Tuesday, June 30, 2009

You can't un-learn things

Now that I've done a fair job of establishing that there's no link (or maybe a slight negative link) between a team's passing proficiency and their running game's yards per carry, I can't help but notice claims to the exact opposite all over the place. And by "all over the place," I mean two places where I generally go for better-than-average football analysis and commentary. I still like these sources and I don't really blame them for holding to a thought process that I would also have believed just a few months ago, but it's difficult for me to pass them by without dying a little on the inside.

Yesterday, Daily Norseman posited that:

I think it's fair to say Peterson would see a bump in his yards per carry average with Favre as the team's starter


While the free pdf download of FootballGuys' fantasy football magazine (a great deal, and only 21 MB!), when discussing Matt Forte and the effect Jay Cutler will have on his numbers, asks, on page 111:

1. Will the running game improve with Jay Cutler under center?


And responds with only the following information:

Yards per carry average for all seven Denver RBs last year = 5.17
Yards per carry for Matt Forte last year = 3.9


This, to me, is an egregious oversight by a group of people who should know better. Putting aside the question of whether it's an erroneous assumption, it's a classic case of small sample size. If the Vikings traded for Drew Brees, I could just as easily ask:

Will the Vikings' running game decline with Drew Brees under center?


And respond:

Yards per carry average for all New Orleans RBs last year = 4.15
Yards per carry for Adrian Peterson last year = 4.85


So, clearly, adding Drew Brees will make Adrian Peterson's YPC worse, just as Cutler will make Forte's better. It has nothing to do with any of the involved teams' offensive line, quality of their backs, play calling, run-blocking scheme -- which, it should be noted, Denver has been superb at for years, long before Jay Cutler took over at quarterback -- nope, it's entirely because Drew Brees/Jay Cutler was at quarterback. End of discussion.

(Note that both of these analyses use straight yards per carry as the measuring stick, not the consistency of the back from carry to carry, which is still possibly related to quarterbacking.)

And while I haven't looked over every player's description, the entry for Ryan Grant (page 113) leaves me scratching my head in a number of ways:

There is always a chance that the Grant we saw in 2007 was the anomaly. Without a Hall of Fame quarterback in the backfield, defenses were able to concentrate more on Grant and lessened his impact.


No, not only do we have the "quarterback affects running back's performance" myth to deal with, but there's also the "Brett Favre makes everything better" myth. Aaron Rodgers threw for 4,000 yards and 28 TDs last year. Maybe for the first part of the year, teams concentrated on shutting down Grant, because they didn't know what they'd be getting from Rodgers. By about midseason, though, if you weren't paying attention to Rodgers, he was going to kill you. Grant's disappointing season might also have had something to do with the Packers' defense being so bad as to necessitate the abandonment of the running game earlier than they would have liked.

There's probably more like this in the FG analysis of running backs, and maybe for other positions. It's still a great resource that I heartily recommend, but don't buy into the notion that Matt Forte, or any other back, will have a great season because of improved quarterbacking.

Friday, June 26, 2009

Brett & Adrian, Part 2

After my last post, Peter posited (oof -- say that five times fast!) that a strong passing game could help Adrian Peterson's average yards per carry and/or his consistency in getting positive yards with every carry. Neither of these are new ideas, as I explained in that post, referencing Pacifist Viking's post and my own previous posts on correlation. I thought it might be a good idea to revisit and possibly add to that information.

First, a warning: I am not a professional statistician or nearly as smart about this kind of stuff as this guy. So, I'm fulling willing to admit that I'm a hack playing around with mathematical concepts that, even if I don't fully understand how they work, I understand what their results signify (assuming I'm using them right).

I do have a pretty good head for numbers, though, and a willingness to give facts more weight than beliefs and adages. And the belief that we continue to hear -- not just from the pro-Brett Favre crowd, but from pretty much any team with a good running back and shaky quarterback situation -- is that a good passing game will help the running game. The argument of eight-in-the-box versus seven-in-the-box and a good passer "opening up running lanes" is posited by amateurs and professionals alike.

