clock menu more-arrow no yes mobile

Filed under:

So you think you're advanced: It's time to DYAR homework

In part two of this advanced stats series, we take a closer look at DYAR and why it's a much better measurement for Russell Wilson than simple yardage totals.

Kaep: "I had a higher DYAR than you" Wilson: "Oh yeah? I'm goin to the Super Bowl, dumb dumb"
Kaep: "I had a higher DYAR than you" Wilson: "Oh yeah? I'm goin to the Super Bowl, dumb dumb"
Kyle Terada-USA TODAY Sports

In preparation for another season in which we take an advanced look at the best team in the NFL (and for the first time, covering the defending champs), I'm doing something that I did not do last season during my Advanced Stats series:

Explaining what Advanced Stats are and mean.

In part one, I showed what DVOA was, how it's calculated, and why it's important to understand when it comes to discussing the Seattle Seahawks and other teams with your friends*. And then after I got done doing that, Aaron Schatz, the founder of Football Outsiders and a creator of DVOA, told me where I went wrong and so I re-told that same story but correctly. It's important to put ego aside when it comes to advanced stats (or any discussion that may be a little out of your element, really) and make sure that at the end of the day you're getting it right.

*If you're talking about teams besides the Seahawks, maybe they're not really your "friends."

I am not an advanced stats expert. I am an advanced stats novice and lover that's spent a couple of years trying to understand the numbers to a greater degree. That's exactly why I do posts like these, because what better way is there to learn something than to write an article about it and then have the source tell you where you fucked up?

If you turn your books back to chapter one, you'll remember that DVOA stands for Defense-adjusted Value Over Average. I think it's fair to say that one of the main cruxes of any advanced stat in any sport is that it computes for average of the era. It's impossible to compare Jim Brown to LeSean McCoy with any meaning if you're just comparing the raw data, but it becomes more meaningful when you look at how those two players -- or any two players -- did against the league when compared to other players of the same time period.

Take Babe Ruth for example.

If you look at Ruth's career statistics simply for what they are, you're still likely looking at the greatest baseball player to ever live. A career .342 hitter with 714 career home runs that is first all-time in career slugging percentage, OPS, and he was a heck of a pitcher too. However, it becomes even stupider when you look at his numbers in context. Ruth led the majors in home runs in 1918 when he hit 11 of them in 95 games with the Boston Red Sox.

The rest of his team hit four home runs combined, and they won the World Series that season.

The following year, Ruth hit 29 home runs and again, the rest of the Red Sox his four combined. In 1920, his first with the New York Yankees, Ruth hit 54 home runs, which was nearly three times as many as second-place George Sisler, who hit 19.

Now, hitting home runs wasn't considered to be a "good play" until Ruth came along and changed the game, but he was so much better at it than everybody else (and continues to be perhaps better at it than anyone that's played since, including Barry Bonds) that it's truly unfathomable to consider the context. If a player came into the major leagues today and hit 2.75 times as many home runs as second place, then that player would have hit 121 home runs in 2013.

If David Ortiz, the team leader in home runs on the Red Sox in 2013, had hit 7.25x as many home runs as the rest of his team combined, he would have hit 1,073 home runs last season. Now, I'm no "expert*" but that seems like a lot to me.

*jk, I'm an expert and I endorse that it is a lot

Just as DVOA compares players and teams against the average, the stat known as DYAR -- Defense-adjusted Yards Above Replacement -- aims to do the same but strictly on a level of trying to measure the difference, in "yardage," between one player and what a replacement-level player would do.

In baseball, we have "WAR" (Wins Above Replacement) and nobody in history has more of them than Ruth does. Though it's accuracy and value have been argued for and against for years, WAR does give us the opportunity to have a baseline understanding of the value a single player can bring to the team. The reason that DYAR will never quite be able to do that?

Because unlike baseball, football is incredibly reliant upon your teammates and defense. Whereas there are certainly moments in baseball games where a great defensive play can steal some value from a hitter, a home run is still a home run. The outfielder probably had no chance to remove that home run outcome. There are not very many "home run" touchdowns in football that don't come with:

A) At least one other player on your team assisting in the score

B) A total defensive collapse

What DYAR attempts to do instead is simply take a player, like Russell Wilson for example, and try to measure how much better he is than a replacement level quarterback. What is "replacement level"?

