113 Comments

Advanced Defensive Stats: Coverage

| July 9th, 2020

I’m continuing to look at Chicago’s defense using advanced defensive statistics from Pro Football Reference (PFR). I already looked at missed tackles, and today I want to look at coverage.


Baseline Rates

There are a whole host of advanced coverage stats available, including completion percentage, yards/target, target depth, yards after catch allowed, TDs, INTs, and passer rating. In order to keep it simple, I’m going to look only at yards/target, as that is a good baseline metric for how effective teams are when targeting a player. I’m intentionally not looking at passer rating because that gets skewed by touchdowns and interceptions, which are notoriously random statistics within a small sample size like this.

I compiled all yards/target stats from the PFR database for 2018 and 2019, the only 2 years it has, and sorted them by position. In order to compare starters to starters and avoid rates skewed by backups, I assumed a base nickel package of 4 defensive linemen (DL), 2 linebackers (LB), 3 cornerbacks (CB), and 2 safeties (S). For all 32 teams over a 2 year span, this would mean 128 LB, 192 CB, and 128 S. This gave thresholds of 30 targets for LB, 40 for CB, and 20 for S.

Looking at those sample sizes, you can see the spread of missed tackle rates in the table below for each position group.

Read More …

Tagged: , , , ,

246 Comments

Advanced Defensive Stats: Missed Tackles

| July 1st, 2020

I’ve written quite a bit about Chicago’s offense so far this offseason, but not as much about the other side of the ball. I want to change that in the next series of articles, using advanced defensive statistics from Pro Football Reference (PFR). We’ll start today by looking at missed tackles.


Baseline Rates

Let’s start by establishing a baseline for what is a normal rate of missed tackles.

I compiled all missed tackle stats from the PFR database for 2018 and 2019 (the only 2 years it has) and sorted them by position. In order to compare starters to starters and avoid rates skewed by backups, I assumed a base nickel package of 4 defensive linemen (DL), 2 linebackers (LB), and 5 defensive backs (DB). For all 32 teams over a 2 year span, this would mean roughly 256 DL, 128 LB, and 320 DB. This gave thresholds of 20 tackles for DL, 60 for LB, and 40 for DB.

Looking at those sample sizes, you can see the spread of missed tackle rates in the table below for each position group.

Read More …

Tagged: , , ,

207 Comments

Ryan Pace Has Gone All-In on the 2020 Season

| June 24th, 2020

After a disappointing 8-8 season, Ryan Pace moved aggressively this off-season to revamp the Bears for 2020.

On defense, he re-signed Danny Trevathan, upgraded Leonard Floyd with Robert Quinn, signed Tashaun Gipson as a cheap replacement for HaHa Clinton-Dix, and drafted Jaylon Johnson to replace the aging Prince Amukamara.

On offense, he traded for Nick Foles to compete with upgrade Mitchell Trubisky, replaced oft-injured veterans Taylor Gabriel, Kyle Long, and Trey Burton with Ted Ginn, Germain Ifedi, and Jimmy Graham, and drafted Cole Kmet to hopefully give Chicago their first long-term solution at tight end since Greg Olsen was shipped out of town a decade ago.

That’s an impressively long list of moves for a team that entered the off-season with surprisingly low amounts of cap space and draft capital. And it has left the Bears with what appears to be a pretty solid roster, at least on paper, though it’s fair to say that questions at quarterback certainly limit the optimism.

But things start to look much more questionable when you gaze beyond 2020. You see, the only way Pace could spend money this off-season was by borrowing from the future salary cap, and he did that quite heavily. Several players have had their contracts restructured within the last year+ to clear up immediate cap space by moving money to 2021 and beyond. This totaled around $20M from a combination of Khalil Mack ($7.8M), Kyle Fuller ($4.5M), Charles Leno ($4.2M), and Cody Whitehair ($3.2M).

On top of that, most contracts Pace handed out this off-season were absurdly back loaded.