Problem is, the numbers don't seem to bear it out.

If Peter's principle (heh) were true, we'd expect that good passing numbers would result in high average rushing gain. As I pointed out a while back, that's not the case, at least using the stats from the 2008 season.

Pass Total/Run Average Correlation: -0.0956

Correlation, if you'll recall, is always between -1 and 1. The closer you are to 1, the closer the relationship between the two sets of data. A positive correlation (0 to 1) indicates a positive relationship, while a negative correlation (-1 to 0) indicates a negative relationship.

Even with the slight negative correlation, the absolute value of the correlation -- less than 0.1 -- is insignificant enough to be practically meaningless. Thus, according to the 2008 data, there is virtually no correlation between total passing yards and rushing average. A strong passing game does not help out the running game.

But raw passing yards (minus sacks) isn't the only measure of a team's success in passing. Just throwing for a bunch of yards doesn't necessarily make you a good pasisng team. What about the various rate stats?

Passer Rating/Run Average Correlation: -0.105
Adusted Net Yards Per Attempt/Run Average Correlation: -0.010

Again, both results are inconclusive, and it would be harder to have less of a correlation than the ANYA/Run average result (actually -0.00954)! Again, this seems to support the notion that there's no correlation between passing efficiency -- by any measure -- and rushing average.

Of course, I'm still just using one year's worth of data. It takes me a while to compile all the data, so I'll only run a couple more years, 2006 and 2007. The three-year data is presented in the chart below:









YearPass Total/Run AvPass Rate/Run AvANYA/Run Av
2006-0.313-0.002-0.023
2007-0.1870.110.151
2008-0.096-0.105-0.01
Total-0.203-0.0070.063

All three correlations grade out poorly, at least in terms of supporting the theory that passing helps running, and using the fairly small sample size of three seasons. In fact, the original hypothesis -- that a lot of passing yards helps improve a team's average yards per attempt -- seems to have a rather significant negative correlation, meaning that one somewhat implies the opposite of the other (though the direction of the correlation cannot be discerned).

I think this is a case of people trying to sound smarter than they are (like when they say that, despite the glitter and hype of passing games that the way to win is to "run the ball and stop the run") and attributing more importance than is warranted to a "hidden" part of the game. I think it's much simpler than that: Teams that are good at passing pass the ball and so they have better passing numbers than rushing numbers. Teams that are good at running run the ball and so have better rushing numbers than passing numbers. That's it. If it is true that running well opens up the passing game, then the passing numbers will look better and the strong-running team will look more balanced, and vice versa.

In the end, I think that, if you want to be a better running team, improve the parts of your team directly related to your running game: your tailback, fullback, offensive lineman/tight ends/receivers who can run block, etc. Improving the quarterback doesn't really help the running game.

(There is, of course, the question of running back consistency, on a per-carry basis that, without per-carry stats, I can't prove or disprove, but the notion of it strikes me as fishy for two reasons: 1. There's no proof that it will help and it sounds like just another platitude to help us accept the unacceptable; and 2. It's based on the notion that Peterson is that type of "boom-or-bust" running back, which is only prevalent because he played that way in his last two games, while ignoring his previous 28.)

All this doesn't mean that a good quarterback doesn't help a team. I'd be delighted if Drew Brees or Peyton Manning somehow wound up on the Vikings, and it would clearly improve the team, but not because it would make Adrian Peterson look better. But the notion that Brett Favre, or any other quarterback, will make Peterson's life easier is, in my opinion, just another way to make us feel better about accepting ol' #4 as our starting quarterback.

Wednesday, June 24, 2009

Thursday, June 18, 2009

Can Brett help Adrian?

One of the persistent reasons I've heard for bringing Brett Favre -- or any quality quarterback -- to the Vikings is that the threat of an improved passing game will open up more holes for Adrian Peterson (and Chester Taylor) to run through, thus improving the running game as well as the passing game. With teams forced to respect the pass more, Peterson will face fewer eight-man fronts and be poised for a spectacular year.