For quarterbacks, we analyzed situations where two or more quarterbacks had played meaningful snaps for a team in the same season, then compared the overall DVOA of the original starters to the overall DVOA of the replacements. We did not include situations where the backup was actually a top prospect waiting his turn on the bench, since a first-round pick is by no means a "replacement-level" player.

At other positions, there is no easy way to separate players into "starters" and "replacements," since unlike at quarterback, being the starter doesn't make you the only guy who gets in the game. Instead, we used a simpler method, ranking players at each position in each season by attempts. The players who made up the final 10 percent of passes or runs were split out as "replacement players" and then compared to the players making up the other 90 percent of plays at that position. This took care of the fact that not every non-starter at running back or wide receiver is a freely available talent.

If you can figure out the expected outcome of a play by a replacement-level player, then you can measure the actual player against the figure and have a difference. In DVOA, "success value" is the measurement to determine whether the outcome of a play was good, not good, or just as it should be:

Every single play run in the NFL gets a "success value" based on this system, and then that number gets compared to the average success values of plays in similar situations for all players, adjusted for a number of variables.

Since people would immediately scoff at a stat that said "Russell Wilson had 101 success value points this year!" Football Outsiders simply translates those points into a yardage total and it becomes.... Defense-adjusted Yards Above Replacement.

Last season, Wilson finished ninth in DYAR, with a final total of 770 Defense-Adjusted Yards Above Replacement. It's a counting stat, but it is not just about yards. Whereas actual passing yardage stats are misleading, a players DYAR is only misleading if you think it's only measured on yards. Wilson passes less than almost every other starting quarterback in the NFL, but his efficiency is better than almost every other quarterback in the NFL. DYAR calculates that Wilson had 770 DYAR last year, but it doesn't mean that a replacement level quarterback in Wilson's position on the Seahawks would have simply thrown for 770 fewer yards.

Wilson was just 16th in passing yards, and that's for a guy that started every game of the season. Only Colin Kaepernick and Geno Smith started all 16 games and had fewer passing yards. So if DYAR was simply a yardage measurement, the replacement player would've had just 161.6 passing yards per game. In reality, a replacement level player would probably pass for more yards than Wilson, but do far less with it. He would have more pass attempts, fewer touchdowns, more interceptions. Take Ryan Tannehill for example, a player that passed it 181 more times than Wilson, but had two fewer touchdowns and eight more interceptions, with 1.5 fewer yards per atttempt.

Tannehill finished with just 54 DYAR, putting him fairly close to replacement despite being one of the more prolific passers in the NFL.

Notable players to finish in the negative last year were: Robert Griffin, Matt Schaub, EJ Manuel, Joe Flacco, Eli Manning, Geno Smith, and Terrelle Pryor.

Brandon Weeden was last, finished 443 "yards" below replacement. He would've been rewarded with a free dinner at Red Lobster, but he already gets a senior discount.

(See, here's the thing about Weeden: He old. Lol. I mean, he's younger than me though.... oh God, he is younger than me. What have I done with my life?!)

Of the players to finish with negative DYAR last season, the only ones to still be starters with the teams they were with last season are Griffin, Manuel, Flacco, and Manning. Christian Ponder, Case Keenum, Chad Henne may get to be in "competitions" with their own teams, but only for posterity's sake. Meanwhile, Pryor and Schaub were traded, while the Jets had to go out and get Michael Vick in order to bench Smith.

It was only a year ago that Mark Sanchez finished last in DYAR for New York, giving them a pretty good track record in this category. In fact, the total MetLife DYAR for home starters last season was -706 between Manning and Smith. (Unless you're also talking about the Super Bowl, and if you're not counting Manning's sad sack brother as the "home" starter. That other guy though, he did pretty good*.)

*And I'm not talking about the sarcastic "pretty good" that I used earlier in the paragraph. I'm talking about the subdued "pretty good" in which the guy was actually pretty great. If this wasn't already in italics, I would've italicized "great" to emphasize how great he is.