  • Robert Quinn has a $6M 2020 cap hit on what is essentially a 3 year, $43M deal (a 2020 savings of over $8M from the average cap hit for the deal). The downside is he will still have total cap charges of $37M remaining in 2021 and beyond, and will likely only play in Chicago for 2021-2022. To make matters worse, those will be his age 31 and 32 seasons, when his play will likely start to slip. He’s a speed rusher that relies heavily on that one skill, so it’s possible that decline will be very pronounced.
  • Danny Trevathan has a $4.2M 2020 cap hit on what is essentially a three-year, $21.7M deal. That saves about $3M in 2020 cap, but means the Bears will still have $17.5M on cap charges for his remaining 2 seasons, in which he will be 31 and 32 and likely start to see his play decline.
  • Jimmy Graham has a $6M 2020 cap hit on what is essentially a one-year, $9M deal. That saves $3M in 2020, but means the Bears will have that cap hit in 2021 when he is likely not on the team (if he is on the team, he’ll have a $10M cap hit, which is not ideal for a player who will turn 35 during that season and has already started showing signs of decline).

Read More …

Tagged: , , , ,

59 Comments

What Should Teams Do at the Goal Line?

| June 9th, 2020


It has become common knowledge that passing is far more valuable than running in the NFL. But I have surprisingly seen very little data about how that changes as teams approach the end zone and the real estate tightens.

I found this excellent article looking at all goal-to-go plays, which found that passing is still more valuable than running and highlighted specific types of runs and passes that work better than others. But that groups plays from the 8 or 9 yard line together with plays from the 1 or 2, and those are drastically different scenarios.

I spent about 15 minutes on Google trying to find something detailing what’s most effective for teams to score a TD from the 1 or 2 yard line, and couldn’t find anything, so I decided to do it myself. I started by using the Pro Football Reference game play finder to get a basic look at how often, and how successfully, teams run vs. pass from the 1 and 2 yard line. The table below shows that information for the years 2016-19. I chose that specific time range to be consistent with available information from later in the study.


Read More …

Tagged: , , , ,

119 Comments

How Consistent are Explosive Players?

| June 2nd, 2020


Recently, I’ve found that explosive plays are really important to overall offensive production and explosive plays are extremely inconsistent from year to year on a team level. Today I want to look at explosive plays on an individual level to see if players can be fairly reliable counted on to be more or less explosive than expected.

The Set-Up

Like with the team-level data, I used performance from 2014-19 as my sample size. I used the Pro Football Reference Game Play Finder to identify all players who had at least 200 pass attempts, 50 pass targets, or 100 carries in each season. I chose these numbers as somewhat arbitrary thresholds to get a good mix of a sufficient data sample each year and a big enough sample size within each data point to make the data as reliable as possible.

I then looked up the explosive plays (runs of 15+ yards, passes of 20+ yards) each of those players achieved in those seasons. I used the data in aggregate to get average explosive play rates for each. Full data can be seen here.

  • Passing: on average, 8.7% of all passing plays (including sacks) resulted in explosive passes. This data did not seem to change much from 2014-19, with each year fluctuating between 8.3% and 9.1% and no clear year-to-year pattern. I also double checked that smaller sample sizes didn’t skew the data, but the rate stayed the same when I only looked at player seasons with 300+, 400+, or 500+ pass attempts.
  • Rushing: on average, 4.8% of all running back carries resulted in explosive runs. I’ll note I excluded QBs with 100+ carries in a season from this, because many of those are scrambles and thus have a much higher explosive rate, and the sample size of QBs with 100+ carries was too small to study independently. Again, this number didn’t change much year-to-year or if I had a larger carry threshold for inclusion (I checked 150+, 200+, and 250+ carries).
  • Receiving: I split this one up by position, since WRs, TEs, and RBs are used quite differently in the passing game. Overall, 5.5% of targets to running backs, 11.1% of targets to WRs, and 9.3% of targets to TEs resulted in explosive completions. Again, there was little variation year-to-year.

