clock menu more-arrow no yes

Filed under:

Georgia Tech Football: Advanced Stats Review - GT vs. Duke

New, 15 comments

There was plenty to critique, but Georgia Tech made explosive plays on both sides of the ball to win a game it had to have

NCAA Football: Georgia Tech at Duke William Howard-USA TODAY Sports

Final Score: Georgia Tech 31-27

Model Prediction: Duke by 1, Duke to cover: correct (closed GT -4.5)

Projected EPA (Offense and Defense) Margin of Victory: GT by 2

GT Win Probability (Based on Success Rate, Yards Per Play, and EPA): 58%

On Saturday afternoon, Georgia Tech went into empty Wallace-Wade Stadium to play a team it was projected to beat by 4 points. It won by 4 points, but it felt so much more difficult than that. Why? Two reasons immediately come to mind.

As Ben covered on Friday, the Georgia Tech fanbase still largely sees Duke as a team it should beat 90% of the time, with most of those wins of the comfortable variety. Of course, David Cutcliffe has turned Duke into a far different program than the one GT fans knew under Ted Roof and Carl Franks. Duke used to occupy the dregs of the FBS, but this year, like others in recent history, they are probably a slightly below average team on balance. On the flip side, they are certainly not “one of the best offenses in the country,” as Coach Collins curiously decided to posit at halftime. GT fans see their program as occupying an altogether different space than where Duke resides, but unfortunately, that simply has not been true for much of the past decade.

The second issue is that GT’s talent flashes in a way that looks overwhelming versus a team like Duke. On the third play of the game, Jahmyr Gibbs looked like he was playing for Dalton again, and it seemed like he could do whatever he wanted all day. He ended up averaging less than three yards per carry. The throws that Jeff Sims made to Adonicas Sanders on the final drive and to Malachi Carter earlier in the game looked NFL-caliber. His two interceptions defied all expectation of the kind of decisions a P5 quarterback should make.

The frustrated expectations and the confusing variance in performance can make games like this one feel so confusing.Let’s where the numbers take us to get a better understanding.

Success Rate Comparisons

GT vs. Duke Success Rates

Success Rate O GT Offense D Opp Offense National Avg
Success Rate O GT Offense D Opp Offense National Avg
Down 1 40.70% 1 51.30%
2 38.10% 2 37.90%
3 40.00% 3 42.90%
4 4 66.70%
Qtr 1 43.80% 1 42.10%
2 31.80% 2 52.40%
3 50.00% 3 33.30%
4 36.40% 4 51.60%
Pass P 40.00% P 58.10% 41%
Rush R 39.40% R 38.30% 42%
Overall 39.70% 45.70% 42%
Success rate is the baseline metric for efficiency. As a reminder, a successful play gains 50% of the needed yards on 1st down, 70% on second down, and 100% on third or fourth down.

The 6% success rate margin that Duke enjoyed was perhaps the most surprising statical outcome of the day. Last year, GT put up a +3% SR margin against Duke, and I thought that the matchup looked to favor GT even more in that department this year. The Georgia Tech offense putting together a spectacular drive with the game on the line, but the middle of the game inefficiency and the back to back three outs that let Duke take the lead were significant cause for concern.

On the Duke side of the ball, Duke ran Durant as we expected, and the defense did an admirable job holding his success below 40% and preventing any explosives. On the flip side, Duke had been much more reliant on explosives in the passing game than sustained efficiency, but they put up an outstanding 58% success rate on pass plays. More on that below.

Advanced Stats Comparison and Positional Breakdowns

GT vs. Duke Advanced Box Score

Adv Box Score GT Offense Opp Offense National Avg
Adv Box Score GT Offense Opp Offense National Avg
Snap Count 67 97 71.5
# Pass Plays Called 31 31 31
Avg Starting FP 76.86 66.07 70.5
YPP 6.04 4.77 5.7
YPA (incl. sacks, scrambles) 10.61 9.32 7.39
% of Passes on 1st Down 41% 29% 40.17%
% of runs on 2nd and long 50% 67% 39.80%
Avg EPA/play 0.09 0.04 -0.01
Avg EPA/pass 0.33 0.47 0
Avg EPA/rush -0.14 -0.18 -0.01
Total EPA 5.69 3.46 -0.96
Avg Air Yards / Completion 11.36 8.55 6.14
Air Yards / Attempt 14.83 9.38 8.89
CP 45.83% 75.86% 62.54%
CPOE -7.94% 15.80% 1.90%
Line Yards per Carry 2.74 2.66 2.55
Opportunity Rate 35.48% 44.83% 42.42%
Power Success Rate 0.00% 57.14% 68.60%
Stuff Rate (Offense) 18.18% 23.33% 19.17%
Havoc Rate 13% 9% 21.00%
Pressure Rate 10% 29% 27.00%