I don't think I'll ever get a writing job for a national publication with garbage like this. Though, I mean, Field Gulls can be seen all over the world so technically I'm writing for an international publication. I wonder if guys in space get the internet, in which case, I'm writing for the solar system, if not the universe. At least the galaxy. Oh God, just stop.

Since DYAR is a counting stat though, it's important to note which players were even more "impressive" than those players already mentioned. Take Kirk Cousins, for instance.

Noted as Wilson's "equal" around the time of the draft, if not a better prospect in the minds of many, Cousins had just 48 pass attempts as a rookie in 2012, but was worth 59 DYAR. That's a figure than what fellow rookie Tannehill posted, and he had 10 times as many pass attempts. (But over the course of a season, Cousins number could go any direction and the small sample size isn't very useful. It's just a fact that Tannehill had just 39 DYAR as a rookie and thus far has been the closest example of a replacement level QB in his career thus far.)

That being said, last season Cousins took over for Griffin with three games to go because Griffin was just not looking very good all year due to his knee injury. On only 155 pass attempts, Cousins was worth an incredible 314 DYAR... below replacement.

He completed only 52.4% of his passes, had only four touchdown passes (the same number he had the year prior on a third the number of attempts) and seven interceptions. In his career now, admittedly a small sample size of a career but a career nonetheless, Cousins has thrown a pick on 4.9% of his attempts; more than twice as high as Wilson's career figure of 2.4%.

Speaking of impressively bad in a small sample size, Josh Freeman posted a DYAR of -176 on only 147 pass attempts. Freeman? More like Free-fall, am I right? You'll tell me if I'm right, right?

A good question to ask about DYAR is if it can signal that a player is not as good as his numbers may suggest. That's one of the benefits of advanced stats in baseball -- a figure like BABIP (Batting Average on Balls In Play) tries to assimilate luck to a players batting average, whether bad or good -- and something that football stats have a much harder time of doing. Naturally, it would be hard for Ryan Tannehill to post great numbers if his supporting cast was bad.

And it would be hard for the supporting cast to make it look like they are good, if actually Tannehill is bad. This would be the key for any advanced NFL stat, if it can start to attribute the truth about which players are actually worth of credit or blame.

It might seem easy to say that Tannehill hasn't had a good supporting cast since being drafted two years ago, but the reality is that the Dolphins have spent a lot of money on their receiving corps. If two years ago, Miami had drafted Wilson and Seattle had drafted Tannehill, how would the two players look at this point in their careers?

It's the same question that many have pondered since the Alex Smith/Aaron Rodgers debate of the 2005 NFL draft and where they'd be if the 49ers had taken Rodgers.

The truth is that we will never know for sure, but I think it's fair to say that a large part of player success is whether or not the player is actually good enough to play on the NFL level. Nobody that I know of would argue against the notion that Warren Sapp would've been a Hall of Fame player whether he had been on the Bucs or the Vikings. The truth is that Sapp benefited a ton from playing in front of Derrick Brooks and John Lynch, and they benefited from playing behind Sapp.

Had Sapp been on another team, he would have still been a force in the middle of a defensive line and the players behind and next to him would still benefit from his talent. Had Wilson been on the Dolphins instead of the Seahawks (*crosses my chest and says a prayer, "Hallow be thy name, please let us never go back in time and have this Hellish nightmare come true. Amen and a-women."*) I still think that he'd be great and that Miami would be better for it. Who Wilson is as a player and a person matters more than the players around him. Yes, I think the coaching staff in Seattle did him a lot of favors by giving him a real chance to compete and then knowing what his limits were and setting hard rules about what he could and could not do until he was ready, but I don't think that Wilson would've been a bad player if he was on the Dolphins.

I think that he would have a significantly higher DYAR than what Tannehill has posted thus far, though there's still time for the latter to make great strides in his career. Wilson was a much more experienced quarterback coming into the league.

Another player of note, and one that posted a very incredible DYAR in 2011, definitely threw up warning signs that he'd be one of the worst players in the game (and the fans threw up too, you guys, because his play was gross, get it) was the rookie Blaine Gabbert. Thrown into action way too early in his career, Gabbert started 15 games in 2011 but threw more touchdowns (12) than interceptions (11) so that's good, right?

Not really.

