Georgia Tech Football: Duke Advanced Stats Review – Offense

Georgia Tech put together an impressive performance this past weekend to knock off a Duke team that looked better than it on paper, despite said Duke team being in year one under new head coach Mike Elko. I’m not quite ready to say that any performances or tactics were particularly stellar, but if you’re a fan of signs of progress, there are definitely a bunch of those in these underlying numbers. Let’s dive in.

Overall

As always, the charts first:

Another good rushing performance from Tech, buoyed by an outright good defensive performance. I’ll leave evaluating the defense to Robert tomorrow, so let’s dive deeper into this day on the ground. QB Jeff Sims (mostly) made good decisions on option reads, netting 96 yards and 0.21 EPA/play on 15 carries, while RB Dontae Smith returned to the running back rotation to add 31 yards and 0.62 EPA/play on three carries of his own. RB Hassan Hall had a down day by advanced metrics, collecting -2.07 EPA (-0.12 EPA/play) and 70 yards on 17 carries. Especially during the second half, Tech seemed to run Hall right into the teeth of Duke’s defensive line and especially on first downs, which cost him tremendously in terms of EPA.

Expected Points

One of the metrics I like to look at to determine how consistently effective a team is its EPA/play without explosive plays. As we discussed last week, averages can be misleading when outliers come into play, so looking at averages with the outliers stripped out can give us an idea of underlying performance. Tech benefitted greatly from explosive plays through the air (a 3.07 EPA defensive pass interference call, a 2.97 EPA touchdown pass, and a 2.73 EPA touchdown pass all count here — GameOnPaper’s filtering and aggregation logic isn’t complex enough to exclude penalties), but were much more consistently effective on the ground without needing to take as much advantage of explosive plays as they did last week. It’s possible that Tech’s low mark in terms of non-explosive passing EPA could be explained by its focus on throwing screens in this game; per SIS, 36% of Tech’s passes were at or behind the line of scrimmage.

This mention of passing tactics brings us to our continuing discussion of Jeff Sims’ performances under center. By the metrics we have available on GameOnPaper, Sims had a meh to decent day: his 0.08 EPA/dropback would have ranked in the 53rd percentile of all 2021 single-game passing performances and through the third quarter, he had a CPOE just under +1.5% on 4.70 air yards per completion (9.94 per attempt). In terms of overarching metrics, GameOnPaper’s expected QBR model rated Sims at 45.5, cementing in my mind that Sims had more or less an average day throwing the ball paired with a pretty good day on the ground. However, Pro Football Focus (PFF) seems to disagree with this assessment of the situation:

We discussed Sims’ performance based on the eye test versus metrics on this week’s episode of Scions of the Southland, but after some review, there’s two dimensions of the situation we didn’t cover there:

1. It’s possible PFF gives rushing and passing performance equal weight in their offense grades.
2. xQBR, QBR, and EPA are inherently based on each play’s result, which is different from PFF’s philosophy on player grades:

We aren’t grading players based on the yardage they rack up or the stats they collect. Statistics can be indicative of performance but don’t tell the whole story and can often lie badly. Quarterbacks can throw the ball straight to defenders but if the ball is dropped, you won’t see it on the stat sheet. Conversely, they can dump the ball off on a sequence of screen passes and end up with a gaudy looking stat line if those skill position players do enough work after the catch.

PFF grades the play, not its result, so the quarterback that throws the ball to defenders will be downgraded whether the defender catches the ball to notch the interception on the stat sheet or not. No amount of broken tackles and yards after the catch from a bubble screen will earn a quarterback a better grade, even though his passing stats may be getting padded.

PFF’s process-oriented workflow will have different outputs than GameOnPaper’s and ESPN’s results-oriented workflows, but that fundamental difference in grading ends up helping showcase a facet of fans’ tension with Jeff Sims’ performances: on a play by play basis, it seems like Sims is doing the right things, but it’s unclear whether said things are the most optimal or effective.

Situational

From last week’s column:

Per [Sports Info Solutions] and Robert’s charting, the Jackets ran constantly on first down (62%), early downs in general (57%), second and long (53%), and third and long (50%). I harped on this last time, so I won’t make too much of a meal of this point again, but after five games of unflinching inefficient rushing, it’s clear that this is a concept Tech is married to, for better or for worse (read: for worse).

