By Bill Barnwell
bill@patriotsdaily.com

Discussion regarding the Indianapolis Colts during the Peyton Manning era has rightly revolved around an offense combining clinical effectiveness with historical staying power. When the Colts won the Super Bowl, people talked about the Colts defense as if it had suddenly taken some leap forward, but it was a case of the media applying the simplest narrative possible to the story. While the Colts defense had taken a step forward in 2005, it had struggled mightily in 2006; it was only once the playoffs started that it returned to its 2005 form. In 2007, the Colts defense has been the best of the Dungy era, which would turn the Colts into a scary dynasty…most years.

What makes the Colts defense fascinating, though, is how it’s constructed. The Patriots got the reputation as the team that would let its stars go and replace them with draftees, but the Colts are a much better example. It’s not just that players like Cato June, Mike Peterson and Jason David have been allowed to leave in free agency; it’s that the Colts have successfully replaced them with middle-round draft picks and players that would normally be classified as roster filler. They don’t make a lot of money, and once they’re about to, they’re let go.

Obviously, the effects of being able to plug in these defenders allow the Colts to employ what’s commonly referred to in fantasy auctions as the “stars and scrubs” approach: Spending loads of money on a group of star players, and then spending the minimum or close to it on a large portion of your players. It’s diametrically opposed to the Patriots strategy of depth and not overpaying your star players, and while an article comparing the two approaches would require more research than is available in mid-week, I wanted to look at the Colts defense and see how it developed over Dungy’s tenure with the club.

I apologize in advance if any of the starters are incorrect, whether they’ve been placed in the wrong position or listed ahead of the real starter. Also, remember that the salary cap has risen dramatically in the last five seasons; the Colts’ salary expenditure in 2002 was a shade under $65 million; in 2006, that was over double, at $131 million. We’ll be looking at players’ salaries (not their cap value) as taken from the USA Today Salary Database; for 2007 data, we’ll be drawing information from coltscap.net.

We ironically start with the position the Colts identify as the most important in their Tampa-2 defense, defensive end. A majority of the Colts’ pressure comes from the defensive line, particularly Dwight Freeney and Robert Mathis. Freeney and Mathis both play most downs and their primary purpose is to rush the passer; while that sometimes ends with them being pushed upfield while a running back runs into the hole they’ve vacated, they are a threat to get to the quarterback against any offensive line.

The Colts also value their defensive tackles as worth spending significant sums of money — not appearing on this chart is DT Booger McFarland, out for the year with a knee injury, who the Colts traded a second-round pick for and are paying $5,000,000 in salary to this season. The Colts found a gem in Raheem Brock, who the Eagles cut when they ran out of rookie cap space — Brock was good enough to be one of the few guys the Colts re-signed following their rookie deal. In addition, the Colts gave Corey Simon a big deal as a free agent, but Simon’s deal went so disasterously that Jim Irsay called it a “bad mistake”. Oops.

Here’s the first position where we really see how the Colts defense operates. Every single one of the Colts’ starters was let go following the expiration of his rookie contract except for Robert Morris, who was a failed middle linebacker and signed a cheap deal. The Colts have a plan on how to spend the cap space they’ve allocated to their defense, and know that there are some places they have to cut back at. When that means letting talented players like Mike Peterson or Marcus Washington go, the Colts bite the bullet, plug in the best guy they have, and let him rack up tackles. The latest example is Freddy Keiaho, who had a breakout game against the Saints in Week 1.

Morris became a free agent after his rookie contract ended, had no interest, and ended up shuffling back to Indy as a backup. His former backup, the undrafted Gary Brackett, had an excellent year in 2005, and then before he could become a restricted free agent in 2006, the Colts locked him up with a four-year, $10 million deal that included a $3.2 million dollar signing bonus, which is now all paid. In 2008 and 2009, Brackett will make an average of $2.25 million, but his contract will inflict no harm on the Colts’ cap if he’s cut to save costs.

The other position where the Colts have no problem getting rid of players is at corner — this is a strategy diametrically opposed to that of the Bears, who also play the Cover-2, but locked up both Charles Tillman and Nathan Vasher to long-term deals.

The Colts have seemingly changed their strategy some recently, though, as while they used overaged undrafted Nick Harper and second-day pick Jason David at corner, they also spent first- and second-round picks on cornerbacks Marlin Jackson and Kelvin Hayden, respectively, to replace them once their contracts came up in 2006. While Harper’s been successful in the stifling Tennessee pass defense, he’s 33 and his career is about finished; David, meanwhile, has struggled mightily in New Orleans.

Don’t expect the Colts to hold onto Hayden and Jackson once their contracts expire, though.

The one place where the Colts have been remarkably stable is at safety. Mike Doss and Idrees Bashir have transitioned nicely into Bob Sanders and Antoine Bethea.

Sanders is the most fascinating case on this defense. A second-round pick in 2004, Sanders held out and ended up receiving a six-year contract that allows him to void the deal after this season. As Sanders makes a relative pittance for his talents, he’s a lock to do so. Troy Polamalu received a four-year, $30 million contract with $15 million in incentives. Sanders would have every reason to expect and demand a similar contract. As you can see, the Colts have not, under Tony Dungy, valued the safety position at the level. It will be very interesting to see how the Colts handle the situation; it’s worth noting that, in a safety-rich draft, the Colts didn’t use one of their 2007 picks on a safety.

What’s more important than deciding whether the stars and scrubs or super-depth strategies are superior or inferior is understanding that the Colts have a plan and stick to it. So many teams in the NFL (everyone stare at Detroit and Oakland) struggle with taking a plan, sticking with it, and trusting it. They’re the dieters who starve themselves for a week and then binge.

The Colts, on the other hand, just eat well. They have a plan: Splurge on defensive linemen, particularly at end, and elsewhere, insert players on their rookie contracts and once they’ve played those contracts out, let ‘em hit the market. The result is a team with loads of money to spend on the places they deem valuable — and an offense that might even match the Patriots come Sunday.

