Bend or Break Part II (Duke Projections)

Yesterday I posted some statistics to shed some light on the whole "Bend or Break" a.k.a "Ted 'Leaky' Roof" debate (see here). In the comments, GT_Jason asked to see what happens if you add Duke's stats. Since the discussion is a bit long for a comment, you get "Bend or Break Part II: Electric Boogaloo".

TL;DR - The average Tech defense doesn’t affect opponent scoring too much. The good news is that Duke doesn’t score too much. Their defense is good, but our offense should still move. Statistically, we have a decent shot at winning a 24-14 type of game.

I'll look at points per possession (PPP), since that's what matters for win/loss. The other metrics I used in the previous post were used to measure the bend-don't-break debate.

I went ahead and added Duke like GT_Jason suggested, and here's what you get:


Honestly, this doesn't tell me much about how the game will go, because this doesn't account for GT's effect on Duke or vice versa. Note the FBS average = 2.2 PPP. What we can learn is:

  1. We will face a pretty good Duke defense (1.5 PPP), one of the best we've see
  2. We will face a pretty weak Duke offense (1.6 PPP), one of the worst we've seen
  3. This is why they're 3-4

So if we neither help them nor get in their way, we're looking at a nearly tie game. Likewise:

  1. Our offense is pretty good (2.6 PPP), although not great
  2. Our defense is pretty average (2.1 PPP)
  3. This is why we're 4-3 (although those stats are more like 5-2...)

So if Duke does nothing, we're looking at a win. But that's not how football works. For example, if two good defenses are on the field, PPP should drop below average for both teams involved.

We can clearly see in the graph above that when faced with a good offense, GT defense will allow more PPP. When faced with a weak defense, GT offense will score more PPP. Let's look the difference between Tech's performance and each of its opponents' season averages. In other words, how often does each Tech unit out-perform or under-perform?


These graphs are nice because they confirm our feeling about the team's last two performances. The offense is performing really well, while the defense has been faltering of late. To their credit, the defense had a strong stretch between Mercer and Miami. In total: (not accounting for special teams)

  1. Offense Worst game = -0.5 PPP, Average = +0.8 PPP, Best = +1.9 PPP
  2. Defense Worst game = +1.0 PPP, Average = -0.2 PPP, Best = -1.1 PPP

Looking at the averages, the offense is actually quite good, generating +0.8 PPP compared to what the opponent’s defense usually allows. The defense is a mixed bag.

Now we can make a meaningful projection for how GT will perform while adjusting for Duke:


For example, an average day for our offense will generate 2.37 PPP once adjusting for Duke’s defense (1.53 + 0.84 PPP). Even on a really good day, we can expect some drop-off offensively because Duke's defense is quite good. An extreme example was Duke at Louisville.

What does this all mean? Here’s a table of different scenarios that might shake out for a game with 10 possessions: (Score = PPP x 10)


So if the Tech offense and defense simply have an average day, it’s likely a 24-14 win, factoring Duke’s average offense and defense. Actually, it’s even possible to win with either unit having an awful day, as long as the other compensates with a very strong day.


Oh, but I will tell you the odds, Han Solo. If we very roughly assume a 50% chance of an average day, 25% of a bad day, and 25% of a good day for each unit, here are the scenario odds:

69% chance of a win. No guarantees, though. As Coach says, that's why you play the game.

Note for the statisticians: I did not directly account for the variability in Duke's performances. I assumed that Duke can affect Tech in a way that has already been encountered with a different opponent, roughly speaking. I'm not getting paid to do it correctly ;)

FanPosts do not necessarily reflect the opinions of the FTRS writing staff, and are spotlighted purely to create discussion.