clock menu more-arrow no yes

Filed under:

Bye Week Reflections: Georgia Tech by the (Better) Numbers

New, 50 comments

Let’s talk intelligently about how the Yellow Jackets are doing.

Georgia Southern v Georgia Tech Photo by Daniel Shirey/Getty Images

You may have heard earlier this week, but Paul Johnson went onto the new “We Never Played the Game” podcast with WSB’s Zach Klein and the AJC’s Jeff Schultz for a fantastic in-depth interview that covers a ton of ground from his personal life to his time at Georgia Tech. If you haven’t listened to it yet, I can’t recommend it highly enough.

Within the podcast, Johnson throws some shade at a few different folks, one of whom is “Jeff’s colleague” — Mark Bradley — who often likes to cherry-pick statistics and use them to suit whatever narrative he likes. That statement clearly got under Bradley’s skin, because it turned into not one, but two articles on Wednesday bashing Johnson and his team. It was a pretty embarrassing look, actually, as Bradley seems to meander through personal musings on the status of the program through history, before proceeding to literally go and cherry-pick some statistics in response to Johnson’s comments about his appreciation for doing so.

That said, as engineers, we know that numbers and statistics can be sliced up to say whatever we want them to. So, instead of just looking at the raw data, let’s use some advanced numbers to look at where Georgia Tech really stands 7 games into this season. To do so, we’ll use Bill Connelly’s S&P+ numbers, a whole rundown of which can be found here for Georgia Tech. Also, I’ll try to give a brief summary of how each works before diving into it, but I’d also highly recommend listening to Podcast Ain’t Played Nobody from two weeks ago, when Bill spent around 23 minutes (beginning at the 52:00 mark) talking through how his numbers work.

Five Factors

Category Offense Rank Defense Rank
S&P+ 29.8 58 26.3 49

The overall S&P+ ranking for Georgia Tech (and every team) comes from rating the offense and defense on what Bill refers to as his “Five Factors”. Effectively, the concept here is that rating teams (and winning games) is based on their ability regarding five things: efficiency, explosiveness, field position, finishing drives, and turnovers. You can read about how each of these correlate with win percentage (and therefore why these measures are used) at Football Study Hall. With an overall national ranking of 55 right now, let’s examine how Georgia Tech is doing so far regarding each of these.

Efficiency

This is a measure of how well a team’s offense does at keeping the chains moving, and how well does this team’s defense does at preventing the opponent from doing so. This is measured with “success rate” using a simple evaluation of the team’s play-by-play data. A “successful” play is defined as one where the offense gains 50% of the yards-to-go on first down, 70% on second down, and 100% on third- or fourth-down. Nationally, average success rate is 40.7%. For Georgia Tech right now, the numbers look like this:

Offense Defense
Avg. Rank Avg. Rank
Success Rate 42.60% 60 44.30% 98

Efficiency-wise, the offense is slightly above average, while the defense is comfortably below average. These numbers make sense — the offense doesn’t often gain 5 yards on first-and-10, and typically thrives on converting third downs. Meanwhile, the defense plays a “bend-don’t-break” style that regularly gives up a lot of yards on early downs and tries to tighten up on third down.

Explosiveness (Big Plays)

This is a measure of just how successful a team’s “successful” plays are. This is measured using “IsoPPP”, or “Isolated Points per Play”, which removes unsuccessful plays from the data set and determines how many points a team scored per successful play. The theory is that making big plays and picking up yards in chunks removes opportunity for error — there are more opportunities for mistakes in a 17-play drive than in a 7-play drive, even if both go for touchdowns. The national average in IsoPPP is 1.27. Here’s how Georgia Tech stacks up:

Offense Defense
Avg. Rank Avg. Rank
IsoPPP 1.35 34 1.14 18

Again, this makes sense in a lot of ways. Where Georgia Tech can struggle in success rate, it gets quite a bit of value out of hitting big plays. It’s why Clinton Lynch is averaging over 17 yards per touch this year, or why all four of the team’s leading receivers are averaging over 14 yards per reception. Defensively, this also makes sense with the team’s “bend-don’t-break” style. Georgia Tech’s defense is 23rd nationally having allowed only 23 plays of 20+ yards this year, and 3rd nationally having allowed only 6 plays of 30+ yards. In some sense, the goal of not giving up big plays on defense is being achieved, even if that comes at the price of efficiency.

Finishing Drives

Another important measure of a team’s success is how they finish drives — are their drives ending in touchdowns, or are they being held to field goals? This is measured in “Points per Scoring Opportunity”, where a Scoring Opportunity is defined as an occasion where a team runs a first-down play from inside the opponent’s 40-yard line (or scores a touchdown on a play of 40+ yards). The national average here is 4.67 points per opportunity. Georgia Tech compares like this:

Offense Defense
Avg. Rank Avg. Rank
Points per Opp 5.05 45 4.64 60

Truth be told, finishing drives has been quite a bit of an issue for Georgia Tech’s offense for a lot of the year. Aside from the Clemson game (where the team scored a touchdown on its only scoring opportunity of the game), the team only has two games where they’ve scored more than 5 points per scoring opportunity — the last two, against Pittsburgh and Georgia Southern. It was particularly ugly against Boston College, when the team came away with 17 points on 5 opportunities (3.4 points per opp), and against Miami, when they came away with 21 points on 5 opportunities (4.2 points per opp). Defensively, things started well but have really degraded lately. After holding Boston College, Mercer, Vanderbilt, and Clemson to under 3.5 points per opportunity (forcing a lot of field goals), the last three games have seen opponents all manage over 5 points per opportunity (lots of actually breaking in the red zone). The offense seems to be improving here, while the defense has really degraded throughout the season.