By Bill Barnwell
bill@patriotsdaily.com

42 points in a half is a lot, even for these Patriots. It’s not an unbeatable total if you’re Bill Belichick, apparently, but it’s enough to inspire thoughts of record totals. It leads announcers to point out that the Patriots, against the Dolphins, were on pace to score 84 points.

The thing is, of course, teams who are “on pace” to score 42 points don’t score at the same pace in the second half. Both teams see their game plan change: The leading team runs the ball more to kill clock, while the trailing team throws the ball almost exclusively to make some attempt to catch up, as unrealistic as it might be. The leading team generally has fewer possessions because, inherently, teams that score 42 points in a half need some luck to get there — namely, an extra possession or two from turnovers. If we look at all drives from 1998-2006, we can see that teams would find it extremely hard to get to 42:

With six possessions on average in each half, teams would need to score a touchdown on every drive. The Patriots are good, but not that good — they benefited from a Cleo Lemon fumble that gave them a short field, and picked up seven points on a Willie Andrews kickoff return.

What we also see in that breakdown is that both scoring and possessions are seemingly consistent — an average performance from a team sees them score about five points a quarter on three possessions.

Of course, describing the Patriots offense as “average” is nonsensical at this point. Let’s split the data by points per quarter and see how many points, for example, a team that goes scoreless in the first quarter scores on average in that game:

So, then, a team that goes scoreless in the first quarter averages 14 points per game — slightly less than the 5-point-per-quarter pace that we’d expect.

The issue with looking at data towards the realm of scoring the Patriots approached is sample size; while 15 teams have scored 38 points in a half, only one has scored 41, 44, or 47 points. That being said, none of them have significantly increased their total, so it’s a pretty safe bet to say that we shouldn’t really expect the Patriots to double their point total in the second half.

If we break point totals down by possessions per quarter and half, there are some interesting trends:

The Patriots, for reference, had two possessions in the first quarter (for 14 points), five in the second (28 points, with the Andrews kickoff return counting as a possession), but only one in the third and three in the fourth — that’s eleven possessions, actually one below average.

It’s interesting that there’s only really a slight upwards trend in the first half for points compared to possessions — you would assume that more possessions would yield more points and a shootout, but it’s also the case when teams just can’t move the ball and are punting at each other all game. In the third quarter, actually, there’s no relationship whatsoever between possessions and points, as it begins to blur together.

Do teams that have a lot of possessions in the first half continue to have them in the second half, though?

Not really. The small increases at the margins aren’t enough to say that possessions in the first half have any predicative value for possessions in the second half.

Next time an announcer says that a team is on pace for a certain amount of points, you can safely ignore him or her. Whether it be due to luck running out, changes in the game plan, a lack of possessions, or combination of the three, most teams’ scoring totals regress to the average over the course of a full game.

By Bill Barnwell
bill@patriotsdaily.com

The Dolphins trade of Chris Chambers this week was both surprising and not surprising. On one hand, rumors pegged Chambers as potential trade bait going back to training camp, where he failed to impress new head coach Cam Cameron. Chambers was also plying his trade for a 0-6 team about to undergo a difficult rebuilding process, the other side of which Chambers wouldn’t be likely to see. On the other hand, Chambers had been an important part of the Dolphins offense this year, and was rebounding back to at least mediocrity following a dire 2006.

The other surprising thing is that the Dolphins would make this deal considering what else they have at wide receiver. Marty Booker, much like Chambers, is a veteran with little value left to the Dolphins. Derek Hagan is a lanky receiver who struggles with drops. Greg Camarillo is, ironically, a San Diego castoff who followed Cameron to Miami, but the elephant in the room is the one least likely to make an impact this year – Ted Ginn.

The Ohio State product was Miami’s surprise first-round pick in the NFL Draft following a high-profile career at OSU, highlighted by a kickoff return in the BCS Championship Game that left him injured for several months with a foot concern. Ginn’s performance at school wasn’t particularly special; he caught 135 passes in three seasons, and averaged only 14.4 yards per catch, not a particularly high number for a deep threat. The problem is, though, that we don’t know much, if anything, about what college statistics and performance mean with regards to wide receivers. I did hypothesize earlier in the year that Big 10 wideouts might be underrated versus those from other divisions because of the propensity of Big 10 teams to run the ball, but that still requires more research.

Besides, people weren’t picking Ginn based upon his college performance, they were picking him based upon the possibility that he might be Devin Hester 2.0. Ginn’s return skills were dynamite, and that was best measured by, well, his measurables. While Ginn did not run at the combine because of his foot injury, the numbers provided by his school were pretty amazing, highlighted by a blazing 4.28 40 time.

The thing is, standardized measurements are a lot like the standardized tests you’ll take while preparing to go to college; they’re a useful indicator of a person’s relative intelligence, but by no means are they exact. Some players actually play much slower than their 40 time, as taken without pads or opponents, while others run just as fast regardless of the conditions.

Much like how we don’t have any idea of the correlation between receiver performance in college and the pros, though, we’re still in the dark about what relevance a good or bad combine has to future success. We can probably infer that a good combine is a better indicator of a good player than a bad one, but what parts mean more? Is 40 time the most important metric, or is it vertical leap?

What gets lost in the whole mess is that Ted Ginn’s not a particularly fantastic prospect when it comes to his measurables. Granted, a 4.28 40 is a great time; that being said, he’s not particularly large and the only other figure we have for him, his vertical leap, is downright mediocre. At 5’11” and with a 34.5” vertical leap and a 9’9” jump, Ginn’s not going to be able to get to the ball in traffic, meaning that unless one of those things change, he’s going to have to run past guys to get the ball as opposed to running around them.

The other interesting thing I discovered when doing research on Ginn and his combine numbers is something I reported in this year’s Pro Football Prospectus; namely, that Ginn’s measurables match up very well to a current NFL player.

The similarities between Ginn and Texans receiver Jerome Mathis are pretty remarkable. Their height, weight, 40 time, and vertical leap are pretty much exact duplicates of one another. The big difference is in the jump, where Mathis has nearly a foot on Ginn. While Mathis has made no impact on the Texans passing game, he was a Pro Bowl returner in his rookie season before missing most of 2006 with an injury.