Putting aside the question of whether Brett Favre will improve the Vikings' passing game, I've always found this type of reasoning questionable. Certainly, on the surface, it makes sense, and you hear this logic frequently espoused by commentators and even coaches. But does a better passing game actually improve the running game? And here's another thought: Adrian Peterson has averaged 101.3 yards per game and 5.2 yards per carry during his professional career. Is it really realistic to expect that any improvement to the team will push him much over a 1,600-yard season on a consistent basis?

When you think about it, in terms of raw numbers, an improved passing game should probably decrease overall rushing numbers. After all, if your passing game is good, you should be using it a fair amount, and that's going to take carries away from your running back. The 2008 Vikings threw the ball 452 times (with 43 sacks) and an unknown number of QB scrambles, for about 500 total pass dropbacks. If they'd dropped back to pass 600 times, that would have been 100 or so fewer possibilities for Peterson and Taylor to carry the ball. That'll subtract from your rushing yards, no doubt.

I took a few stabs at this concept last year, and my admittedly amateurish results showed that a good passing game does not help the running game (or vice versa). Brian Burke over at Advanced NFL Stats has talked about a similar topic a few times, and uses what I like to call "The Princess Bride Paradox":

* Your team is very good at rushing. Thus, it should rush the ball.
* The defense knows you're good at rushing. Thus, they'll play a stout run defense.
* You know that the defense knows you're good at rushing. Thus, you'll surprise them by passing!
* The defense knows that you know that they know that you're going to try and stop the run. So you'll outsmart them by playing the pass!
* You know....

Y'know?

Brian's follow-up post used the 2007 Vikings as a specific example by asking what their optimal run-pass mix was, given their talent. His conclusion was that talent on one side of the offense (running or passing) shouldn't affect play-calling on the other side, though he did acknowledge that the passing game improved slightly when Peterson was added in 2007.

But all this still doesn't answer the basic question: Would an improved passing game help Adrian Peterson? Since he ran for 1,760 yards last year at a 4.8 yards-per-carry clip, there doesn't seem to be much that could help him. So I got to wondering, in the best seasons ever by running backs, did those backs have a strong passing game to "open up lanes" for them?

The answer: not so much. I took the top 27 single-season rushing performances -- that's every season of 1,700 or more yards -- and checked out their passing games. Here are the results:


































YearRunning BackRush YdsPass AttPass YdsRating
1984Eric Dickerson2,1053582,38265.9
2003Jamal Lewis2,0664152,51764.7
1997Barry Sanders2,0535403,60575.4
1998Terrell Davis2,0084913,80893.5
1973OJ Simpson2,0032131,23642.7
1980Earl Campbell1,9344633,27170.4
1994Barry Sanders1,8834593,08580.2
2003Ahman Green1,8834733,37790.5
2005Shaun Alexander1,8804743,63296.8
1963Jim Brown1,8633222,44978.3
2005Tiki Barber1,8605583,76275.7
2002Ricky Williams1,8534553,06979.3
1977Walter Payton1,8523052,07061.8
1998Jamal Anderson1,8464243,74492.7
1986Eric Dickerson1,8214032,38063.7
1975OJ Simpson1,8173542,66180.2
2006LaDainian Tomlinson1,8154663,41293.0
1983Eric Dickerson1,8084893,41176.0
2006Larry Johnson1,7894503,24384.7
1995Emmitt Smith1,7734943,74191.7
2008Adrian Peterson1,7604523,21781.5
1985Marcus Allen1,7595063,48168.5
1997Terrell Davis1,7505133,70487.4
2005Larry Johnson1,7505074,01490.1
1985Gerald Riggs1,7194623,02566.5
1992Emmitt Smith1,7134913,59788.8
2000Edgerrin James1,7095714,41394.7
























1993Average1,8554483,19780.5

The five running backs with 2,000-yard seasons had Jeff Kemp, Kyle Boller, Scott Mitchell, John Elway, and Joe Ferguson as their primary quarterbacks. Apart from Elway (who missed four games with injury and was replaced by Bubby Brister), that's a pretty sorry group.

