The Inches We Need is now available on Amazon!

Kindle Version

Paperback

Join the mailing list!

Subscribe to The Inches We Need mailing list

* indicates required

Intuit Mailchimp

NESSIS 2023

On Saturday, I attended the New England Symposium on Statistics in Sports (NESSIS). Compared to the Sloan Sports Analytics Conference, I found NESSIS much smaller, focused, and academically rigorous. The presentations revealed fascinating dimensions to analytics in a variety of sports, and the small size afforded me the opportunity to speak with many of the presenters. I even talked to some statistical legends I've been following for years. I will certainly be returning; unfortunately NESSIS only occurs every other year.

Ryan Brill, a PhD student at University of Pennsylvania, stole the show with his talk "Analytics, have some humility: a statistical view of fourth down decision making" (Abraham Wyner also credited on the paper). He revealed some troubling flaws in widely-circulating fourth-down calculators based on Expected Points Added (EPA) and Win Probability Added (WPA). Specifically:

  • Good teams run more plays in the red zone than bad teams, so EPA calculations based on statistically averaging without regard for team quality overweight the chances that good things will happen for the offense. Accordingly, fourth-down calculators based on EPA also show this flaw.
  • While there are hundreds of thousands of plays that EPA / WPA models pull from, WPA formulas autocorrelate to one outcome (win or lose), drastically reducing the data set (Brill estimates to about 4000 data points over 15 years). Accordingly the error bars are much higher than fourth-down / win probability calculators want to admit, an 8% Win Probability Interval on average. Fourth down bots should be a lot less confident.

There were a lot of other great talks, and I filled a notebook with terms to look up and try to understand better: Extreme Gradient Boosted Regression Trees, Markov Chain Monte Carlo, Bayes Additive Regression Trees, Random Forests for Conditional Density Estimation, etc. By the time 2025 rolls around, I'll understand more of this stuff. Probably. Either way, I found both the presentations and the conversations inspiring, and I'm looking forward to learning more.

 

Most draft capital on edges since 2017

Another discussion arising from the Scho Show slack channel: which teams have used the most draft capital on edges since 2017?

I've compiled a list of all edges drafted in the top 200 since 2017. The columns include number of players drafted, projected Approximate Value per the Chase Stuart draft chart (AV), and points according to the Jimmy Johnson draft chart (JJ):

 CountAVJJ
JAX791.05732
SFO468.05114
LVR673.43972
WAS457.23748
CLE555.83532
DET452.53355
NOR655.62581
NYG653.82567
DAL859.32520
IND864.02496
HOU542.02464
DEN548.72451
GNB443.02432
NYJ547.32214
SEA855.02198
ARI548.82160
ATL650.52064
CAR750.82033
BAL752.21916
PHI640.21833
MIA435.91737
CIN852.51622
NWE747.41432
KAN532.21300
BUF330.51272
TAM424.8863
LAC323.3838
PIT321.5752
TEN316.4553
LAR721.1348
MIN414.5251
CHI35.770

Tags

Precedents for Sam Howell

The Washington Commanders showed little inclination to use quarterback Sam Howell during his rookie season. Consigned to third string behind veterans Carson Wentz and Taylor Heinicke, he only saw action in one game, the season finale. Washington won that contest, an impressive 26-6 victory over division rival Dallas Cowboys, and Howell performed credibly enough for a rookie. His 19 passes sold the Commanders braintrust, who added only seasoned backup Jacoby Brissett and did not select a quarterback in the draft 2023.

Howell had once been a highly-esteemed college prospect, but by the time the draft rolled around, he slipped to pick 144—the first pick in the fifth round. Few late-drafted quarterbacks develop into starters, and even fewer get to start Week 1 with little in the way of competition. The Commanders’ gambit may pay off—or prove disastrous. The franchise seems to be in a holding pattern with new owner Josh Harris transitioning in, but Howell will get a shot to start some games and prove whether he can be the quarterback of the future, for a team and a city that could use one.

So far, so good, as the Commanders have started the season 2-0. Howell hasn’t lit up the scoreboard, but he’s thrown three touchdowns against just one interception. Still, with such a small sample size, we’re better off looking at past precedents for his likelihood of success than digging into Howell’s statistics.