I then used those rates as a baseline for how many explosive plays an individual should be expected to produce based on their volume for the year. For example, a RB with 100 carries and 100 pass targets should be expected to have 4.8 explosive carries and 5.5 explosive receptions. If they actually produced 6 explosive carries and 4 explosive receptions, they had 1.2 more explosive runs and 1.5 fewer explosive catches than expected.

To save words, from here on out I’m going to refer to that as the explosive differential.

Read More …

Tagged: , ,

126 Comments

Nick Foles Will Be the Starting Quarterback

| June 1st, 2020


For the Bears, there is no more important issue looming than which man will be under center receive the shotgun snap when the Bears take the field against Detroit in Week One. Today I want to dig into the stats to see what we can learn about Foles vs. Trubisky, as well as what to expect from whoever wins that derby compared to other QBs around the NFL.

The table below shows basic efficiency statistics for Trubisky and Foles in the Reid offense (so Trubisky in 2018-19 in Chicago and Foles in 2016 in KC and 17-18 in Philadelphia), plus the other three notable recent Reid QBs (Smith 13-17, Mahomes 18-19, Wentz 16-19). I’ll note I included playoff stats for everybody because otherwise Foles’ sample size is just so small (less than 350 with just regular season, just over 500 with playoffs included). I also included the NFL average for 2018-19 as a frame of reference for what’s roughly normal around the league. I split up the data into short and long passes (targeted more than 15 yards past the line of scrimmage) using Pro Football Reference’s game play finder.

That’s a lot of information to digest, so let’s look at short and deep passes separately.


Read More …

Tagged: , , , , , , ,

174 Comments

How Consistent are Explosive Plays?

| May 26th, 2020

The Bears produced the fewest explosive plays in the NFL last year, and given the importance of explosive plays to overall offensive output, that largely explains their status as one of the worst offenses in the NFL.

So I want to look at how consistent explosive plays are. We’ll start with a team-by-team basis, and then look at it on a player-by-player level in a follow-up article.


The Setup

I used Pro Football Reference’s Game Play Finder to track explosive runs (gained 15+ yards) and passes (gained 20+ yards) for each team season since 2014. I did this to have 5 years to compare season-over-season consistency (2014 vs. 2015, 2015 vs. 2016, etc.), giving a respectable sample size of 160 data points without going too far into the past, since the NFL is a constantly evolving league.


Results

I started by doing a simple comparison of explosive plays a team had in one year compared to explosive plays they gained the following year. As you can see in the chart below, there wasn’t much of a relationship.

As a reminder, correlation (R²) is a measure of how strong the relationship between two variables is. It ranges from 0-1, with 0 meaning there is no relationship whatsoever. So a value of 0.027 tells us there is basically no relationship between how many explosive plays a team has in one year compared to how many they will have the following year.

I’ll note I did similar looks for explosive runs and passes when separated out from each other and got similar results (R² < 0.07 for both). I also looked at all three in terms of explosive rate (explosive plays/total plays), and got similar results. I don’t feel the need to pepper this article with a bunch of similar graphs that show no results, but if you’re curious, the full data set and graphs can be seen here.

This then, would seem to suggest good things for the Bears. Just because they were unexplosive in 2019 does not mean the same will be true in 2020. 

Read More …

Tagged: , , ,

195 Comments

Is Producing Explosive Plays More Important Than Avoiding Negative Ones?

| May 18th, 2020


I did some work last off-season examining how important explosive plays are to an offense’s production, and found that there is a strong relationship between the number of explosive plays (runs of 15+ yards, passes of 20+ yards) and overall offensive performance (measured in either points/game or DVOA rank). I have updated that information to now include 2018 and 2019 data and still found a strong relationship, as you can see in the graphs below.