Only eight of the 27 seasons featured quarterbacks with passer ratings of 90 or higher and the overall weighted average is 80.5 and the better entries seem to be clumped down near the bottom, with the lower rushing totals. That's not real impressive, and only four of these 27 seasons occurred before the changes to open up the passing game in 1979.

Of course, what I should really be looking at is overall rushing totals for an entire team (minus quarterback rushing numbers), but that would require more effort than I'm willing to put in :P Still, consider the following additional two seasons:


































YearRunning BackRush YdsPass AttPass YdsRating
2007C. P.
1,7564322,93874.2
2008C. P.
1,727452

3,21781.5

Those are a couple more excellent 1700+ yard seasons that would be added to the list above if "C. P." were a real person. He's actually the combined rushing numbers of Adrian Peterson and Chester Taylor ("Chester Peterson"), multiplied by 80% to give a reasonable number of carries (316 and 372 for 2007 and 2008, respectively) and represent that C. P. probably wouldn't be actually able to carry the ball 859 times over consecutive seasons. Again, the quarterbacking for C. P.'s team has been mediocre to poor for two seasons, yet he's put up spectacular numbers. How would adding a better quarterback improve them?

Without question, better quarterback play would improve the Minnesota Vikings as a team, though whether it would improve the running game is, I think, questionable. Even if we could somehow acquire the in-their-prime versions of Joe Montana and Jerry Rice and plant them on the '09 Vikings, Adrian Peterson would be hard-pressed to improve on his stats from the previous two years. There are just only so many plays to go around, and if the passing game is that much more improved, we should use it more.

Pacifist Viking does raise a good point, though. It might be that an improved passing game will make Peterson more consistent on a per-play basis. I brought up Barry Sanders in the discussion, as a perfect example of a player who very rarely had a good passing game to support him and could be maddeningly inconsistent on a carry-by-carry basis. Without access to play-by-play data, it'll be tough to prove/disprove this, but it's a decent enough assumption. Personally, I think consistency is overrated, but that's a topic for another post.

Based on the evidence from NFL history, though, it's hard for me to believe that an outstanding rushing season is the result of good passing from the same team. Some of this might be because, with limited resources (i.e., cap room), teams can't invest in as much. So you have a great running back and so-so QB and receivers (and a line that's better at run blocking than pass protection); all that might result in some of the disparity we see here. But when it comes to improving your running game, the most important factor, I feel, is the talent of your running back and his blockers -- not your quarterback and receivers.

Wednesday, June 17, 2009

Another possible implication

I'm probably dwelling too much on the worst-case scenario, but consider this:

Tarvaris Jackson is a free agent after 2009. Suppose that, somehow, he emerges as a good quarterback this year -- either the team doesn't sign Brett Favre and T-Jack wins the starting job or Favre does come in and is replaced by T-Jack, either due to injury or ineffectiveness.

So, after the 2009 season, you're Tarvaris Jackson and you're a relatively sought-after free-agent quarterback. Will you sign with:

a) The team that tried to cast you aside twice to pursue an aging legend; or
b) One of 31 other options.

Of course, another alternative is that T-Jack does nothing to make the Vikings wish to retain his services. One way or another, I'd wager that 2009 will be T-Jack's last season in purple. Maybe that's not a bad thing, but you never know.

Thursday, June 11, 2009

The dubious importance of timing

(Don't worry, there's football contained herein.)

So I've been following a discussion on my favorite comedy/baseball site about the importance of RBIs. My views on the validity of RBIs as a stat pretty much boils down to a comment by Hossrex, in response to the original article writer, Patrick:

Patrick: “You seem to think that hits with men on base is luck, like a roll of the dice.

No. Getting a hit is skill. However, whether or not there are runners on base when he gets the hit is luck… like a roll of the dice.