The other NFL receivers on this list also have superior numbers to Ginn. Chambers had 25 pounds of bulk on Ginn, and was a freak athlete outside of his blazing speed. Santana Moss had more than a half-foot of vertical on Ginn, and likely represents the best possible career path Ginn could hope for. Tim Carter’s found it impossible to stay healthy, while Troy Williamson remains a bust now into his third season.

Furthermore, there’s no real indication that even a great returner is easily findable in the draft. Take a look at the last ten years of Pro Bowlers at the return spots. Only Charles Woodson and Eric Metcalf were first-round picks; while 2006 selectees Justin Miller and Hester were both second-round picks, most of the others were either undrafted free agents or Day Two picks.

Ginn’s been virtually nonexistent this season. He’s been a decent return man, if not a great one, but his role in the Miami offense has been reserve receiver. While that puts him ahead of Robert Meachem in the scheme of things, that’s no great shakes. It’s entirely impossible that Ginn could turn out to be a great receiver, but based upon all the data we have up to this point, there’s not much evidence pointing in his favor.

By Bill Barnwell
bill@patriotsdaily.com

Although the Cowboys almost didn’t make it, their narrow victory on Monday does present us with a rarity coming up this week: a matchup of two teams, both 5-0. In fact, since 1983, there’s been exactly one 5-0 vs. 5-0 game, and it also involved the Patriots. In 2004, the 5-0 Patriots met the 5-0 Jets in Week 6. The end result? A 13-7 victory for the Patriots, and while both teams made the playoffs, the Patriots won the Super Bowl. So, then, if the Patriots beat the Cowboys this week, we can just pencil the Patriots in as champions and move on with our lives, right? No? Oh, statistics are annoying.

Obviously, there’s much more to the whole winning-the-Super Bowl-thing than going 6-0 and beating a 5-0 team. What we can do, though, is attempt to quantify what going 6-0 instead of 5-1 might do for the Patriots.

To do that, we’ll take the game-by-game records of every season from 1983-on, excluding the strike season of 1987. We’ll do game-by-game instead of week-by-week so that performances match up over the different bye week patterns that we’ve seen throughout the last 20+ years.

First, let’s see how likely a team is to even be 5-0 in the first place.

From this, we can see that 5.4% of teams have been 5-0 following Week 5. That means that that, on average, we could expect 1.72 teams a year to start their seasons 5-0. Now, you can understand how unlikely it is (a .28% chance, actually) that two 5-0 teams would be meeting when there’s not even a 50/50 shot of two 5-0 teams even existing come Week 5. In 2006, only Indianapolis and Chicago were 5-0, and in 2005, it was only Indy.Remember that this is based on real data, not simulations, so while there’s obviously a chance of a team going 16-0, because it hasn’t happened in the timeframe, 0.0% of the teams have won 16 games out of 16.

While it’s good to know the Patriots are one of a rare breed, what does that mean? Well, what we can do is compare those winning percentages week-by-week to a team’s end result; namely, how many wins did a team that started the year 5-0 average?

So, then, the Patriots catapulted themselves from a team that averaged 11.3 wins per season last week to one that now averages 12.2 wins per season. If they were to beat the Cowboys next week, they would go even higher and average 12.7 wins per season, but if they lose, they pretty much are back to where they were following their fourth game.

The bigger question, then, is what that means for the Patriots playoff chances. How often do teams that win 11 games make the playoffs? Or, alternately, how often does a team that starts 5-0 make the playoffs?

So, a team that wins 11 games will make the playoffs 98.1% of the time. The only 11-5 team to not make the playoffs in the course of this study were the 1985 Broncos, who missed out on a tiebreaker to two other 11-5 wild card teams, one of whom was, of course, the Eason Express.

On the other hand, a team that goes 5-0 will make the playoffs 94.1% of the time. A team that goes 6-0 will make it 96.0% of the time, but a team that’s 5-1 will only make it 85.4% of the time.

Finally, using this data, we can actually answer the age-old question. What’s the most important game a team can play? Judging by the playoff percentages, it’s actually a tie. In Week 9, a 4-4 team can really determine its destiny with its performance. If they lose, their odds of making the playoffs are a woeful 11.8%; if they win, an even 50%.

That 38.2% difference based upon the outcome of a sole game also comes up, not surprisingly, in Week 16. An 8-7 team is on the hotseat this time. Win, and you’ll make the playoffs 50.7% of the time; lose, and it’s only a 12.5% shot. Maybe some games really are more important than others.

By Bill Barnwell
bill@patriotsdaily.com

The Bills’ selection of Marshawn Lynch was a simple one. Buffalo needed a running back after trading the wantaway Willis McGahee to Baltimore. Lynch was, outside of Adrian Peterson, the best running back available in the draft. If only all draft picks could be this easy!

Unfortunately, they’re not, and even though Lynch was the second-best running back in this draft, it’s no guarantee he’ll become a star. My own opinion on Lynch after some pre-draft film study was that he was likely to be a starting running back in the NFL, but was unlikely to become a star; mainly, his 4.46 40-time seemed to be a product of effective workouts as opposed to a speed Lynch actually played at. That’s my only real complaint about Lynch, though. He bounces out of trash very well, he stays low to the ground and gives defenders little to hit, and is a quality blocker and receiver already (if you’ve watched Laurence Maroney come out in passing situations over the last two weeks, you realize the importance of those skills to your playing time). There was one thing I didn’t really take into account when I was looking at Lynch; his college stats.

The running back preceding Lynch at Cal was J.J. Arrington, currently the third-string halfback for the Arizona Cardinals. Arrington was a JuCo transfer who had one gigantic year in a California uniform, when he rushed for 2018 yards on 289 carries, averaging 6.89 yards per carry. In Lynch’s two years as the starter for California, he averaged 6.21 yards per carry. Arrington’s of a similar build to Lynch; Lynch is an inch shorter and weighs about 10 more pounds. His 40 time was 4.49, and his vertical jump 1/2 inch smaller than Lynch’s. They’re very similar, but Arrington’s performance in college was superior to Lynch’s. Does this mean he’s likely to be the better player? Not necessarily.