Gabbert completed just a hair more than 50% of his pass attempts and managed just 2,214 yards for a ridiculously low 5.4 yards per attempt. On the surface you're thinking, "Well, not as bad as Geno Smith's rookie season" but you'd be very wrong. Despite have a much better TD/INT ratio than Smith, Gabbert posted a fairly bad DYAR.

Smith's DYAR was -371 last season, and he threw 21 interceptions and 12 touchdowns.

In Gabbert's rookie season, he posted a DYAR of -1,009. More than twice as "low" as the next guy on the list, Caleb Hanie, who had -470 on a low number of attempts.

Few QBs to post bad DYAR seasons usually last very long as starters, that's obvious, but let's take a quick look at the other skill positions and look for interesting clues. Because often with running backs and receivers we are wondering if they're just a product of a good offensive line or a high number of pass attempts, while not making good on their yards-per-target.

Running backs

Advanced stats for running backs includes DYAR, DVOA, Effective Yards, and Success Rate.

Effective Yards, listed in red, translate DVOA into a yards per attempt figure. This provides an easy comparison: in general, players with more Effective Yards than standard yards played better than standard stats would otherwise indicate, while players with fewer Effective Yards than standard yards played worse than standard stats would otherwise indicate. Effective Yards are not the best way to measure total value because they are more dependent on usage than DYAR.

The final statistic is Success Rate. This number represents the player's consistency, measured by successful running plays (the definition of success being different based on down and distance) divided by total running plays. A player with higher DVOA and a low success rate mixes long runs with downs getting stuffed at the line of scrimmage. A player with lower DVOA and a high success rate generally gets the yards needed, but doesn't often get more. Success Rate is further explained here. It is not adjusted for opponent.

Most people can probably guess that last season's leader in DYAR among running backs was LeSean McCoy, who posted a DYAR of 341. As the league's leading rusher, that means that a replacement back in Philly would've had 1,266 yards, which was the exact amount of yards that Adrian Peterson posted as the league's fifth-leading rusher.

What may be more of a surprise is that second in DYAR was DeMarco Murray of the Cowboys. Despite only rushing for 1,121 yards, Murray had 5.2 yards per carry (highest among starting running backs) and nine touchdowns. Murray was also first in DVOA. (A good tip for you fantasy "freaks" out there, as I like to call you fantasy freaks.)

And while Rashad Jennings was 25th in rushing yards with the Raiders last year, he was 7th in DYAR. On just 163 carries, he had 4.5 YPC and six touchdowns. Teammate Darren McFadden had -38 DYAR, and while Jennings signed a four-year deal with the Giants, the Raiders replaced him with Maurice Jones-Drew, a player that had -49 DYAR last season.

Whatever happened to predictability. The milkman, the paperboy, Oakland suck-ing.

Meanwhile with the stat "success rate" you may want to look for players that excelled when given the opportunity to make plays. One guy stood out far above everyone else and was only 18th in DYAR. That guy was Danny Woodhead of the Chargers, who had a success rate of 60% last season, which was 6% higher than anyone else in the league.

And while you can guess that Ray Rice was last in DYAR (more opportunities + bad play = higher chance to finish last than a worse running back that doesn't play as much) and that he only had a success rate of 35%, which players were surprisingly bad at success rate?

Frank Gore was at just 42% and posted a DYAR lower than Chris Johnson, while Adrian Peterson had a success rate of 44% and had a lower DYAR than Gore, Mike Tolbert, and Joique Bell.

Were there any warning signs among players from 2012 that we saw come to fruition in 2013? Yes and no.

Rice was 7th in DYAR but 33rd in success rate. That being said, Willis McGahee was first in success rate and the fact that teammate Knowshon Moreno was second in success rate may signal that it had something to do with Denver's offense.

C.J. Spiller was 3rd in success rate in 2012 and then 44th in 2013.

The players with less than 100 carries in 2012 to have the most success were Andre Brown, Kendall Hunter, Mike Tolbert, and... Danny Woodhead.

There certainly could be some signals, but I'm not sure how you can differentiate them and know who is for real and who isn't. Sometimes it's still impossible to predict.

(Marshawn Lynch finished 5th in DYAR, but 17th in DVOA and 19th in success rate. However he scored 12 touchdowns and was second in effective yards.)


