This won’t be a normal Advanced Stat Review because frankly, Saturday was bad, and we know it was bad. In fact, I contemplated just coming in and leaving the article at one of these statements:
- I made the decision to turn the game off before halftime and go with my wife to the Georgia Southern game. I made the right decision.
- Fire Andrew Thacker.
One of those sort of came to fruition with Thacker’s demotion from Co-Defensive Coordinator to Safeties Coach. Anyway, this article will be a bit more of an amalgamation of random here-and-there stat things that stood out to me, and then I want to look a little deeper into what Kevin Sherrer brings to the table as the sole defensive coordinator.
How the hell do you lose a game and end up with a post-game win expectancy of 75.3%?
So in my advanced stat review of the Ole Miss game, I dug deeper into what goes into the stat of post-game win expectancy to figure out why Georgia Tech ended that game at 0% despite it seeming closer than it was.
As a quick refresher, Bill Connelly’s post-game win expectancy considers 5 different in-game statistical categories:
- Explosiveness (using points per play)
- Efficiency (using success rate)
- Drive-finishing (using points per trip inside the 40)
- Field position battle (using average starting field position)
- Turnovers (using turnover margin)
After getting all of these numbers, I think I’m even more confused.
Georgia Tech only beat Bowling Green in Success Rate, it seems reasonable to think that the PGWE should almost be the other way around. That was my first thought. I chatted with Akshay about this, and he brought up that although Bowling Green won the turnover battle by a large margin, that also produced significant turnover luck for Bowling Green.
The Yellow Jackets were projected to turn the ball over 1.4 times, but because they turned the ball over 3 times, that produces a significant amount of turnover luck for Bowling Green. The way that GameOnPaper describes turnover luck is through points. Bowling Green finished at 12.8 points of turnover luck. The final score of the game was 38-27, so if you take away the roughly 13 points of turnover luck, that game becomes a 27-25 Georgia Tech victory.
It’s quite strange how that works, but that is my understanding of why Georgia Tech finished with such a high PGWE: because those penalties above expected are not guaranteed, thus causing turnover luck. Why did that cause such a drastic swing in PGWE? Honestly, I don’t have a clue, because Bill C.’s PGWE model is not public, so I can’t see the specific formula used to calculate it.
So every week, I’ve been including a graphic of P5 quarterbacks that compares how they scored in ESPN QBR vs. their PFF rankings. Every single week—without fail—Haynes King has been in the right quadrant, scoring higher in QBR comparatively to PFF. This week, though, things are a little different.
King had his best performance by PFF’s standards to date, despite having a QBR around 30. From my understanding of how PFF’s grading works, this is saying that King’s poor performance is more due to the performance of those around him, that he is not necessarily at fault.
Generally, this whole game just sounds like a bit of a weird anomaly statistically.
What does Kevin Sherrer bring to the table?
After several long years, the words that many Georgia Tech fans wanted to hear have finally come to fruition: Andrew Thacker is no longer Georgia Tech’s defensive coordinator. Taking his place is Kevin Sherrer. Sherrer has more experience as a linebackers coach, but he also served as the co-defensive coordinator for Tennessee in 2018.
How’d he do there?
I would say he did okay. Overall, his defense kept opponents to a negative EPA in passing and rushing, but that success rate is concerning. Georgia Tech’s pass defense has already noticeably struggled this season, and a 46.4% success rate allowed on pass plays is really bad.
Given the work he has done with linebackers at different spots (producing NFL guys like Leonard Floyd and Roquan Smith), my hope is that he will be an improvement, but unfortunately, the sample size is small, so it’s tough to say.