Things were better this week! Per Sports Info Solutions (SIS), Tech rushed on 54% of its first downs, 33% (!!!!) on second and long, and 14% (!!!) on third and long. Credit where it’s due: the offensive staff seems to have progressed towards optimal play calling quite a bit in a single week, though the first-down rushing rate still leaves something to be desired. It’s possible that better weather and more effective quarterback play may have facilitated this shift, but regardless of the reasons, it’s good to see.

However, there is some bad with this good: Tech’s playcalling while up two or three scores in the second half got extremely, extremely conservative at a time in which Tech should have been both bleeding clock and putting the game to bed. This downturn in performance can be seen pretty well in the Expected Points chart:

From GameOnPaper.

Tech let its foot off the gas during the latter stages of the third quarter and all of the fourth quarter: lots of rushes on first and second down that then telegraphed medium-range throws on third down, which later forced punts. Some grace should be allowed here: it’s hard to properly manage to finish off games when you’ve rarely done it before. However, in the future, Tech needs to take advantage of these kinds of opportunities to take shots to end games.

To that end, nailing some critical fourth down decisions over the course of the game could have helped Tech keep things clean late on. The first flubbed decision ended a drive with a field goal from the Duke 5 during the first quarter:

Acquiring four more points here most likely means Tech avoids the penalty and overtime tomfoolery that actually ended the game. Sure, the historic conversion rate in this situation is 33%, but there are three reasons why I would go for it in this spot:

1. It’s the first quarter. It’s not like failing to convert here would put Tech behind the eight ball in the middle of the fourth overtime or something.
2. A 1% difference between the win probability after a successful FG and a failed conversion attempt is virtually negligible.
3. Even if Tech were to fail the conversion attempt, Duke would be backed up on its own side of the field and would most likely have been forced to punt from its own end zone, meaning that Tech would have gotten the ball back in plus territory.

Another frustrating fourth down decision of the weekend saw Tech punt on fourth and two from the Duke 36 right before halftime:

Let’s walk out the possible negative outcomes here:

1. Tech fails the conversion attempt and gives Duke the ball back at its own 36.
2. Tech misses the field goal because college kickers usually don’t hit kicks from 53 and gives Duke the ball back at its own 36.
3. Tech punts the ball and it bounces into the end zone for a touchback, giving Duke the ball at its own 20.

Let’s toss out the field goal option due to the kick distance and focus on punting versus attempting the conversion. It is very, very difficult for a punter to get a punt to bounce either A) out of bounds at the opponent 5 OR B) on a dime in-bounds in the middle of the field without going into the end zone. Punting in this spot is effectively betting that the likelihood of the punter pulling off this manuever is higher than the chance the offense will get two yards.

You can get two yards. You have a bunch of space to work with.

Extending this drive and potentially scoring points before half-time increases your chances of winning the game. It’s better to attempt to do that that than rely on the extremely-low-percentage chance an oblong ball will bounce in just the right way to guarantee the worst possible field position for your opponent.

Granted, neither of these critical fourth-down decisions came during the aforementioned period of bad playcalling during the third and fourth quarters. However, making the right choice at each of these junctures could have set Tech on a path to avoid putting the fate of their win in the hands of a flag-happy referee late on.

We had a great discussion about some of the coaching decisions from this game in this week’s episode of Scions of the Southland, but the critical portion is this: you can write off Key’s game management (see: fourth-down decisions above) as a first-time head coach making mistakes with a career of all of two weeks long, BUT you can’t as easily write off other parts of Tech’s offensive identity, like its dogmatic adherence to running the ball on early downs or its screen-heavy passing game, especially when Tech has opportunities to take even marginal risks to finish off games.

The pieces are all there for a very dynamic, effective, and eminently watchable offense: motion, RPOs, and formational diversity are fundamental to some of the best offenses at both the college and pro levels. The complete picture might not be unique as a program brand as Tech’s previous offensive identity, but it could be very effective at moving defenders around to create space and then exploiting that space to devastating effect. It’s just up to Tech to put all of these pieces to good use.