Among the receivers advanced stats, they also get effective yards but no success rate. Instead, there is catch rate, which is incredibly easy to understand: How many of their targets did they catch? Of course, sometimes it's difficult to say if the player that receives the "target" is actually the one being targeted, but more importantly, if it was a catchable pass or not.

Certainly if a player dives full body length and reaches out and fingertips the ball, it's a target, but one that maybe only a handful of players can make a part of the time. Should he be penalized for that? Maybe not but then in those moments when he does catch it, should we not count it? You have to consider both. (ProFootballFocus does it's best also to eliminate "meaningless" passes for quarterbacks and receivers, like throwaways.)

Last season, Josh Gordon led the NFL in receiving yards with 1,646 of them. He was ninth in DYAR at 336 yards, in part because he caught just 55% of his 159 targets. I feel like I said this in the DVOA article, but not every pass that goes to Gordon is a good one, especially since that passer was usually Weeden.

First in DYAR was Demaryius Thomas. He had a 65% catch rate on 142 targets with 1,431 actual yards and 14 touchdowns. Eric Decker was fourth in DYAR, which again, shows that playing in Denver is just a good thing to do; you can expect to see Emmanuel Sanders rise up from finishing 69th in DYAR last season, while Decker could potentially take his place in the sixties since he plays with the Jets now.

You simply can't eliminate a football player's numbers from his teammates, even with advanced stats. (Yet.)

But, can we try to spot players being undervalued and overvalued with DYAR? Hell yeah.

The player to rank highest in DYAR despite not hitting 1,000 yards last season was Marvin Jones of the Bengals. He only had 712 yards, but he finished 11th in DYAR by catching 64% of 80 targets with 10 touchdowns. Keenan Allen of the Chargers was barely over 1,000 yards as a rookie but was 8th in DYAR and caught 68% of his targets with eight touchdowns.

Comparatively speaking, rookie Kendall Wright of the Titans had 32 more yards than Allen but was 47th in DYAR. Why? Wright had 140 targets (36 more than Allen) and only two touchdowns. He also had 3.2 fewer yards per catch and posted the lowest DYAR of any player to have at least 1,000 yards last season. That being said, he did catch 67% of his targets, and if Tennessee can open up space for him to gain yards, maybe there is potential for him to get better.

Otherwise, his 1,000-yard season was pretty meh, while Allen's was pretty yeh. (that's the opposite, right?)

But again, Wright worked with Ryan Fitzpatrick and Jake Locker, while Allen (and Eddie Royal, who was shockingly 15th in DYAR last season) played with Philip Rivers during his career season under head coach Mike McCoy, a noted passing guru.

You absolutely can't get anywhere in this league without some help, and yet that help only matters if you're talented enough to do something with it. Josh Gordon and Kendall Wright both had terrible quarterback situations last season, and yet one of them was clearly much better than the other by a lot.

However, Wright will probably have more yard than Gordon next season, so there's that.

(Doug Baldwin 13th in DYAR on just 73 targets. He was ahead of Alshon Jeffery, who had more than twice as many targets. Get off my back about Doug Baldwin, I like him!

Golden Tate was 24th in DYAR and caught 66% of 100 targets.

Jermaine Kearse had 116 DYAR on only 38 targets. His DVOA wasn't as high as Baldwin's but was more than double what Tate posted. Kearse was used sparingly, but his four touchdowns was only one fewer than Baldwin and Tate.)

Chuckleberry Fin

(Because you're laughing a lot at this article while learning, and because this is the end.)

In conclusion, DYAR is Football Outsider's attempt at a WAR-like stat, but acknowledges that unlike baseball, your teammates value above replacement also matters. When we talk about DYAR next season and whether or not Percy Harvin posts a DYAR of 250, I hope we all have a better understanding of what that means. It's not necessarily how many yards better he is than an average receiver, but the value he presents over someone else presented in a number that most people are already used to seeing: Yards.

And whereas Wilson can get lost in the shuffle of yardage totals, let DYAR be your guide to how valuable he really is.

Not that his radiant smile or cat-like reflexes weren't evidence enough. There are certain intangibles that no stat can measure.