Football: Yellow Jackets @ Duke Advanced Stats Preview

Author's Note: I'm gonna be introducing some new stuff this season, so if you are confused by anything please ask away so I can make sure I'm explaining everything correctly.

You guys and gals obviously remember last year's disappointing loss to the Dukies, where an injury to Justin Thomas severely limited the capabilities of our offense. Since then Tech has improved rather drastically and that improvement, at least in the computer models, has carried over into this season. Vegas has Tech as anywhere from an 8 to 9 point favorite.

Computer Ranks

S&P+ is a model built by Bill Connelly, SB Nation's resident Analytics guru, FEI is the Fremau Efficiency Index, a possession based rating system, FPI is ESPN's CFB ratings metric, and MELLS is my own model I'm working on to predict the score of each game, it still has some kinks and isn't as accurate as the others but I'm the writer so we are gonna use it.

When Georgia Tech has the Ball

If there is one thing people are clear on its that Duke's pass defense is phenomenal. I have them rated as the 6th best pass defense in the country, while they rank 4th and 17th in S&P+'s passing success rate and explosiveness measures. Their run defense isn't as great, ranking 43rd in my metric and 25th in S&P+. Still better than average, though. I still have Tech's offense has one of the top 5 rushing attacks in the country, even after getting shut down against ND. S&P+ rates us a little worse at 10th overall. Our passing attack, however, is struggling to start the season. S&P+ rates us as the 107th best passing attack in the nation, which needless to say isn't good. My model has us at 30th best, but I think that will drop as the bad teams we beat up on in the beginning continue to play badly.

One thing I've been working on with my own model is how to show the impact of these different unit's ratings on the actual amount of points a team can be expected to score in a game. The visual below shows how many points above average each unit contributes towards the predicted points scored by Georgia Tech on Saturday, in this case 24.5 points.

For example, Duke's Passing defense is 4.5 points better than average, while Tech's passing offense is worth only 1.9 points better than average, so overall the passing advantage when Tech has the ball is 2.6 points for Duke. Do this for all the units and you get that Tech is projected to score 1 points less than an average team, which is normally 25.5 points. I'll have a post later about how accurate the model is, but it definitely gives you a rough idea of the value of each unit.

Why is Duke rated so highly on defense? They have shut down most every passing attack they have faced. The following chart shows the percentile performance of Duke's Expected Points Added per Passing Play in the games they have faced so far. For example, against Northwestern, Duke's defensive effort on passes was better than 92% of defensive efforts this season. The percentile is from the offense's perspective, so its also the 8th percentile of passing performances by Northwestern. These numbers aren't adjusted for opponent (the numbers that go into the MELLS model are) but they at least give an idea of each team's performance in each game.

When Duke has the Ball

This matchup is much less exciting. S&P+ has Duke's rushing offense rated as the 101st best rushing attack in the country. My model has them as 63rd best, still a decidedly average rushing attack. Tech's Rushing defense isn't any better. S&P+ rates them as the 10th worst team in the country at allowing explosive rushing plays. Though at this early in the season that may be mostly a product of Notre Dame's one long rushing touchdown.

Duke's Passing game is rated as 23rd in the country according to S&P+, but their raw stats are much worse, rating 96th in the country in passing success rate and only 45th in explosive plays. Their QB's stat line is actually pretty impressive though. Thomas Sirk has 754 yards through the air already with 5 TDs and only 1 INT. He has also only been sacked once all year, and Georgia Tech's Adjusted Sack Rate ranks 78th in the country so don't expect a big night of pressure. Although, anecdotally, it does seem like we have gotten a lot of pressure this year even if we haven't gotten many sacks. S&P+ rates our pass defense as 45th in the country while I have us at 38th best in the country.

My model projects Duke to score 28.8 points, though they receive a nearly 5 point boost for being at home.

Negative Values mean that unit is worse than average. So Duke's rushing attack is about one point worse than an average team, while our defense is .7 points worse than an average rush defense. In total our defense has a .3 point advantage over Duke in the rushing category.

Game Prediction

All the models think Georgia Tech is the better team, and all but the MELLS model favor Georgia Tech even on the road. While my model thinks GT is the better team it thinks Duke has a 56% chance of winning thanks to home field advantage, with a projected score of 28.8 - 24.5. Brian Fremau just released his game projections and has GT winning 36-24, with a 76% chance of winning outright. Has us 13 points better than Duke, but Duke has about a 3 point home advantage in their model, so it projects us to win by 10. They also list us having a 78% chance of winning outright. Bill C just released his projects and he has Tech winning, but only by 2-3 points with a projected score of 25.6 - 23. He gives Tech a 56% chance of winning outright.

As you can see there is a pretty good consensus, even with Vegas, that Tech should be favored but by how much does vary. The odds of us losing are just a bit better than the odds of you flipping a coin twice and getting two heads in a row, so it's not like this is a guaranteed victory by any means. If Tech's offense continues to struggle we need to hope our defense is for real this year to shut down Duke's offense.