Quarterbacks of Little Renown

Unheralded quarterbacks have started games in the NFL throughout history. Tom Brady famously was pick #199 back in 2000. Tony Romo had a long and distinguished career after entering the league as an undrafted free agent. Brock Purdy, taken with the last pick in the 2022 draft, will start for the San Francisco 49ers this year.

The difference between Howell and these players is that they didn’t enter the season as the #1 option until they had already established some baseline level of NFL competence. Brady entered 2001 as the backup to Drew Bledsoe and only took over when a chest injury put Bledsoe out of commission. The Patriots won the Super Bowl and Brady never relinquished the job. Romo was also a backup who took over, coincidentally also for Bledsoe, when Bledsoe was struggling near the end of his career. Purdy began 2022 as the third-string quarterback, taking over only after injuries to Trey Lance and Jimmy Garoppolo. By the time Brady and Romo entered a season as the starting quarterback, they had bodies of work. Even Purdy, counting the playoffs, attempted 233 NFL passes entering the 2023 season; Howell, just 19.

There have been less distinguished Week 1 starters, but few have been by design. Trevor Siemian started in Week 1 for the 2016 Denver Broncos, but he had to beat out first-round pick Paxton Lynch. Kurt Warner started an improbable Hall-of-Fame career Week 1 of 1999 with the St. Louis Rams, but that was only after Trent Green suffered a preseason ACL tear. Dennis Dixon started Week 1 of 2010, but only after the league suspended starter Ben Roethlisberger and backup Byron Leftwich tore his ACL.

We’ve also seen many stopgap starters, undistinguished veterans starting Week 1 before inevitably giving way to a highly-drafted rookie. That was likely the intention with Siemian, but Lynch fared so poorly that he never took over. Longtime backup Doug Pederson started early on for the 1999 Eagles to ease the pressure on second-overall pick Donovan McNabb. That same year, Shane Matthews started the year for the Chicago Bears to give 12th pick Cade McNown more time to develop. Nathan Peterman started Week 1 for the 2018 Buffalo Bills, but the team inserted first-round pick Josh Allen that game was even over. Tom Savage similarly started Week 1 for the 2017 Houston Texans, but he was only a speedbump on Deshaun Watson’s road to success, lasting only a half before the rookie took the reins.

So who are actually precedents for Howell, players who were neither highly-drafted nor game-experienced? There are only a few:

  • Tyrod Taylor made his first career start in Week 1, 2015 for the Bills. He faced stiffer competition than Howell, however, having to beat out former first-rounder E.J. Manuel and experienced starter Matt Cassel. But the bloom was off Manuel’s rose by that point, so I distinguish this instance from the stopgap situations listed above. Taylor wound up making the Pro Bowl and starting three years for the Bills; he remains a quality backup.
  • The Texans traded two second-round picks to the Atlanta Falcons for Matt Schaub and inserted him as the 2007 starter. Schaub had started only two games for Atlanta. Schaub would start seven years in Houston, making two Pro Bowls.
  • Tim Rattay, a seventh-round pick in 2000, got the starting nod in 2004 after acquitting himself competently in three spot starts the year prior. He did not fare as well in more extended use, throwing 10 interceptions in nine starts and taking a whopping 37 sacks while completing barely 60% of his passes. The 49ers finished 2-14 and took Alex Smith with the first pick of the 2005 draft, dealing Rattay to the Tampa Bay Buccaneers.
  • Chris Redman, like Howell a highly-touted recruit that fell to the middle rounds (he was a third-round pick in 2000), started the first six games in 2002 for the Baltimore Ravens. The team averaged only 17 points in his starts and the team benched Redman for veteran Jeff Blake. Blake didn’t fare much better, and in the next draft the Ravens traded up for Kyle Boller, who also didn’t impress. Redman would bounce around as a backup for most of the next decade, making a handful of starts.
  • Matt Hasselbeck was a former sixth-round pick who had only thrown 29 NFL passes when the 2001 Seattle Seahawks made him the opening-day starter. Seattle had made something of an investment in Hasselbeck, however, dealing a third rounder and swapping seven spots in the first round to obtain him from the Green Bay Packers. Seahawks coach / GM Mike Holmgren had worked with Hasselbeck in Green Bay. After some early struggles, Hasselbeck would go on to a distinguished career, including three Pro Bowl appearances and a Super Bowl berth.
  • Jay Fiedler had started only one game before the Miami Dolphins inserted him as the Week 1 starter in 2000 (and heir to Dan Marino). He did have to beat out Damon Huard, who had started a handful of games for the Dolphins the year before. Fiedler never won accolades but served as a quality starter for five seasons.
  • Brian Griese, like Fiedler, took over for a legend: Denver Broncos quarterback John Elway. A teammate of Brady’s at Michigan, Griese was a third-round pick in 1998 but threw just three passes that season. The team elected to insert him as their opening day starter in 1999 after Elway’s retirement, and he went on to start 83 games over the next 10 years with four different organizations. The results were mixed, but he did make a Pro Bowl in 2000.