That's about as good as I can put it. I've held the belief for a long time that when a player gets a hit -- or scores a touchdown or a goal or whatever -- is less important and very random when compared to his ability to get hits. A player who gets 200 hits in a season or who hits 40 HR in a season is going to do some of that with runners on base and do some of it with runners not on base. Whether he does so with men on or not is a factor almost completely beyond his control. His RBI total is a factor of his overall hitting ability and the skills of the batters hitting in front of him -- not in his ability to deliver "clutch" performances.

A few other people in the comment chain bring up Albert Belle, possibly the best RBI man of the 1990s. A look at his stats shows the following splits (listed as AVG/OBP/SLG/OPS):

Bases Empty: .296/.363/.571/.934
Men On Base: .293/.376/.557/.932

Take away Belle's intentional walks with men on base (reducing his OBP to .359 and his OPS to .916) and the fact that OPS rises when men are on base as a rule, and you can argue that Belle was worse with men on base! His OPS with runners in scoring position is a healthy .991 (dropping to .965 when you take out intentional walks), but those encompass less than 60% of his plate appearances with men on.

Two other noted "run producers" of the 90s fare similarly (removing intentional walks):

Juan Gonzalez
Bases Empty: .290/.332/.552/.884
Men On Base: .301/.340/.569/.909

Joe Carter
Bases Empty: .255/.294/.467/.761
Men On Base: .264/.306/.461/.767

So why did these men accumulate so many RBIs and gain a reputation as "RBI men"? Part of it is that they were, in actuality, good hitters. Guys who hit 30-40 HR a year are going to drive in runs, just by happenstance. Another factor is the guys hitting in front of them who get on base a lot and generally run well. Albert Belle had Kenny Lofton leading off for him. Gonzalez had Will Clark, Rusty Greer, and a few others.

But nobody had their table set better than Joe Carter, who almost certainly has the worst non-RBI numbers of anyone with his hitting line. How can you not drive in runs when you have, at various times in your career, Roberto Alomar, Paul Molitor, and even, for a few months, Rickey Henderson, batting ahead of you? Carter ranks 57th all time in RBI but 261st in slugging percentage and 625th in OPS, and, as demonstrated above, was no better at driving runners in than he was at hitting with nobody on base.

But this isn't supposed to be a "bash Joe Carter" post. The point is, instead, to try and illustrate that just because a player does what he's supposed to do -- such as get hits or score TDs -- he is probably not any better (or worse) at doing it at specific times than his overall skill level would indicate.

And that brings us to situational stats in football and this question. If you give the ball to Adrian Peterson (or Emmitt Smith, or Walter Payton, or Jim Brown) on 3rd-and-2 at your opponent's 20 down 7-3 with three minutes left in the first quarter, does he have any better of a chance of gaining that first down if it's 3rd-and-2 at the opponent's 20 down 7-3 with three minutes left in the fourth quarter?

I say "no."

Putting aside unquantifiable things, like player fatigue level, general play calling during the day, and so on -- and I arranged the example to make a run or pass pretty much equally likely in either scenario -- I don't believe that a player performs any better or worse in the same scenario at different points in the game and that his accumulation of "good stats" (whether they be RBIs or first downs or whatever) are due more to their being in a position to succeed more often than their peers and performing at a rate roughly identical to their established skill level.

In my baseball examples, a player is going to approach each at-bat pretty much the same way, whether there are runners on or not. In my football example, the running back is going to try to gain that first down (or more) with everythign he's got, whether it's the first quarter or the fourth (and, frankly, if he puts forth "extra effort" in the fourth, why can't he do that throughout the game?). Obviously, some game circumstances can change things, but, for the most part, baseball players are trying to hit the ball as hard/well as possible and football players are trying to gain as many yards as possible every chance they get.

You can approach this any number of ways. Does Peyton Manning pass well on third downs becuase he's good on third downs or just because he's good, period? Does Sidney Crosby have a lot of game-winning overtime goals because he "turns it up a notch" in overtime or becuase he's a great player overall? Alternatively, does Adam Dunn strike out a lot with the bases loaded because he's not "clutch" or because he just strikes out a lot?