Correlation (R²) can be loosely interpreted as how much of the pattern is explained by that variable, which means explosive plays account for roughly 40-60% of overall offensive production, which is quite a high number, and consistent with values from the 2018 season alone. Seeing the same relationship across multiple seasons of data provides additional credibility to the relationship.

(Side note: just like in 2018, total explosive plays shows a stronger relationship with both points/game and DVOA than the % of offensive plays that are explosive, so I’ll probably just track total explosive plays from now on.)

Read More …

Tagged: , ,

268 Comments

Establishing Realistic Expectations for Cole Kmet

| May 13th, 2020


The Bears spent their first pick (43rd overall) on Cole Kmet, a big tight end from Notre Dame who has a chance to plug a Bears’ roster hole from day one.

It should be noted, however, that tight end is a position where conventional wisdom says it’s hard to make a big impact in your rookie season due to a steep learning curve. In order to establish realistic expectations for Kmet, let’s take a look at how comparable tight ends have fared in their first few years of the NFL.

In order to do so, I looked at all 18 tight ends drafted in the 2nd round between 2010-19. I tracked their playing time and statistical contributions on offense after extrapolating to a full 16 game season to normalize the data since several players missed games with injuries.

The full data can be seen here, but I’m just going to show the range of snaps played, targets earned, passes caught, and receiving yards, which can be seen in the table below.



Read More …

Tagged: , , , ,

209 Comments

Why Do Day Three Draft Picks Hit (or Miss)? A Deep Dive…

| May 6th, 2020


Former NFL executive Joe Banner did an interview a few years ago where he referenced a study by an NFL team that found most day 3 picks who turn into successful NFL players are guys who slip through the cracks either because they were from small schools, had an injury in their last year of college, or were undersized for their position.

This made me curious, and since it was a private study without information published, I decided to do it myself.


The Setup

I used the Pro Football Reference database to grab information about every day 3 draft pick from 2007-16. I stopped at 2016 because I wanted players who had finished their 4 year rookie contracts, and started at 2007 to give me 10 seasons’ worth of data. This gave a sample size of 1509 picks.

I then identified players who were a hit based on 2 criteria:

  1. They were a primary starter on offense or defense for at least 2 seasons (as defined by Pro Football Reference).
  2. They had a career AV (a Pro Football Reference metric that attempts to quantify overall impact of each player) of at least 15. I chose this value as the cutoff because Nick Kwiatkoski finished his four years in Chicago with an AV of 15, and that feels about right for the cutoff for a hit.

Any player that hit at least one of these thresholds was considered a hit, while all others were not. I also found that the majority of players who hit this threshold also hit the 1st one, though there of course some outliers.


Results

Let’s take a look at some different factors and see how they influenced hit rates on day 3 of the draft.

Small School

We’ll start with players from a small school, which I defined as anything but the “power 5” conferences (ACC, Big Ten, Big 12, Pac 12, SEC). The table below shows hit rates for Power 5 picks vs. small school picks for each round of the draft’s 3rd day.

A few thoughts:

  • Generally, the small schools hit at a slightly higher rate than Power 5 schools, a difference that is more pronounced later in the draft.
  • Given this, it seems weird that teams spend far more day 3 draft picks on Power 5 players than they do small school guys. Between 63%-74% of picks in each round were spent on players from Power 5 schools. Of course, fewer small school players getting drafted probably helps explain why they have a higher hit rate. If you take more small school guys just because, they probably won’t be able to sustain that higher odds of success.
  • Notice rounds 4-5 are fairly similar, but very different than rounds 6-7. In order to have larger sample sizes, I will split day 3 into those 2 groups going forward.
  • With that in mind, general rules of thumb to keep in mind are that roughly 1/3 of picks in rounds 4-5 pan out, compared to roughly 15% (1 in about 6.5) of picks in rounds 6-7. As we look at other factors, we’ll look for anything that changes appreciably from those numbers.

Read More …

Tagged: , ,