At Football Outsiders, we’ve begun to scrape the surface on college performance and how it relates to a player’s likelihood of succeeding in the pros. Our first efforts in this arena is the Lewin Career Forecast, named after its researcher and writer, David Lewin. His research (the original essay for which is available here shows that for quarterbacks selected in the first two rounds, a quarterback’s college completion percentage and games started are inextricably linked to his propensity for professional success.

We’re currently working on similar research into the performance of running backs and wide receivers. We’re not there yet in being able to point to a certain stat as immutable proof of quality like we are with quarterbacks, but the predictability of quarterbacks and the reliability of translated statistics across levels in baseball, basketball, and hockey leads me to believe that college statistics for running backs and wide receivers have some predicative value.

This was a topic I examined last year when I was attempting to analyze the selection of Laurence Maroney amongst the broader spectrum of both Big Ten backs and heavily-used ones. Since that essay was written, Maroney’s predecessor at Minnesota, Marion Barber, had an excellent year with Dallas, while Maroney had a good year for the Patriots. The track record of other recent Big 10 backs, though, has been hit-and-miss, and while doing research into college team statistics, I realized one of the reasons why: their statistics are bloated.

When I was compiling college statistics using the NCAA’s database (which covers teams from 2000-2006), I organized them by conference, normalized the teams’ schedules to a 12-game season, and then calculated the average performance each team put up for each year. By doing that, I was able to find the offensive performance of the average team in each conference over the six-year span.

What I found was that the run/pass ratio across different conferences was noticeably different. Run/pass ratio is a simple metric that measures how often a team runs the ball as opposed to passing it; for example, all D-1 teams over the six-year span averaged 1.28 rushing plays for each passing play. If you limit the figures to major conferences (the ACC, Big 12, Big East, Big Ten, Big West, Conference USA, MAC, Mountain West, Pac-10, SEC, Sun Belt, and WAC), the figure is 1.24. The Big Ten and the Pac-10, though, stand on different sides of the divide. Big Ten teams averaged 1.36 rushes for each pass, while Pac-10 teams averaged only 1.08, the lowest figure for any conference.

What this means is that Big Ten rushing figures are likely to be inflated, while the passing statistics of the Pac-10 are likely depressed some. Meanwhile, the opposite is likely to be true for Pac-10 players.

While teams don’t draft based upon college statistics, players who have bigger games do tend to receive more publicity, which results in more exposure, and then potentially, a higher draft slot than would be expected or maybe deserved. What we can do is look at the different players who have come out of these conferences at the respective positions since 2000, and see if NFL teams are potentially overvaluing or undervaluing them.

These backs have not been particularly successful in the NFL. Dayne, Bennett, Duckett, and Perry were all busts — of the first-rounders, only Larry Johnson was a starting running back within his first three years, and while Johnson put up great numbers in 2005 and 2006, few people remember that the Chiefs were ready to cut bait on him in 2004. While I happen to think Maroney will be successful, the jury is still out on him. The real finds have been later picks like Ladell Betts and Marion Barber, but they’ve also only been platoon backs.

The similarity in draft value between these two lists is remarkable: 6844 points of draft value were spent on the Big Ten backs, while 6970 were spent on the Pac-10 backs. The latter has a big bust in Trung Canidate, but the Pac-10 yielded starting running backs in Steven Jackson and DeShaun Foster, the two most promising running backs in football in Reggie Bush and Maurice Jones-Drew, and a guy who was some marijuana away from being a starter in Onterrio Smith. The Pac-10 backs aren’t outclassing the Big Ten guys, but they’re the better group of running backs.

What if we take the same look at wideouts?

There are two guys on this list who are useful NFL receivers — Chad Johnson and T.J. Houshmandzadeh, both of whom went to Oregon State and were drafted by the Bengals in 2001. They’re much better than useful, actually. No one else on this list has evolved into much of anything except for, perhaps, Dennis Northcutt, who’s a third wideout and returner. The big busts are obvious and almost legendary at this point: R. Jay Soward, Freddie Mitchell, Reggie Williams, and Mike Williams are three washouts and a guy who was inactive in Week 1. On the other hand, that bust list is missing two players…

This is a list of more useful players, short the two big busts, Charles Rogers and David Terrell. I’ve written at length about Chris Chambers’ mediocrity in the past, but even he’s a better player than everyone besides Houshmandzadeh and Johnson on the Pac-10 list. There are almost a dozen Big 10 wideouts from the last six years better than any non-Bengals Pac-10 wideout.

Finally, the same comparison with quarterbacks:

The Pac-10 has notable busts in Tuiasosopo, Harrington, Boller, and perhaps Rodgers, but Palmer and Leinart. The Big Ten doesn’t really have much outside of the miracle that was Tom Brady, sixth-round pick, and Drew Brees. The difference, though, is that the Pac-10 produced multiple first-round picks at quarterback; the Big Ten didn’t have a single one (although Brees was close). There’s a possibility these Pac-10 quarterbacks could be overvalued because of the numbers they were racking up throwing the ball so frequently.

Of course, this isn’t a catch-all saying to never draft Pac-10 players involved in the passing game or Big Ten running backs. It’s merely an interesting data point in Marshawn Lynch’s favor. We’ll be revealing more findings in Pro Football Prospectus 2008, only ten months away!

By Bill Barnwell
bill@patriotsdaily.com

Since the videotape controversy was already discussed in a post on Football Outsiders (and Patriots Daily) this week, this week’s column will be a bit of a respite from Videogate.