Quarterback

GT vs. Duke QB Play

Player Success Rate EPA/play EPA/pass CP CPOE CPOE10+ CPOE20+ EPAwPressure EPAnoPressure
Player Success Rate EPA/play EPA/pass CP CPOE CPOE10+ CPOE20+ EPAwPressure EPAnoPressure
Sims 0.4 0.1 0.33 0.46 -0.08 0.11 0.09 -0.86 0.82

We got the full Jeff Sims experience on Saturday. The two interceptions demonstrated a lack of situational awareness and an inability to calmly respond to pressure. The two throws to Adonicas Sanders on the game’s penultimate drive showed poise and touch that would stack up with the best of the best in the country. The two strikes to Malachi Carter - one in the 1st quarter and one in the 3rd - showed arm strength and accuracy in enviable combination. The second quarter saw Sims unable to find a rhythm or deliver a ball where it needed to be.

Once again, Sims’s performance without pressure was outstanding, and he continued to show real struggles against pressure. The biggest takeaway for me though: in a game where his CPOE was -8%, it was +11% on 10+ yard throws and +9% on 20+ yard throws. The last drive showed that the capabilities of this offense in addition to the way that opposing defenses continue to play it should lead to more deep throws in the weeks to come. Let’s see if Coach Patenaude will latch onto that in the bye week.

Rushing

GT vs. Duke Rushing

Runner Carries Rushing Success Rate Highlight Yards
Runner Carries Rushing Success Rate Highlight Yards
Gibbs 13 0.31 17
Smith 8 0.5 9.5
Sims 6 0.5 7
Mason 5 0.4 0

Once again Sims and Dontae Smith post an excellent fifty percent success on called run plays. Once again, Jahmyr Gibbs and JP Mason are doomed to failure on runs that require the offensive line to maintain blocks for too long or require athleticism to successfully pick up a defender while pulling that the current version of this line does not possess. Once again, there were too many called runs on first down and on second and long plays. The issue with the running game has nothing to do with the backs.

Though he still lacks the snap count to fully qualify, Dontae Smith is now the highest graded running back in the country according to PFF. Gibbs and Mason are both solidly graded in the 70s, which is well above average on PFF’s scale. The the issue here, clearly, is the line and the play design. Run outside zone. Run power. Simplify the running game, and let the outstanding talent in this group do what it can.

Receiving

GT vs. Duke Receiving

Receivers Receiving Success Rate Avg Target Air Yards Targets % of Team Air Yards YACatch
Receivers Receiving Success Rate Avg Target Air Yards Targets % of Team Air Yards YACatch
McGowan 0.5 20.5 4 13.06% -4
Smith 0 2 1 0.32% 0
Harris 0 2.5 2 0.80% 0
Carter 0.33 17 6 16.24% 39
Sanders 1 36.5 2 11.62% 2
Ward 0 23 1 3.66% 0
Gibbs 0.33 -0.67 3 -0.32% 84
Norris 0.2 13.8 5 10.99% 15

The Georgia Tech passing game was all about explosion on Saturday. McGowan, Carter, Sanders, and Norris all had average depth of target over 13 yards - about 60% higher than the national average. Sanders, of course, had the two catches that swung the game on the final drive; the first was a good route and a great throw; the second was a great throw and a dynamite catch. Carter and Norris both added significant yards after the catch, even on some of these deeper targets, As usual, Gibbs got shorter targets but was dynamic with the ball in space.

Sims missed a couple of open guys - Ward and Norris immediately come to mind - or else these numbers would have been far more impressive. The receiving group is coming into form. The question that remains for me is the insistence on deploying tight ends; their blocking is largely mediocre, and they have been such a small factor in the passing game whenever Sims plays. Sending out double slots with a combination of McGowan, Rutherford, and McCollum or using more snaps with the Smith-Gibbs combination both seem like better personnel usage for this offense.