In addition to Howell and Purdy, mentioned above, the Atlanta Falcons are giving 2022 third-round pick Desmond Ridder the starting nod in 2023 after he played only four games in his rookie season. 

Mid-Round Pick Precedent

The data clearly shows that quarterbacks hit less often the deeper you get into the draft (see chart), and things look downright bleak in the area Howell was selected. But there’s more to the story.

Quarterbacks picked first overall "hit" more than 50% of the time; quarterbacks picked where Howell was selected (101 to 200) only about 3%.

 

From 1998 to 2020, teams selected 102 quarterbacks between picks 100 and 200, the range Howell landed about in the middle of. Only ten of those passers had multiple seasons with 2,000 yards passing, a modest total. Of those ten, only three (Brady, Kirk Cousins, and Dak Prescott), had multiple seasons where they passed for more than 6.5 Adjusted Net Yards Per Attempt, a solid figure (2022 league average was 5.9 ANY/A). Ten of 102 becoming starters and only three becoming good starters—those are long odds.

But we must also consider Bayes’ Theorem, which Wikipedia defines as a technique “used to update the probability for a hypothesis as more evidence or information becomes available.” And while there isn’t a lot of information available with Howell, he has already cleared some of the barriers to success. 

Take, for instance, Tyler Wilson, a fourth-round pick of the Oakland Raiders in 2013. Like Howell, Wilson was a once-heralded college prospect whose stock fell. Unlike Howell, he played so poorly that he could not even win the third-string quarterback job—on a Raiders team that finished 4-12! He never threw an NFL pass. We already know Howell, by virtue of sticking on an NFL roster for a year, is better than Wilson, so it doesn’t make sense to include Wilson in the comparison population.

With one career start, Howell already has distinguished himself from much of the group; of quarterbacks picked 100 to 200; only 54 of the 102 ever started an NFL game. And If we assume Howell will start at least eight games, about half a season (and perhaps not a safe assumption), we’re looking at just 28 passers. Out of a set of 28, getting 10 multi-year starters and three quality players doesn’t look too shabby. Add in that Pro Bowlers Hasselbeck and Marc Bulger fell just shy of the ANY/A criteria, and Howell might have a fighting chance here.

Scenario: Yeah But Howell is Awesome!

Howell impressed his first two years at North Carolina, throwing 68 touchdowns against 14 interceptions and improving his completion percentage (61% to 68%) and yards per attempt (8.6 to 10.3) from his freshman to sophomore season. He looked poised to cement a high draft selection after a strong junior season, but things didn’t play out the way he wanted.

Top receivers Dyami Brown (now a Washington teammate) and Dazz Newsome left for the draft, as did the backfield tandem of Javonte Williams and Michael Carter. The Tar Heels slipped from an 8-5 record in 2020 to just 6-7 in 2021. Howell’s production dropped off, too—his 63% completion rate and 8.8 yards per attempt resembled his freshman figures rather than his sophomore improvement, and his 24 touchdowns and nine interceptions represented career worsts. His prospect star dimmed somewhat.

Still, most expected Howell to hear his name called sooner than pick 144. Arif Hasan compiles a consensus big board, crowd-sourcing the opinions of several draft pundits, and Howell ranked 56th, which would put him in the back half of the second round. If we assume the Commanders got a bargain in Howell and he should be more rightly compared to earlier-drafted quarterbacks, what would that mean for his chances of success?

That assumption does change things, but maybe not by as much as you’d think. Higher picks get much more opportunity than lower picks, but only convert that opportunity into success at a slightly higher rate.