Field Position

This is far simpler than the previous measures. Quite simply, this is a measure of a team’s starting field position. That may seem painfully simple and meaningless, but a better starting field position means less ground to cover before achieving a scoring opportunity -- fewer chances to screw up and stall out. This is measured, quite simply, with an average of where the team is starting drives. Here’s how that looks for Georgia Tech, with a national average of starting at the 29.6 yard line:

Offense Defense
Avg. Rank Avg. Rank
Average FP 29.4 77 27.8 42

Here, the defense benefits from the offense’s ability to move the ball, and from good special teams play when the offense breaks down (or on kickoffs). Meanwhile, this is where the offense suffers from the defense’s “bend-don’t-break” style. Giving up a lot of yards can be fine if they don’t turn into points, but those cases can also end up with the offense backed up deep into its own territory.

Here’s another interesting quirk: Georgia Tech is 4-1 in 2016 when winning the field position battle (the lone loss was to Miami, where the Jackets had a 1.1-yard advantage), and has yet to win a game where they lose the field position battle. (Clemson had a 13-yard advantage, while Pittsburgh had a 7-yard advantage.)

Turnovers

Admittedly the toughest to quantify, especially from someone who strictly considers turnovers to be luck-based. Still, there are ways to quantify a team’s “turnover luck” as compared to that of others, based on how many passes defensed turn into interceptions, or how many fumbles are recovered. This is the section that, admittedly, I don’t understand particularly well, although Bill does an excellent job of talking through it on the podcast above. Effectively, pass defense statistics and fumble statistics are crunched to determine a team’s expected turnover margin, then compared to its actual turnover margin, with the difference converted into a “Turnover Luck” number in terms of points-per-game. Average turnover luck per game would be 0.00 PPG. Right now, Georgia Tech averages +0.13 PPG in turnover luck — meaning, based on how the defense plays the pass and how the fumbles have turned out in their games, the Yellow Jackets have gotten marginally lucky in the sense that their turnover margin on the year is -2, as compared to the projected margin of -2.18.

Aside from “death by turnovers” that was the Miami game, this played a big role in the Boston College game, where Georgia Tech was +2 in turnover margin versus a projected +1.17 margin -- good for 4.2 points in turnover luck, in a game that ended up a 3-point win.

Win Probabilities

From here, the S&P+ takes in all of this information on a game-by-game basis and calculates win probabilities based on margins in each statistic. In other words, look at the Vanderbilt game: Georgia Tech won success rate by 3%, IsoPPP by a comfortable margin, points per opportunity by 2.42 points, field position by 13.9 yards, and got 0.04 points in projected turnover luck. Given those “five factors”, Georgia Tech could be expected to win the game 100% of the time.

Based on that, Georgia Tech’s win expectancies have looked like this so far this year:

Date Opponent Win Expectancy
3-Sep Boston College 21%
10-Sep Mercer 99%
17-Sep Vanderbilt 100%
24-Sep Clemson 0%
1-Oct Miami 6%
8-Oct Pittsburgh 25%
15-Oct Georgia Southern 99%

Georgia Tech has won 3 games without any real concern, lost two games where they had no real shot of victory (according to the statistics), and has split a pair of games that they probably shouldn’t have won (again, according to statistics).

These numbers are all crunched for all other teams, too, and can be used to project win probabilities when playing a certain opponent. Here’s what the rest of the road looks like for Georgia Tech:

Date Opponent Opp. S&P+ Rk Win Probability
29-Oct Duke 72 68%
5-Nov at North Carolina 34 34%
12-Nov at Virginia Tech 16 22%
19-Nov Virginia 79 70%
26-Nov at Georgia 67 55%

There might be a row on this table that sticks out to you, in particular, and I’d recommend that you approach it with a bit of a sense of caution. Yes, on October 22, S&P+ gives Georgia Tech a 55% chance to win in Athens against their rivals on the Saturday after Thanksgiving. That number will change quite a bit between now and then, so don’t count your chickens before they hatch on that one (as you’re fully aware by now, if you’ve watched any of the last 15 years of that rivalry). That said, the conclusions you should draw from this are as follows:

  • Duke and Virginia really should be wins.
  • UNC is beatable, though you shouldn’t necessarily expect a win there.
  • Virginia Tech is not very beatable. Adjust your expectations accordingly.
  • georgia is beatable. Quite beatable. (Rivalry weirdness aside.)

Conclusions

Here are a few things that you should take away from all of this:

  • Georgia Tech has been an average-to-above average team this year, depending on the week. Their 3 losses are all to teams ranked higher in S&P+, and the 4 wins are over teams ranked lower. If you weren’t shocked at the outcome of any of the games so far, that’s a good starting point as for why.
  • The offense is trending upwards over the last few weeks, while the defense is trending downwards. Statistics back that up, particularly regarding the “points per scoring opportunity” margin. (This is what happens when a “bend-don’t-break” defense starts to consistently break.)
  • The road ahead is manageable, and it’s entirely within reason that Georgia Tech finishes the regular season anywhere from 6-6 to 8-4. (Cumulative probability of that happening, per S&P+: 82%. That includes a 68% chance the team finishes either 6-6 or 7-5.)
  • Advanced statistics can really help us to objectively evaluate a team’s abilities and performance. They are far more useful than “Total Defense ranking”. Looking at you, @MarkBradleyAJC.

There are a lot more methods of advanced statistical analysis available to us regarding college football (specifically the F/+ numbers from Football Outsiders, which are more per-drive focused rather than per-play focused and tend to like Georgia Tech a little more), and I’d love to dive into them deeper in the future. For now, use these to adjust your expectations and understand what we’re watching when the Yellow Jackets are on the field. Keep in mind, too, that there’s plenty of room for improvement (or degradation) left this year.

Just stop letting your defense break, Coach Roof.