Offensive Line

GT vs. Duke OL FLOPs

Player # of FLOPS
Player # of FLOPS
Williams 2
Minihan 1
Johnson 2
Cochran 1
Lay 1

I expected this to be one of the few games on the schedule where the offensive line could largely control things. The continuing rash of injuries, the difficult schemes these guys are being asked to block, and the continued execution failures made that a fantasy on Saturday.

The pressure rate on Jeff Sims ended up at 29%, which is about a standard deviation worse than average. Some weeks, it’s easy to spot plays where Sims brings himself into pressure, but this week I would pin 90+% of those pressures on the offensive line and its failures. The run stuff rate came in just about average at 18% - well down from last week - but this was a matchup to turn in a great performance, not an average one.

Cochran surprisingly struggles in run blocking. Johnson reverted to early season form. Williams can’t figure things out. Vaipulu looks lost; on one play, he pulled and watched three different defenders go by him without touching one. It’s hard to see where the improvement will come from in this group over the duration of the year. Coach Patenaude is going to have to scheme around the weakness, not expect magical improvement.

Defensive Disruption

GT vs. Duke Disruption

Player Defending Havoc Plays # of Pressures # of Run Stuffs
Player Defending Havoc Plays # of Pressures # of Run Stuffs
Brooks 0 0 1
J. Thomas 1 0 0
Carpenter 1 0 1
Swilling 1 0 0
Jackson 2 0 3
Eley 2 2 1
Ivey 1 0 1
Sims 1 0 0
C. Thomas 1 0 2
Domineck 2 1 2
Lockhart 1 0 1

Unfortunately, we got more evidence that the UNC disruption fest was an aberration. The pressure rate of only 10% on Duke drop backs is the third-worst of the past two seasons (ahead of only the Clemson and BC games last year), and the havoc rate was still well below average. This came against an offense that gives up an average amount of havoc plays and pressures against, so it’s discouraging to once again produce so little disruption. The one positive note was a solid run stuff rate of 22%; Quez, Charlie, and Domineck led the way on that front.

Michael Lockhart flashed in his still limited snaps; he must move higher up in the DT rotation. Jared Ivey didn’t control the game, but he’s a good player. Once again, there’s almost no disruption happening from guys on the backend.

Pass Coverage

GT vs. Duke Coverage

Player in Primary Coverage Targets CPOE Allowed YAC Allowed
Player in Primary Coverage Targets CPOE Allowed YAC Allowed
J. Thomas 2 -7.50% 8
Carpenter 2 -18.67% 2
Swilling 7 32.44% 41
Jackson 1 35.00% 0
Walton 2 -0.13% 3
Oliver 2 31.50% 14
Walker 2 36.33% 1
King 0 0
Sims 5 21.04% 13
Brooks 1 27.67% 0
C. Thomas 1 -61.00% 0

And now we get to the worst of it, once again. Gunnar Holmberg - who is no Kenny Pickett - went for +16% CPOE against this Georgia Tech pass defense. He put up 0.47 EPA/pass, which was almost four times his season average coming into the game. This week, the corners bear much more of the responsibility than the safeties. That starts with Tre Swilling.

The number of times a defensive back is targeted usually tells you a lot about how the opposing offense views that defender. Swilling was targeted 7 times. For +32% CPOE+. And an additional 41 yards after the catch. The transformation this team is undergoing requires making hard personnel decisions. The “old guys” in the secondary have lost the chance to prove themselves. Give me more Miles Brooks, more Jaylon King (who was never even targeted in 17 snaps!), and more Zamari Walton.

EPA Highlights

EPA calculates the expected number of points added (or lost in the case of a negative number) on a particular play based on the down and the location on they field.

The EPA totals for this game leave us with a 2 point projected win for GT. This game was filled with swing plays going both ways, but the last three went for GT. An 8 point EPA deficit turned into a 2 point EPA advantage in the span of about one minute of game time.

As always, we’ll take a look at the most helpful and hurtful plays for GT.

