Georgia Tech Football: UNC Advanced Stats Review – Offense

A tale of two halves unfolded in Chapel Hill on Saturday afternoon, as Georgia Tech dispatched a ranked North Carolina team for the second year in a row. For a moment in the first half, it looked like 17 points would be enough for UNC to claim victory, given the ineffectiveness of Tech’s offense. However, the fortune of continued Tar Heel mistakes and a few well-executed drives in the second half helped the Yellow Jackets secure victory.

First, the charts:

Three major thoughts for you, based on the numbers:

1. Consistency

I was extremely impressed with Tech’s offensive performance in EPA/dropback (both with and without explosive plays) AND in success rate, both at the game level (more on this in a second): producing 0.30 EPA/dropback (75th percentile of 2021 FBS performances) and 50% success rate (85th percentile) for the game is one of the best performances of the season and, given the struggles of the offense this year, a hopeful sign of progress.

However: if you filter down to the half level, things get a bit muddy:

This difference also plays out in the total EPA over time chart above. While Tech played extremely well in the second half, it looked extremely out-of-sorts in the first. Like I led with: for most of the first half, it seemed like 14 points would be an impassable margin for Tech, given what we know about its offense. But Tech put two well-structured, steady drives around halftime to get within three points and then turned in two more late to take the lead and later ice the game. There are a few ways to look at this:

1. Tech’s offense really struggled in the first half and made some key adjustments to win in the second.
2. Tech’s offense executed well when the game was on the line.
3. UNC plays extremely, extremely poor defense.

I think it’s a mix of all three: Tech made adjustments and executed when it needed to, taking advantage of a UNC defense that ranks in the bottom 30 in the nation in most advanced metrics.

2. Offensive Line Play

The most scathing indictment of UNC’s defensive line might be able to make is that it failed to generate a single sack against a Tech offensive line that has a tendency to be a liability AND lost a few starters to injury over the course of this game. That isn’t to say that UNC didn’t make quarterbacks Zach Gibson and Taisun Phommachanh uncomfortable in the pocket: per Robert’s charting, UNC generated pressure on 46% of Tech’s snaps. But despite that, the Tar Heels only generated four havoc plays (6% havoc rate, 18th percentile performance) over the course of the game (and only one — an interception — in the air) and again zero sacks.

Granted, the other three havoc plays were three stuffed runs AND UNC limited Tech to two or fewer yards on 43% of its rushes, but that doesn’t mean Tech didn’t find some joy on the ground. Tech generated rushing opportunities on 41% of its rushing attempts, with the offensive line responsible for ~2.16 (line) yards per carry.

On the whole, Tech’s offensive line had a fairly competent day, even if the unit they faced wasn’t quite up to par.

3. Early Downs

This is a drum I’ve beat multiple times this season that it’s nearly time to (hopefully) retire: Tech needs to stop “establishing the run” and start being more aggressive on early downs. Tech was effective overall on early downs overall in this game: 0.10 EPA/play with a 49% success rate, but it’s clear how that effectiveness was generated: Tech produced a 67% success rate on 15 early downs passes, compared to a 43% success rate on 42 early downs rushes. Put another way, Tech was 1.5x more successful on early downs passes despite rushing three times more often.

It’s completely irrational to continue to hammer away at first-and-ten and (almost assuredly immediately) second-and-long rushes before telegraphing a pass on third down. Even replacing the aforementioned obvious third-down pass with a (somehow even worse) third-and-eight draw late in the fourth quarter nearly cost Tech this game. Tech has long called its offense “pro-style”, but even pro teams don’t do this kind of thing anymore — it’s completely ridiculous to continue to stick to this playcalling dogma despite knowing you can generate success other ways.

As we head out: a note on Georgia Tech’s evolving coaching search, starting with something I wrote a few weeks ago after Tech lost in Tallahassee:

Given Tech’s difficulties with recruiting, developing, and deploying quality offensive lines, it seems curious that the offensive line coach has become a golden goose to keep in house to some parts of the fanbase. Given the production Tech has seen at that position throughout the last four years, it just doesn’t seem reasonable to make that call.

Over the last few weeks, I’ve seen no reason to vary from that stance. But let’s take it a step further and be more data-driven — here are the scores and post-game win expectancies for these games:

• Pitt: W 26-21 – GT 43.7%
• Duke: W 23-20 – GT 97.7%
• UVA: L 16-9 – GT 2.5%
• FSU: L 41-16 – GT 0%
• VT: W 28-27 – GT 61.5%
• Miami: L 35-14 – GT 0.1%
• UNC: W 21-17 – GT 41.3%

Tech’s won four games of seven — great on the surface, right? However, the Jackets were only expected to win ~2.5 based on their performances in these seven games. Tech is outperforming its expected wins by 1.5, a mark that would have them tied for fourth nationally in terms of over-performance.

It’s possible that one might raise an eyebrow at some of the processes involved in determining expected wins, so let me offer a different way of looking at the same point for those skeptics: Tech has collected four wins by a combined margin of 11 points (+2.75 PPG).

Those marks are not sustainable, period. In general, Tech has not put together good performances in metrics key to future success throughout this stage of the season. Moreover, these performances have not been of the type that (I feel) should make a strong hiring case. If Tech had won its games versus Miami and Virginia — both of which were winnable both on paper and at various points in-game, the hiring case certainly gets much stronger. However, it failed to take advantage of those opportunities, which has (or, at least, should have) consequences as we near the (supposed) end of the coaching search (given that it would be prudent to have a coach in place before December’s signing day).

At the end of the day, a coaching search is a process. I touched on this theme in a previous column:

As a newer fan (freshman class of 2015), my understanding of Tech’s ethos as a program begins and ends with optimizing for competitive advantage when at talent disadvantage (see: option, flexbone). When we talk about maintaining or building Tech’s program culture, that’s what I think of — fundamentally, I expect the smart school to make smart decisions to win football games.

Smart decisions are borne of good process. Tech has had the opportunity to design good (and notably, data-driven) process here. Will the powers that be execute properly on that process, or will they be beholden to emotional decision-making? We’ll find out soon enough.