The Chargers have been left with a conundrum regarding the redoubling of their skill position assets twice now in the last several years. First, Drew Brees emerged as an excellent NFL quarterback just as the Chargers nabbed the promising Philip Rivers in a trade with the Giants (that also netted them Shawne Merriman, in one of the great draft-day swindles of recent memory). When Brees’ contract expired, the Chargers let him walk away to New Orleans without any compensation. While Rivers has been mostly superb in his time as an NFL starter, Brees was an MVP-caliber performer last year. Had the Chargers dealt him after 2004, they could have recouped at least something for the player. Instead, general manager A.J. Smith decided that there was more value in Brees’ skills for a year as opposed to what he could’ve acquired in return in trade (whether it be draft picks and/or veteran players in addition to cap space).

Another somewhat similar situation brewed this past offseason, as Chargers backup RB Michael Turner was a restricted free agent. As per the rules of restricted free agency, the Chargers were allowed to tender him a contract at one of several levels. Turner received a $2.35 million tender, the highest available and accordingly, the one with the most compensation due the Chargers were a team to sign Turner away from them. The Chargers would have received a first- and third-round pick in exchange for Turner had he signed elsewhere; while there were trade talks around the running back, nothing was consummated, and Turner signed his tender offer and is continuing to backup LaDainian Tomlinson this year.

Now, while people may be aware that Michael Turner is a fine running back, allow me to point out how fine he actually is. Turner’s career numbers through three seasons and one week are astoundingly good; he’s averaged 5.9 yards per carry, exceeding Tomlinson’s yards per carry by more than a yard each full season.

That sort of argument, while a point in Turner’s favor, is a misuse of statistics. Namely, Tomlinson gets the ball in situations where he couldn’t possibly gain significant yardage (e.g. inside the five and in short-yardage situations). That’s where DVOA, our pet stat at Football Outsiders, comes in. DVOA measures how a player does each play versus what the league average is, accounting for down and distance, the yardage between the offense and the end zone, and the defense being faced. In addition, Tomlinson carries the ball much more frequently than Turner, which needs to be accounted for — there’s a benefit to having a back who can carry the ball 300 times at a rate better than a guy like Kevan Barlow. We define “replacement-level” as the level of a player the caliber of Barlow, freely-available talent, and then measure a player’s value on a per-play basis as opposed to that freely-available talent cumulatively over the course of a season.

Somewhat surprisingly, while Tomlinson’s DPAR obviously far outweighs Turner’s, Turner’s DVOA is still better than Tomlinson’s.

As you can see, while Tomlinson was the best back in the league on a cumulative basis last season according to DPAR, Turner was the best back in the league on a per-play basis. It’s, in fact, the best DVOA for a back with more than 50 carries in the eleven years we’ve got DVOA numbers for.

While DVOA is a more accurate indicator of performance than unadjusted yards, again, we see that Turner’s still performing better than Tomlinson on a per-play basis. There’s a second set of reasons for that, none of which are easily quantifiable or even provable. He gets more rest. He doesn’t start, and comes in when the defense is tired. The defense doesn’t prepare for him the same way they do for Tomlinson, and they don’t adjust accordingly. There’s probably some truth in those things, but there’s every reason to think Michael Turner’s a pretty good running back.

There’s a reason to think that he’s not, though, and his name is Richard Huntley.

Some of you may remember Huntley; he was a relatively nondescript running back for a few years in the NFL, playing for four teams. He was drafted by Atlanta in the fourth round in 1996, spent a year there, was waived, and then went to Pittsburgh, where he sat on the practice squad for a season. In 1998, he had 55 carries for 242 yards; decent numbers, but a product of context and full of lots of third-and-long useless yardage, as was reflected in his -35.2% DVOA. 1999, though, brought a new Richard Huntley to the fore. He gained 6.1 yards per carry on 93 carries, became a solid receiver in the rushing game, and put up a 21.9% rushing DVOA, good for fourth in the league. The Steelers responded by giving him a three-year, $4 million contract and a chance to compete with Jerome Bettis (who was in the final year of his contract) for the starting job in training camp. Bettis won out, but as if to prove his 1999 was not a fluke, Huntley’s rushing DVOA in 2000 rose to 30.3%, good for fourth in football. Bettis’ DVOA was significantly lower, at 10.7%, but that was still good enough for seventh in the league.

Huntley was cut after the season and signed with Carolina, where he would challenge the oft-injured Tshimanga Biakabutuka for the starting gig. In traditional fashion, Biakabutuka got hurt and Huntley was the starting back for most of the season. Huntley’s line: 165 carries, 658 yards, a -25% DVOA, and -6.2 DPAR — in other words, replacing Huntley with the best free agent you could find at the minimum salary would have been preferable to actually using him. Carolina went 1-15 that year, George Seifert was fired, and Huntley had three more carries in his career before he was out of football.

Huntley’s an example of something DVOA can’t account for, but a scout or an observer with access to DVOA can: The rest of the offense. As useful as DVOA is, it can’t account for the quality of an offensive line or a quarterback’s ability to put a running back in a good situation more often than not. It’s just impossible to get a first down on third-and-8 running the ball every time. Huntley played with a very good offensive line in Pittsburgh. He certainly did not in Carolina. In Pittsburgh, his quarterback was Kordell Ste…people thought he was good then. In Carolina, his quarterback was Chris Weinke. The situation in Pittsburgh was just more conducive to being a successful running back than in Carolina. DVOA can help strip some of that information out, but not all of it.

So, then, Huntley presents the cautionary tale in thinking that signing a successful backup from a pretty good rushing team is a good way to upgrade your running back spot. Of course, there are positive examples too. What we can do to more accurately ascertain whether the new running back was an improvement on the old one is to judge their DVOA versus that of the starter’s the previous year’s. While Huntley’s -25.0% DVOA was certainly abysmal, Biakabutuka’s DVOA the year before was an even more abysmal -26.1% DVOA. We track a statistic called “Success Rate” which measures what percentage of the time a running back has a “successful carry” — meaning he got a chunk of yardage significant enough to push his team closer to a first down than closer to losing the ball. The average starting running back succeeds about 43% of the time. In 2000, Tshiminga Biakabutuka succeeded 20.5% of the time. That’s astoundingly poor. Huntley got that up to 39%, which is at least remotely close to competency.