Most Helpful Plays

  1. The Jahmyr Gibbs play - taking a 3rd and 5 wheel route 77 yards to the house on the opening possession. 7.50 EPA.
  2. Jeff Sims’s 36 yard touchdown pass to Adonicas Sanders. Self-explanatory. 4.40 EPA.
  3. Jeff Sims’s 39 yard completion to Malachi Carter on 3rd and 6, setting up 1st and goal at the Duke 2. 4.20 EPA.
  4. Juanyeh Thomas’s interception to end Duke’s final drive. 4.19 EPA.
  5. Jeff Sims’s 40 yard completion to Malachi Carter on 3rd and 10, reaching the Duke 18 in the middle of the 3rd quarter. 3.38 EPA.

Four passing explosives and the game-sealing defensive play reveal a team with significant capability. These passing explosives especially need to continue shaping the pattern for offensive scheme for the rest of the season.

Most Hurtful Plays

  1. Jeff Sims’s second interception on a baffling throw to Harris on 3rd and 18. -5.65 EPA.
  2. Jeff Sims’s first interception on 3rd and 5 from the GT 30. -3.85 EPA.
  3. Duke’s 7 yard touchdown pass on 4th and 2 to tie the game at 24. -3.76 EPA.
  4. Holmberg’s 37 yard touchdown pass to Bobo to get Duke on the board. -3.27 EPA.

Two interceptions, a botched 4th down play, and a busted coverage on a deep ball are all very revealing about the ongoing weaknesses that this team experiences.

Tracking Season Goals

*I set these goals for the 2021 season in some of my offseason preview work. We will be tracking them as we go this year.

GT vs. Duke Season Goals

Metric Season Goal This Week Season Long
Metric Season Goal This Week Season Long
GT CPOE >= 2% -8% 2%
Pressure Rate Allowed <=26% 29% 27%
Pass Rate on 1st Down >=50% 41% 47%
Defensive Passing EPA/play <= -0.06 0.47 0.24
Defensive Havoc Rate >=21% 13% 12%
Defensive Pressure Rate >= 27% 10% 25%

These numbers all went the wrong way on Saturday. Sims missed too many throws, and Patenaude didn’t give him enough easier chances to throw on first down. The line couldn’t protect, and the defense couldn’t disrupt. The secondary got torched. Fortunately, a few big plays allowed GT to come out with the win anyway, but there’s plenty of bye week work to be done.

Takeaways

  1. Jeff Sims didn’t play at the level he reached against either UNC or Pitt, but he hit explosives, including two when it mattered most. The first play to Adonicas Sanders on the last drive showed the formula - a nice straight drop on a 1st and 10 with a quick release, facing no pressure, to hit the receiver on the go ball. Georgia Tech can find more chunk plays by throwing like this on first down. The rest of the explosive pass plays came in much more difficult third down situations; yes, Sims can sometimes hit those, but it would sure help the offense to give him more of those chances on early downs.
  2. The offensive line can’t execute what it is being asked to execute. Coach Patenaude loves the inside zone, the split zone, the counter - all good football plays - but this line struggles to make those blocks. They struggle to pick up blitzes on obvious passing downs. So, let’s run more sprint outs, more powers, more outside zone. Give the guys up front a chance to success based on what they have shown in the first half of the season.
  3. The secondary continues to be the most under-performing unit on the team. Last year, GT gave up 0.22 EPA/opponent dropback, which came in at about the 40th percentile nationally. I saw potential for significant improvement in this area, setting a goal of -0.06 EPA/opponent drop back (which would put GT about the 75th percentile of passing defense). Instead, it’s gotten worse. The experienced guys in the secondary have all been here for the entire Collins tenure; if they were going to figure it out, they would have by know. It’s time to pass more of the baton to the young talent in the secondary - with King, Brooks, and Huff at the top of my list.

Heading into the season’s lone bye week, Georgia Tech sits at 3-3 and has a puncher’s chance of reaching bowl eligibility. This is an average FBS football team, which will face 5 relatively toss-up matches and one death-star in the second half of the season. 5-7 is most likely, but the passing offense and rushing defense this team has shown give reason to believe that at least 3 wins are possible in the homestretch. The causes for concern rest on the offensive line and in the secondary. The coaches have two weeks to assess their personnel and figure out how to scheme them into more advantageous situations.

Here at FTRS, we’ll be watching and writing, and we hope you’ll join us for the rest of the 2021 ride.