2020 Advanced Stats Season Review Part 3: Defense

In the first two parts of our look back at the advanced stats from 2020, we’ve seen the big jump that GT took from the rock bottom of 2019 performance and then identified some significant areas in need of focus for the offense to take another step forward in 2021. This week, we’ll dive deeper into the defense, which failed to show the improvement that many desired and predicted this past season.

Looking at some of the top line defensive numbers, we see why so many GT fans were disappointed in the defensive performance this year. Success rate allowed went up from 43% to 45%, yards per play allowed dipped slightly from 5.8 to 5.5, and EPA/play allowed rose from 0.08 to 0.11. The offenses faced in 2020 were roughly equivalent to those of 2019, so the top line numbers tell a story of stagnation or slight regression, which was not the expectation with the amount of experience that GT returned last year on that side of the ball. Let’s take a deeper dive to see just what happened on defense this past year.

To Disrupt, or Not to Disrupt

One of my favorite numbers to track is defensive havoc rate. For a refresher, a havoc play includes any tackle for loss, forced fumble, pass broken up, or interception. These havoc plays disrupt opposing offenses, either taking the ball outright or setting them behind schedule for moving the chains. One of the preseason goals I set for GT’s defense this year was to improve our havoc rate from the 19% level of 2019 up to the national average of 21%. Unfortunately, that went the other way and fell to 15%. The scatter plot below shows just how strong of an impact that havoc rate has on winning football games for GT.

The trend line paints a clear picture, and the data shows that GT’s defensive havoc rate explained 68% of the total variance in TEPA outcome this past season. That’s an enormous number, higher than any other defensive metric that I compared to total EPA outcome.

Let’s take a look at another metric as we try to pinpoint the shortcomings on the GT Defense. Pressure rate tracks the percentage of pass drop backs on which the defense pressures the quarterback. This data is not publicly available, so a friend and I started charting it in games that we were tracking this year. On the year, GT pressured the quarterback on 23% of drop backs, which falls below the national average of 27%. This lack of pressure bolsters the effectiveness of the opponent’s passing attack.

Though the correlation is not nearly as strong as what we saw with havoc rate , there is a clear effect here; getting more pressure lowers the efficiency of opposing passing attacks. The lack of disruption that GT was able to create allowed opposing offenses to throw the ball effectively.

I do want to note one bright spot here, and that is Jordan Domineck. Before the season, we pegged him to make an impact, and he came through. According to data shared by David Hale, Domineck had an individual 14.7% pressure rate. Which ranked 4th in the conference for all defensive linemen. He added 4 sacks, created 7 incompletions, and had 10 tackles for loss or no gain, which tied for 7th among DL in the conference. We need more Jordan Domineck on the pass rush next year! This ties into a concept that we looked at last week, the “or” vs. “and” positional distinction of Eric Eager. Defensive End is one of those “or” positions, which means one elite player can vastly change performance. The hope here is that Domineck can be the guy, and flashes from Jared Ivey and Kyle Kennard as well as the incoming production expected from Keion White provide hope for a significant jump in this kind of production next season.

As they always do, injuries played a factor, especially limiting GT once again on the defensive line. There were 2 games in which GT started four defensive lineman entirely different from those projected to start before the season and another 3 games in which only one projected starter was available. That is a significant amount of games lost to injury for a crucial unit. The bright side, of course, is the number of reps that went to young defensive linemen, hopefully paving the way for improvement this fall.

Let’s take a look at one more important number that had an outsized impact on the overall performance this past season.

Run stuff rate measure the percentage of run plays where the defense holds the offense to zero or negative yards on called rushing plays. The national average is 19%, and the GT defense came in just below that at 18%. The above plot shows the game to game impact of this metric. Run stuff rate had an R² of 0.59, meaning it explained 59% of the variance in Total EPA outcome. Once again, the lack of disruption that GT was able to create had a huge impact on (not) winning. Linebackers have a significant role to play in this, of course. Once again using numbers from David Hale, there are a couple of interesting insights from the linebacking corps. David Curry had 11 tackles for loss or no gain, which tied for third among ACC linebackers (good!), but 15 missed tackles, which was 3rd most among LBs (not good!). Quez Jackson had only 3 run stuffs all year, and his average tackle on running plays was 4.6 yards downfield.

Takeaway #1: Lack of disruption continues to be the achilles heel of the GT defense. Havoc rate, run stuff rate, and pressure rate were all at below average levels, and the GT defense paid a price for that, giving up much more efficient offense when those rates went down. An absolute must for 2021 will be the development of a consistent disruptor. My money is on Jordan Domineck or Keion White.

Takeaway #2: The defensive play calling needs to do a better job of matching aggressive calls with the right personnel. To grow our disruption numbers, more blitzing is needed, but it has to come from the right place. My hope is to see blitzing coming from players like Khatavian Franks, Trenilyas Tatum, and Charlie Thomas, with the aim of getting athletic guys who can avoid blockers into the backfield.

Passing Defense Woes

As was true in our offensive analysis last week, the passing game had an outsized impact on GT’s overall performance compared to stopping the run.

As on offense, focus for play to play improvement must be in the passing game. Passing success rate allowed explained 39% of the variance in total EPA outcome, while rushing success rate explained exactly 0%. As you’ll recall from last week, the numbers for all of FBS showed a similar importance in the passing game (R² of 0.43); nationally, the rushing success rate had an R² of 0.29, which is much higher than it was for GT but still significantly lower than the passing success rate’s affect on total EPA outcome. Pass defense has to be at the forefront of offseason preparation and in-game execution.

Once again using Eric Eager’s framework, the defensive backfield is an “And” unit. Failure by one player allows touchdown. Over the long haul, one position of weakness will be exploited. Georgia Tech’s two starting corners both ranked around the 60th percentile in PFF grades for the season; unfortunately, the two starting safeties were around the 30th percentile, and nickel play was graded right around national average. This left far too many opportunities for opposing offenses to find holes in the defensive backfield. Coach Burton has his work cut out for him to shore up those holes in the coming year.

One interesting finding is that GT was particularly exploited by teams who relied heavily on the quicker, shorter passing game.

This chart shows that as the average depth of target for our opponents increased, their passing efficiency went down. Teams that averaged throws at around 7-8 yards of depth were lethal against this GT defense. That highlights the weakness of linebacker and safety play in the passing game. According to David Hale, our starting middle linebacker allowed 8.6 yards per pass attempt when he was targeted. Improving pass coverage in the middle of the field is of paramount importance in the upcoming year.

Takeaway #3: If the starting safeties are performing more like 2020 than 2019, it’s time to try somebody new. Derrik Allen showed improvement as his reps increased over the course of the year, and Kaleb Edwards may be too athletic to keep off the field. A slow start against the pass in 2021 should lead to trying the young guys at safety.

Takeaway #4: Deploy linebackers and nickel backs who can play man coverage within 10 yards of the line of scrimmage. Quez Jackson returns after allowing only 6.2 yards per pass attempt when targeted. Coach Thacker must identify who else can play alongside him without getting exploited in the intermediate passing game.

Conclusion

For the Georgia Tech defense in 2021, disruption up front and consistency in pass coverage will tell the story of the season. We’ve offered 4 ways forward: identify one reliable disruptor on the defensive line, blitz aggressively with the right people, try younger guys if the safety play doesn’t improve, and identify the second starting linebacker who can hold up against the pass. Personnel and coaching improvements will both be necessary. Just like on offense, it comes down to recruitment, development, and deployment. It’s time for a defensive-oriented coaching staff to prove their bonafides.