Perhaps the best way to see what teams might be getting into if they sign Michael Turner after this season is to look at the performance of teams who have acquired another team’s backup running back (defined as the back with the second-most carries on the team) and made them their starters. We’ll then track their performance the year after and compare it both to how they performed the year prior and how they did relative to the back they replaced. Since we only have DVOA and DPAR for seasons from 1996-2006, this analysis will only include backs from those seasons.

1. Lamar Smith (1997-98 and 1999-00)

Lamar Smith was one of two backs (both of whom were named Smith) to pull this move twice in his career. Starting off his career with Seattle, Smith was an average back by all measures — he averaged right around 4.5 yards per carry, and his DVOA in 1996 and 1997 was right around there: 3.2% and 0.2%. Nothing awful by any means, but not a star about to breakout, either. The Ditka administration in New Orleans brought him in to replace the awful Ray Zellars-Mario Bates hydra of suck, and Smith only made it to 138 carries. His DVOA, a miserable -38.6%, was actually worse than Zellars’ -34.2% the year before, and way behind Bates’ -19.0%. The result? Mike Ditka traded his whole draft for Ricky Williams (who put up a -30.6% DVOA his rookie year), and Smith returned to his backup role for a year.

Smith then moved onto Miami, where Jimmy Johnson was fooling around with Cecil Collins, J.J Johnson, and the last vestiges of Karim Abdul-Jabbar’s career. He made Smith his starter and Smith proceeded to not only bump his DVOA back up to a very respectable -1.6%, he had one of the great out-of-nowhere fantasy season of all-time, with 309 carries for 1139 yards and 16 touchdowns. He improved on Collins’ -18.5% DVOA dramatically. A year later, he averaged only 3.1 yards per carry, and in 2002, Ricky Williams replaced him again.

2. Charlie Garner (1998-99)

Garner’s mercurial career saw him start off as the backup to Ricky Watters in Philadelphia; when Watters left, it seemed likely that Garner would get the starting gig, but he was beat out by Duce Staley. Garner took the opportunity to replace the injured Garrison Hearst in San Francisco, and had an excellent year on a terrible 49ers team (this was the year that Steve Young went down at the hands of Aeneas Williams in Week 3). Garner’s DVOA went from -15.9% to 7.6%, and he kept his DVOA pretty close to Hearst’s 1997 performance of 9.4%. On a significantly worse team (the 49ers went from 12-4 to 4-12), it was an admirable performance and the beginning of Garner’s short career as a starting running back.

3. Ahman Green (1999-00)

Very few people, by this point, remember Green starting his career in Seattle. After being drafted in the third round in 1998, he spent two years struggling to get a foothold on any significant playing time, and was then dealt to the Packers for CB Fred Vinson, a second-round pick the year before. Vinson never played a game for the Seahawks. Green gained over 8,000 yards for the Packers. Oops. Green’s DVOA as a Seahawk in 1999 was good (21.4%), but it was gained on only 26 carries and that’s not enough of a sample to be of any reliability. As a Packer, his DVOA was a solid 6.9%, which was a significant improvement over Dorsey Levens’ -13.5% DVOA the year before. It’d be hard to find any way to say this wasn’t an incredibly successful move for both Green and the Packers.

4. Priest Holmes (2000-01)

The big one. Holmes had a strange start to his career. His first year with any playing time, he ran for 1008 yards and gained 4.3 yards per carry. The next year, he was hurt and only played eight games, but gained a ridiculous 5.6 yards per carry, good for a 22.9% DVOA, second in the league. The Ravens drafted Jamal Lewis and won the Super Bowl in 2000; Holmes backed him up and had a 12.5% DVOA, fourth in the league. If Football Outsiders had been around then, Holmes would have been one of the players we championed the way we do Jerious Norwood now.

Holmes went on, of course, to gain 1555 yards behind a great offensive line in Kansas City the next year. His DVOA that season was 20%, a slight improvement over the criminally underrated Tony Richardson’s 18.1% the year before. It was the third-best DVOA in the league that year, though, and Holmes would be in that rarified air until he got hurt in 2005. He’s the inverse Richard Huntley.

5. Terry Allen (2000-01)

The running back who would not die, Allen came back from injuries to have seemingly three careers. He was solid (4.5% DVOA) as Ricky Williams’ backup in New Orleans, and when Jamal Lewis went down for the season in 2001, with Holmes gone, Allen became the starter in what was his last NFL season. His DVOA wasn’t very good at -5.8%, but it wasn’t a disaster, either, when you compare it to Lewis’ 5.5% the year before.

6. Antowain Smith (2000-01, 2004-05)

The other Smith who pulled this move twice, Antowain lost his job to Travis Henry in Buffalo and was one of the thousands of veteran refugees brought in by Bill Belichick for the magical 2001 season. His DVOA stayed remarkably similar to his performance in Buffalo, going from -9.5% to -10%, but it was a ways better than Kevin Faulk’s -20.7% the year before, or J.R. Redmond’s -19.9%. I think everyone here knows the Smith story well enough.

In 2004, he rose again! After being let go by the Patriots, he ended up in Tennessee, where he backed up Chris Brown. He made his way the year after to New Orleans in its Katrina-riddled season, where stats are pretty irrelevant. Again, his DVOA stayed remarkably consistent (-12.7% to -12.5%), and while he was slightly worse than Deuce McAllister had been the year before, there’s no real predicative value to the 2005 Saints season.

7. Richard Huntley (2000-01)

8. Trung Canidate (2001-02)

Canidate was the speed demon and late first-round pick who was supposed to make the terrifying Rams offense even more so, but he never really made an impact. He barely played in 2000 and 2002, although he was fantastic in 2001, when he averaged 5.8 yards per carry and his 24.1% DVOA was best in football. The Rams shipped him to Washington, where he served as the starter in Steve Spurrier’s offense, splitting carries with Rock Cartwright and Ladell Betts. He actually was an improvement on Stephen Davis’ -8.4% DVOA, putting up a respectable -2.2% DVOA, but he couldn’t take the workload of the starting role, Washington traded for Clinton Portis, and Canidate never played again.

