clock menu more-arrow no yes mobile

Filed under:

Advanced Stats Week 2022: Sticky Metrics and Advanced Stats Goals

What stats are stickiest from season to season, and how can we measure improvement for GT in 2022?

NCAA Football: Georgia Tech at Miami Jasen Vinlove-USA TODAY Sports

A year ago, we wrote this:

From the coaches to the players to the message boards to the stands, there is not one person associated with Georgia Tech football who would be content with the 2021 team faring no better than GT has the past two years. Another 3 win campaign, with below average statistical performances across the board, would give major cause for concern about the trajectory of the program.

Well, that’s happened. That’s where we are. But, I want to use this space today to look at a few places where last year’s performance might not guarantee a repeat this season. Are there places where improvement is more likely than others? What does that mean for where we might most realistically expect improvement for GT in 2022?

Sticky Metrics

To try and glean some insight here, I analyzed three years of play by play data using 12 different measures of performance. These included success rate and PPA on offense and defense, for all plays, rushing plays, and passing plays. The goal was to see which metrics are most consistent from year to year and which metrics were the least sticky.

Metric Stickiness, Year to Year

Side of the Ball Play Type Metric R-Squared GT 2021 Ranking
Side of the Ball Play Type Metric R-Squared GT 2021 Ranking
Offense All SR 0.256 83
Defense All PPA 0.199 112
Offense Rush SR 0.194 93
Offense Pass SR 0.191 81
Defense All SR 0.189 116
Offense All PPA 0.185 91
Defense Rush SR 0.165 102
Offense Pass PPA 0.146 95
Defense Rush PPA 0.135 77
Offense Rush PPA 0.134 69
Defense Pass PPA 0.111 129
Defense Pass SR 0.077 112

Causes for Optimism

First and foremost, the bottom of the table should give slight relief to the Georgia Tech faithful. Pass defense, particularly success rate and PPA (an Expected Points style metric from that we discuss from time to time), are the least sticky metrics from season to season. That is, of the twelve metrics we looked at, 2021 performance would be least predictive of 2022 performance in those two areas where Georgia Tech was especially weak last year.

Secondly, the top of the table, offensive success rate, is a metric where Georgia Tech was comparatively better last season. This indicates a modicum of offensive efficiency that GT should be able to count on and hopefully build on under Chip Long’s tutelage this season.

Causes for Concern

Although defensive pass PPA is not sticky, overall defensive PPA is. As the second stickiest metric from year to year, Georgia Tech’s defense came in at #112 last year and thus wouldn’t have much expectation for overall improvement there. That tracks quite well with the most recent SP+ projections from Bill Connelly, which has Georgia Tech 111th in the country in overall defense.

Secondly, Georgia Tech’s best categorical finish last season came in offensive rushing PPA. That metric is the third least consistent from year to year, and last year’s relative success by GT in that area was largely driven by explosive runs from Jeff Sims and Jahmyr Gibbs. One of those players now resides in Tuscaloosa, but the other remains in position to drive explosive offense with his explosive presence on the ground. However, if rushing explosiveness does fall off, that’s going to further reduce the ways this offense can get the ball in the end zone, especially given the perpetual offensive line weaknesses.

Most concerning is the possibility that the defense would persist in its flammability. Most hopeful is that the pass defense can recover and the offensive efficiency has a solid floor underneath it. I’ve made my opinion very clear that it won’t be enough, but if improvement is coming, what will it look like? Where should we be looking for improvement, and how do we quantify some goals for 2022?

Advanced Stats Goals

Pass Defense

We’ve repeatedly described the absurdly bad performance of Georgia Tech’s pass defense in 2021. There’s not a better metric at our disposal for evaluating this performance than Expected Points Added per pass play. Every time an opponent dropped back to pass, how did our pass defense affect the overall state of the game? Looking at EPA/play for pass defense for the Power 5 and FBS Independents reminds us just how badly things went.

In 2020, the Jackets gave up 0.08 EPA/pass play, which landed in about the 40th percentile nationally. Last year, that fell to almost 0.5 EPA/pass play, which landed in the 0th percentile. Looking for improvement there is a tautology. It can’t get worse. So how much improvement should we be targeting? How much improvement might come from coaching and personnel overhauls throughout the defensive front and secondary?

Average feels beyond reasonable expectation. An improvement of about two standard deviations would get GT back to where things were in 2020, and that seems about as much as we could hope for. Allowing 0.08 EPA/pass play and ranking about 80th in the FBS in EPA/pass allowed would be massive improvement for this team.

As usual, there are a couple of other areas on defense that will help us achieve that big picture goal in pass defense. Both defensive havoc and pressure rates have a strong correlation with improved pass defense. The GT defense, already struggling in this area, managed to take another step back in havoc rate last season. The national average is about 21% of defensive plays producing havoc (tackle for loss, forced fumble, interception, or pass breakup), but GT managed only 15% in 2020 and then a paltry 10% in 2021.