And then there's my favorite "circumstantial" football stat, the fourth-quarter comeback. I've heard more than a few times the last few months that Brett Favre has 42 fourth quarter comebacks. Putting aside the notion that, in order to have to make a fourth-quarter comeback, you have to be behind (*coughinterceptionscough*), can anyone tell me how many times Brett Favre has taken his team into the fourth quarter while behind and not come back? Anyone?

Didn't think so.

Favre has 100 losses as a starter. His team trailed at some point in the fourth quarter in all those losses, so that makes Favre's "4th quarter comeback rate" 42/142, or about 30%. Seems rather "meh," I think. And yes, several of them might have been the other team scoring with little to no time left and Favre not having a realistic chance to come back, but I'd also wager some of the victories involved the Packers taking the lead (maybe on a run or a defensive score) with 14:50 to go in the fourth quarter and not relinquishing it. Hardly dramatic, but it still counts as one of the 42.

Then there's the question of whether Favre (or any other QB famous for his late-game heroics) is actually even better in the fourth quarter than he is in the other three. (Looking at split stats is deceiving, since offenses and defenses do radically change their approaches depending on the time of the game and the score. Still, his 79.6 career passer rating in the 4th isn't anything to get excited about.) If so, why doesn't he play that well throughout the game? Such monikers are usually the result of an overall high quality of play and a few legendary performances in big situations (Joe Montana, John Elway, and Reggie Jackson are all good examples) etching the player's name into posterity. Remember, just a few years ago, Peyton Manning wasn't "clutch." Then he won a Super Bowl.

But this isn't even a "bashing Brett Favre" post. (I've done plenty of that lately.) If you've gotten this far, I just hope you can open your mind to the thought that opportunity and overall skill level matter as much when evaluating a player as his counting stats and that when a player does something is less important, in the long run, than doing it consistently and getting plenty of opportunities for "big moments."

Tuesday, June 2, 2009

Fun with numbers

Using the Historical Data Dominator:

* Only Eric Dickerson, Edgerrin James, and Earl Campbell have ever rushed for more yards in their first two seasons in the league than Adrian Peterson.

* Peterson already ranks ninth on the team's all-time rushing leaders list, and could vault all the way up to fourth if he amasses more than 1,445 yards in 2009.

* Bernard Berrian was a major deep threat in 2008, but he doesn't compare to John Gilliam in Vikings lore. Among pass-catchers with at least 25 receptions in a season, Gilliam owns the #2, #3, and #4 seasons in yards per reception. Berrian clocks in at #9.

* Who's the Vikings' single-season leader in receptions by a tight end? Nope, it's not Steve Jordan. Not Jermaine Wiggins. Not even Byron Chamberlain. It's Joe Senser. Who?

* Last year, I ran a comparison using the HDD and comparing quarterbacks with numbers similar to Tarvaris Jackson's in his second year. Running a similar comparison based on last year's numbers gives me a much smaller, not to mention less favorable (John Fourcade?!) list, partially because of Jackson's limited playing time.

Expanding the comparison to take Jackson's three years and total stats into account, I get this larger list of "similar quarterbacks" through their third years. The results are a little more favorable -- Steve Young and Roman Gabriel stand out -- and includes several decent-but-not-great QBs who experienced some success in their careers: Steve Beuerlein, Bobby Hebert, Elvis Grbac, and Kordell Stewart.

I tried the same trick with Sage Rosenfels, but it's an iffier proposition with him, because there aren't very many eight-year veterans who have only thrown 500-odd passes. His list is understandably small and undistinguished and includes a few QBs -- like A.J. Feeley and Matt Cavanaugh -- who were career backups given the occasional chance to start, with predictable results. This only further cements my notion that Tarvaris Jackson is probably the best option the Vikings have at QB.

* And finally...here are all the seasons ever turned in by 40-year-old quarterbacks. Gee, good thing we don't have one of those!