RangeNumber1+ GS8+ GSMultiyear StarterQuality Starter
100 to 2001025428103
65 to 9928241862
33 to 6423211873
16 to 3219191794
16 to 99706453229

 

The group from 65 to 99—roughly the draft’s third round—shows virtually identical rates to the 100 to 200 group. If we imagine Howell should have been a second-rounder, as the consensus board suggests, his odds increase a little more. Players taken in the latter half of the first round meet with more success, but that’s lofty for Howell. And even there, the general pattern holds. Note that these are all small sample sizes.

Only six of 70 quarterbacks selected between 16 and 99 failed to start an NFL game, and only 17 didn’t get to start at least eight. Opportunity drives the vast majority of the differences in outcomes shown in the first chart in the article. Howell’s going to get his opportunity. He’s not a great bet to succeed—but he’s not a significantly worse bet than any other unproven player.

Rookie Contract Value

The other day I discussed whether the New England Patriots had a good foundation with some Scho Show Slack compatriots and I opined that their rookie contract talent probably ranked in the 20’s somewhere. It occurred to me that I had the data to answer this question.

This isn’t strictly rookie contract players as I’m using first-rounders from 2019 (who could or might be on their fifth-year option in 2023) and draft picks from 2020-2022. The most recent draftees haven’t played enough to factor in here. I’m only capturing draft picks, not undrafted free agents. And I’m relying on Pro Football Reference’s Approximate Value statistic, for better or worse.

RankTeamAV
1NYG193
2CIN193
3ATL188
4MIA183
5BAL180
6DAL177
7JAX174
8MIN173
9DET168
10PHI161
11CAR161
12WAS156
13SFO153
14PIT153
15GNB152
16LAC149
17DEN149
18BUF147
19NYJ146
20TAM145
21KAN141
22LVR138
23ARI136
24CHI134
25NWE122
26IND122
27HOU115
28SEA115
29CLE109
30TEN94
31LAR75
32NOR67

The New York Giants surprised me by ranking tied for first, but they’ve made a whopping eight first-round picks during the time period. Even with Deandre Baker flaming out (just 6 wAV), their earlier picks from 2019, Daniel Jones (38 wAV) and Dexter Lawrence (31 wAV), have carried a lot of the weight. Perhaps that’s a flaw in the methodology; older picks factor in more heavily because they’ve had more time to accrue AV, while arguably they are less valuable because they are approaching the end of their rookie deals (Jones already signed a big extension).

The Cincinnati Bengals also had a great run, but more expectedly. Drafting Joe Burrow (38 wAV), Tee Higgins (26 wAV), and Ja’Marr Chase (23 wAV) will do that. The Atlanta Falcons are more unexpected at number three; like the Giants, they are buoyed by multiple 2019 first-rounders (Chris Lindstrom, 26 wAV, and Kaleb McGary, 28 wAV).

If we exclude the 2019 picks, the Bengals hold at #2, but Dallas leapfrogs them to #1, keyed by Micah Parsons (36 wAV), CeeDee Lamb (32 wAV), and Trevon Diggs (27 wAV). The Falcons slide to 10th, while the Giants drop all the way to 18th.

At the other end of the spectrum we have two teams who haven’t made a lot of picks. The Saints didn’t have a first-rounder in 2019 and made just 15 total picks between 2020 and 2022. The Los Angeles Rams famously haven’t made a first-round pick since 2016. The Tennessee Titans have made a full complement of picks, but the disastrous choices of Isaiah Wilson (0 wAV) and Caleb Farley (1 wAV) have hurt. They finish last in this metric if we exclude 2019 first-rounder Jeffery Simmons (37 wAV).

Pats Nation Network - Podcast Apperance

Got a chance to have an extended conversation with my friend Matt St. Jean of Pats Nation Network. If you like this conversation, you're going to love the book.

Book Release Day!

The Inches We Need: How NFL Teams Find their Competitive Edges by Dave Archibald

The book is out! Check out the Kindle version. Paperback will be up as well shortly. I'm so excited to be able to share it with all of you.

Locked On Patriots - Podcast Appearance

Had a lot of fun chatting with Mike D'Abate on the Locked On Patriots podcast.

Kindle book releases tomorrow, Friday September 1! Watch for the paperback shortly thereafter.