9. Warrick Dunn (2001-02)

Dunn wasn’t a backup inasmuch as he was a complimentary back to Mike Alstott. He was the featured back from 1997-98 and in 2000, but was technically the backup by seven carries in 2001 by virtue of missing three games. He went to Atlanta and was a big improvement over Maurice Smith, who also had one year as a starter and never played again.

10. Amos Zereoue (2003-04)

Zereoue was hyped as a good running back while backing up Jerome Bettis, but the numbers don’t show it. His best DVOA was -9.4%, and he was usually closer to -20%. He never had an above-average success rate, and was eventually dumped off to Oakland for the first year of the Norv Turner era. Replacing Wheatley (5.3% DVOA), Zereoue’s -14.1% DVOA did not help the Raiders’ many problems one bit. He made it to three games for the ‘05 Patriots and was done.

11. Thomas Jones (2003-04)

Jones is another guy who had a strange career dictated by context; he went from first-round bust in Arizona to backup material in Tampa Bay to successful back on a Super Bowl team in Chicago. He wasn’t particularly effective in Tampa Bay, putting up a -12.4% DVOA while backing up Michael Pittman (-2.8% DVOA), but the Bears acquired him to supplant the A-Train, Anthony Thomas, who had actually put up a reasonably close to average -3.8% DVOA the year before. Jones got all the way up to -2.3% DVOA before really hitting his stride in 2005.

12. Corey Dillon (2003-04)

Another well-known story to readers of this site. Dillon’s -5.8% DVOA in Cincinnati in 2003 wasn’t fabulous by any means. He got it up to 21.2% in 2004 in New England, supplanting Smith, who was also at -5.8% in 2003.

13. Lamont Jordan (2004-05)

For my money, the closest comp to Turner. Jordan put up some extreme DVOAs as a Jet: in 2001, he was at 46.6% on 39 carries; in 2002, -40% on 84. 2003 saw him actually be average, with a 4.2% DVOA on 46 carries, and then back with 93 carries in 2004, his final year as a Jet, his 36.6% was second-best in the league. By comparison, Curtis Martin’s DVOA that year was 20%, good for eighth in the league, but he led the league with 54.9 DPAR.

The salary cap-strapped Jets had no way to keep Jordan to backup Martin, even though Martin was old and likely to breakdown the next year (which he did). Jordan moved onto Oakland, which gave him a five-year, $27.5 million deal. As you might remember, he was not the best back in football again. His DVOA dropped down to 1.5%; this, however, was still an improvement on Zereoue’s -14.1%. He was hurt for almost all of 2006.

14. Chester Taylor (2005-06)

Another Raven who flew the coop. He was no Holmes, as his -8.1% DVOA as a backup in 2005 was by no means spectacular. With Minnesota needing a running back after Whizzenation, they signed Taylor and gave him the bulk of the carries behind their restructured offensive line. Taylor wasn’t particularly great for a guy who gained 1,216 yards. His -9.4% DVOA was worse than he’d been in Baltimore, and was a step down from Mewelde Moore’s 1.9% DVOA the year before. Taylor was more durable, though, which is why he got the rock.

So, then, combining empirical judgement of the results along with some of my own opinions, I count eight improvements, five declines, and three relative washes. Of course, a back with the success of Turner isn’t really comparable to the failure Ahman Green was thought to be, or the starter Warrick Dunn was. In more likelihood, you can classify Turner in a group of elite backups alongside Jordan, Holmes, and Huntley. Those players went to three teams with very different offensive lines and played accordingly. Turner’s career, and his future as a star back, depends upon the same.

By Bill Barnwell
bill@patriotsdaily.com

Hello! Welcome to the year’s first “Outside Foxborough”. I’ll be your host, Bill Barnwell. Although the column has a new name, this is the second year I’ll be penning a weekly column looking at the Patriots from the Football Outsiders perspective.

For those of you unfamiliar with our site, Football Outsiders is a website dedicated to analyzing the NFL through viewpoints not seen anywhere else. Primarily, that’s through our work in developing new metrics and means of judging performance that are more accurate than established measures like yards. Our primary metric, DVOA, measures performance against the league-average while accounting for context in ways that those traditional measures don’t, and are a better predictor of wins than the previous year’s performance. Now that I’ve explained all that, I’ll say that this week’s column has nothing to do with DVOA. In the future, I’ll explain DVOA and our statistics further when they come into play. Although our work has been featured in places like ESPN The Magazine and The New York Times and on FOXSports.com, if you enjoy these columns and want to read more of our work, I’d recommend you go out and purchase Pro Football Prospectus 2007, our annual that features essays on each team, original research looking at those teams, player and team projections, fantasy football information, and everything else you’ll need by your side come Sunday.

The archive of last year’s columns is available here; among other things, I looked at whether Deion Branch and Adam Vinatieri were worth the money, how wide receivers from different college conferences performed in the pros, the historical injury rates suffered by the Patriots defensive backs, and whether running quarterbacks age faster than pocket passers. Several of the essays were expanded and updated for PFP 2007.

This week’s column, though, was inspired by Cut Day, and what the Patriots did to get to the 53-man limit; namely, remove all their Day Two (Rounds 4 to 7 of the NFL Draft) draft picks short DT Kareem Brown, who might also have been cut had Richard Seymour been ready to play. While three of the players went to IR, the Patriots also cut four more players. This wouldn’t be surprising for some teams, but the Patriots have been known for their drafting acumen during the Belichick era. A look at the players they selected between 2002 and 2006 on the draft’s second day reveals several key Patriots that were found on Sunday in April instead of Saturday:

Not only did the Patriots find stars like Asante Samuel and Dan Koppen, but useful role players like Tully Banta-Cain, Jarvis Green, David Givens, Ryan O’Callaghan, and Willie Andrews have all came out of the later rounds of the draft. So it would seem to be a surprise, then, that the Patriots were only able to find one active roster spot for their Day Two picks this year. (G Clint Oldenburg has already been signed onto the practice squad.) While this is a deep, veteran team, it isn’t exactly a new situation for Belichick and Pioli.

