished first in PRD, 60% finished top three, and 95% finished top ten. It's kind of amazing that this holds true even back in the days before passing was a big part of the NFL, but it does.

4) In reverse chronological order: '11 Rodgers, '10 Brady, '07 Brady, '04 Manning, '03 McNair, '01 and '99 Warner, '94 and '92 Young, '89 Montana, '88 Esiason, '84 Marino, '81 Anderson, and '80 Sipe.

5) I'll be the first person to tell you that volume is not the be-all, end-all of passing stats. In fact, it's usually massively overrated (see here and here). The ONE exception is that if two players are playing at comparably high levels of efficiency and one has significantly more volume, it's reasonable to say that he's contributed more overall value to his team. This does NOT apply if there's a big gap in efficiency or if the passers in question have low efficiency.

6) Essentially, DVOA is per-play value added, while DYAR is total value added. But--and this is important--DYAR is still based on efficiency. Higher volume with less efficiency produces a lower DYAR, while relatively small volume with very high efficiency can produce a high DYAR. This is also why high volume at low efficiency is not a good thing, but high volume at high efficiency is (see footnote 5).

7) Of course not.

8) In fact, the only passing stats in which Cam ranks top-5 on the season are TDs, TD%, and Yards per completion, all of which, I'd argue, are due at least in part to--are you ready?--Ted Ginn. Having a deep threat like Ginn improves your performance in exactly those areas (as well as the yards-per-attempt stats, where Cam is top-10 but not top-5). Maybe that's the thread linking the '07 Patriots and the '15 Panthers! I have goosebumps. (Just kidding.)

9) For the record, I'm not at all convinced that Cam's rushing touchdowns are all that valuable. The Panthers have one of the best short-yardage backs in the league in Mike Tolbert and one of the better surviving (non-injured) RBs in the league in Jonathan Stewart. Is running the ball at the goal-line with Newton really that much better than running with Tolbert or Stewart? Surely the benefit isn't big enough to overcome Cam's weaknesses. This all stems from the idea of duplicability, which asks the question: "If Cam didn't rush for that touchdown, would the Panthers have gotten it?" In most cases, I think the answer is yes. Whereas if you ask the analogous question about Wilson ("If Wilson didn't make that throw, would the Seahawks have scored?"), I suspect the answer would be no in almost every case. This is actually a great way to assess value, but unfortunately it's almost impossible to do in the NFL.