Similarly, the pressure rate on pass plays was a relatively anemic 22%, below the national average of 27% and the 2020 performance of 23%. The emergence and health of Keion White, the improved depth at linebacker, and the fresh faces in the secondary give me reason to think that both the pressure and havoc numbers can jump up. But these have been perpetual weaknesses. Will the new members of the defensive staff finally be able to unlock something to make Georgia Tech a more disruptive defense?

Defensive Goals

  • EPA/pass play allowed: <= 0.08 EPA/pass (0.45 in 2021, 0.06 is average)
  • Havoc Rate: >=18% (10% last year, 21% is average)
  • Pressure Rate: >=27% (22% last year, 27% is average)

Pass Offense

Consistent rushing success is helpful, but unless we are returning to the days of the spread option, this offense’s overall effectiveness will hang mostly on the passing game. We could look at EPA/pass here as well, but instead we will focus more directly on the quarterback’s contribution, as measured by Completion Percentage over Expectation. We’ve discussed this before, but a reminder won’t hurt.

The concept is simple: given where and how far down the field I’m throwing the ball, am I completing more or less passes than expected? It’s quickly become one of the gold standards for evaluating QB play. Josh Hermsmeyer detailed a few years ago how valuable CPOE is as a measuring stick for QB prospects entering the NFL draft. Elsewhere, Jason DeLoach, also known as @CFBNumbers, has laid out the case for why CPOE is one of the premier ways to evaluate quarterback play that we have at our disposal.

Georgia Tech is in an interesting situation that is framed well by considering this metric. Jeff Sims has been below average in CPOE each of the last two years, while recently arrived transfer Zach Gibson is a CPOE darling. All of the public comments from the program indicate that Sims is entrenched as the starting QB, but Ben unpacked a few weeks ago why that might not be the case for the whole season. Sims hasn’t been good enough to have the spot guaranteed for the whole year. Below, you can see where Sims ranked nationally in CPOE (this chart does not include the Yates snaps for GT). If this included all of the FBS, Zach Gibson would be up there in the Purdue neighborhood at the top-right.

From my vantage point, the situation is simple. Jeff Sims has to be better in metrics like CPOE, or Zach Gibson should get a shot as the starting quarterback. Of course, my career doesn’t hinge on the outcome of decisions like that.

Especially given his explosive rushing abilities, Sims is given the benefit of the doubt to succeed at quarterback this year. But if he hasn’t shown a step forward in the first third of the season, it’s perfectly reasonable to expect Gibson to get a shot.

What is a step forward? Let’s set the benchmark at the same place we did last year, +2%, for Georgia Tech’s team CPOE. After two straight years around -2 or -3%, that would be a meaningful step forward.

There are a couple of other underlying factors that will help the quarterback achieve his CPOE goals as well.

As the mirror for the defensive goals, the offense must limit the amount of pressure affecting the quarterback. The offensive line actually performed slightly better in this metric in 2021 than in 2020, but it still allowed the QB to be pressured at above average rates. Still, more pressure lowers passing effectiveness. The once-again reworked offensive line and the new play-caller have to find ways to keep a clean pocket and to get the ball out on time.

Secondly, let’s talk about play-calling patterns. And here I have a mea culpa to offer. When we have talked about situational play calling over the last few years, I have called for a marked increase in Georgia Tech’s pass percentage on first downs. First down passing rate has been repeatedly shown to improve efficiency in the NFL, but work I did this offseason has shown that this does not carry over into college football.

However, teams can improve their efficiency by not calling running plays on second and long. The national average is just over 40% of runs on second and long plays, while Georgia Tech ran on just about 50% of those plays. That didn’t help the offense.

On second down and long plays, passing is better. Further, third down plays on average become much more manageable after passes on second and longs. Coach Long can help Jeff Sims (or Zach Gibson) by reducing the number of called runs on second and long plays.

Let’s look at one more way that offensive design might help the offense this year. Last season, Georgia Tech averaged just over 8 air yards per pass attempt, which was about a yard below the national average. Increasing average depth of target has a moderately strong correlation with more efficient passing offense. You can see GT as slightly below average in average depth of target and Passing EPA/play last season.

Last season, Jeff Sims’s two highest average target depth games were UNC and Duke. Those were two outstanding offensive performances. Coach Long can help the quarterbacks by scheming up route combinations that work to get guys open further down the field. That creates better passing offense.

Offensive Goals

  • CPOE: >= 2% (-3% last year, +2% is average)
  • Pressure Rate Allowed: <=26% (28% last year, 27% is average)
  • 2nd Down Run Rate: <=40% (48% last year, 40% is average)
  • Average Target Depth: >=9 yards (8.3 last year, 9.2 is average)

The deck is stacked against Georgia Tech in 2022, no doubt about that. But even with the challenges ahead and the daunting schedule to come, we’re always looking for improvement that we can measure. As our stickiness measures showed, pass defense may well see a solid bounce this year, and we’ve laid out some clear ways to track that potential. On offense, the passing must improve, and that involves the line, the QB, the receivers, and the play-caller. We’ve laid out the tracks that the team can run on this year, and we’re just about two months away from watching things play out.