Let’s evaluate Day Two, though, and see how the Patriots have done there. It’s sort of a forgotten day in the modern draft cycle; it gets sent to ESPN2, where most people don’t watch it, 99% of mock drafts don’t talk about it, and it’s seen as somewhat of a crapshoot for most, if not all teams. Of course, it’s precisely these areas seen as crapshoots or otherwise underappreciated that smart franchises like the Patriots exploit to their advantage.

To look at how teams did on Day Two, I went through the five drafts that took place between 2002 and 2006 and evaluated the results of each of the 799 players taken as measured by their games played and games started in the years following their selection. Games played and started were counted for all teams, not just the team who selected the player. You’ll also note the “Value” column above — that stands for Draft Value, as determined by the generic NFL Draft Value Chart, which you can find here. While the specific charts vary, the chart quantifies the value of a pick and assigns each a numerical value, ranging from 3000 to the first overall pick all the way down to two for the 224nd pick. (Since each of the drafts had more than 224 picks, I assigned all picks following #224 zero points.) The chart’s origins are attributed to Jimmy Johnson, and while it stands as a decent generic base for judging potential trades, I’m sure the Patriots’ chart is different from the one being used here.

The Games Played chart is led by the team that I believe has the best scouting of any team in the NFL right now, even if they don’t often receive the credit for it they deserve. Note that the “2002″ column, for example, measures the number of games played over the past five years by people who were selected in that draft, while “2006″ measures only those players who were selected in 2006.

The Baltimore Ravens franchise is impeccable at finding talent in the lower rounds and integrating them into their lineup (Adalius Thomas comes to mind). Amongst the other top teams, you see a mix of awful teams with nothing to lose in Tennessee and San Francisco, but then a series of relatively successful franchises, including the Patriots.

Indianapolis also stands out as a franchise who have often used Day Two picks, especially on defense, to fill in gaps. Keep in mind that this is strictly games played, so all a player needs to do is play one snap to count as much as someone who plays the full game. Matt Cassel has eight games played in his career. To see whose draft picks are really making a difference, we’ll need to look at the games started figures.

Again, the same mix of teams is near the top, even though the Ravens drop slightly. The 49ers in 2002 were the last team of the Steve Mariucci era, and the number of Day Two starts from that draft is due less to the quality of the players and more to the confluence of salary constrictions and the failure of both the Mariucci and Dennis Erickson eras to draft successfully on Day One. This is the franchise that chose Mike Rumph, Kwame Harris, and Rashaun Woods with successive first-round picks. The Colts and their Cover 2 guys are second, with players like David Thornton, Robert Mathis, Cato June, Jason David, and Antoine Bethea all stepping into starting roles.

Chicago, meanwhile, has Alex Brown, Ian Scott, Nathan Vasher, Kyle Orton (hey, he started 15 games), and Mark Anderson in their haul.The teams at the bottom of the list are also a mix of surprises and usual suspects. The Broncos are generally thought of as a team that drafts well, but all they have to show from Day Two over the timeframe is Elvis Dumervil, Jeb Putzier, and the argument against you being able to gain 1000 yards behind Denver’s offensive line, Quentin Griffin.

Pittsburgh’s also been famed for their drafting abilities for decades, and they unearthed contributors like Larry Foote, Brett Keisel, and Verron Haynes in 2002; after that, though, the well has gone dry. Chris Kemoeatu is a depth guy at defensive tackle, and Noah Herron got some carries in Green Bay, but no one else has really made the breakthrough into being a steady NFL player. On the other hand, Washington’s well-known for both trading away draft picks and making sure it has a very famous starting 24 with nothing else behind it, while Oakland and Detroit are the teams you associate with futility in everything.

The interesting thing here, and what surprised me when I was doing the research, is that there’s no consistent “year-after” effect where a team in salary cap hell or otherwise awful starts 15 second-day picks for two years in the hopes that a bunch make it through. There’s no correlation between a team’s wins the year before and the number of games played or started by their Day Two picks the year after, nor is there one between Day Two picks starting and the team’s wins that season.

Now that we’ve seen which teams are successful, we can look and see which positions have the best return on value.

Generally, we see running backs, receivers, linebackers, and defensive backs offering the best return on value, while quarterbacks and linemen tend to lag behind. Well, there’s a reason for that; those four positions make up the majority of special teams players, and many teams grab their special teamers from Day Two. Again, we’ll have to look at games started to get a clearer picture.

The first thing that stands out is those fourth-round tight ends! No wonder the Patriots went after Garrett Mills. Justin Peelle, George Wrighster, Darnell Sanders, Robert Royal, Donald Lee, Bo Scaife, Putzier, Courtney Anderson, and Owen Daniels have all seen plenty of playing time as starters, while Randy McMichael has started every game of his professional career. Fifth-round offensive linemen include Koppen, David Diehl, Tony Pashos, Jacob Bell, and Jake Scott. The clear trend is that late-round defensive players are more likely to be successful than offensive players (tight ends excluded), with skill-position players in particular lagging behind their defensive counterparts.

As you can see from the games started chart above, the numbers do seem to trend downwards as the draft goes along, which indirectly answers another question: Is there a point at which the draft becomes a crapshoot, and picks in one round are just as valuable as those from another round? The chart indicates that the answer is no, and it’s easy to confirm by checking for a correlation between draft value and both games started and games played.

The correlations are both positive (-.28 for games played and -.23 for games started) across all draft picks; if you remove the fourth round and just run the correlation for the fifth round on, the correlations are weaker, but still trend positive. The same is true for the sixth round. This means conventional wisdom is true; the draft becomes more of a crapshoot as it goes deeper, but a fourth-round pick is still much more valuable than a seventh-rounder.

So, with all this data compiled, we can finally look and see what we’d expect the average Day Two class to do in its rookie year. The result? 30.2 games played and 10.5 games started. It’ll be hard for Kareem Brown to accomplish that all by himself, no matter how long Richard Seymour is out for.

Bill Barnwell’s ‘Outside Foxborough’ will appear every Thursday. We welcome your comments here.