This season I have seen a lot of chatter about how poorly our defense has played, especially on third downs. Well me being me, I decided to look into just how bad Georgia Tech is performing on third downs this year. The raw numbers don't paint a pretty picture; Georgia Tech's defense ranks 110th out of 128 teams in the country in third down percentage. Georgia Tech's offense ranks 1st in the country. But the fact that our offense always ranks so high in third down percentage got me thinking that perhaps some teams just have easier 3rd downs than others. Intuitively third downs with longer distances to go are harder to convert, and the numbers back that up. I came up with an expected conversion rate for each third down a team faced based on the distance to a first and if the team was close to their own endzone (for some reason third down rates plummet when you are backed up near your own endzone, coaches are probably scared to try and gain a first for fear of a safety). Using this we can see if Georgia Tech's offense and defense have been lucky or unlucky this year in their third down performance. I'll save a lot of the caveats of this analysis for the end of the article.
3rd Down Conversion Percentage; Actual vs Expected
So how has our defense fared? The following plot shows each FBS teams defense's third down conversion rates; their actual third down percentage allowed is on the x-axis and their expected third down percentage based on the distance the offense has to go is on the y axis. Georgia Tech is the gold square. Also my numbers are slightly different from what cfbstats.com reports and I couldn't find any reason that was so, but this is what we have. The size of each dot is the number of third downs each defense has faced. I restricted this analysis to only look at games between FBS opponents.
Georgia Tech's third down defense has been bad this year, but this shows they have been even worse than you would expect given the opposing offense's distance to a first. By my calculations we allow opposing offenses to convert 45.4% of their 3rd downs, but we would only be expected to allow a 39.6 conversion percentage if only the distance to the first down line (and if an offense was backed up near their own endzone) mattered. The average 3rd down conversion percentage is right around 40%, so our defense is about average in the difficulty of 3rd downs we force. Our issue is that we allow offenses to convert those 3rd downs at a much higher rate than average.
Our offense has been a completely different story. Here is the same chart as before but now comparing Georgia Tech's offense to the rest of the country.
Georgia Tech's gold square is almost off the charts. Our 3rd Down rate of 60.3% is tops in the country but our expected 3rd down rate of 42% is also very good. This means our offense is putting itself in easier than average to convert third downs and is converting them at an incredible, and probably unsustainable, rate.
Conversions Over Expected
Another thing we can look at is seeing how many third down conversions we are gaining or allowing over the expected amount we should be based on my model. One thing that I noticed is that both our offense and defense have very few third downs compared to the rest of the country. There are probably many factors responsible for this, one being we don't run that many plays to begin with. Our defense may be so bad that we don't even get to third downs, the offense just converts on 1st or 2nd down. And the reverse of that could apply to our offense. Regardless, the following plot shows how many conversions we allow over expected against our expected third down conversion percentage. A positive amount is bad for the defense, because we have allowed more 1st downs than we should.
This kind of gives me hope, at least we aren't the worst. Of course, the number of first downs we allow over expected is going to be lower than other teams because we don't even face that many third downs relative to everyone else so this could be understating our terribleness. And here is the same thing for our offense.
Because we are busy with the actual season I didn't have time to make this analysis too in-depth like I would in the off-season. Here are a couple further questions I would like to answer:
- Adjust these numbers for opponent. Is converting 3rd downs against our defense the same as against Virginia Tech's?
- This may simply be a measure of how good an offense is. If you gain a lot of yards on average on 1st and 2nd downs you'll probably do the same on third downs. Is there a way to isolate the skill of converting third downs specifically?
- Is this even a skill? Are third downs random over time?
Any questions or insights of your own? Put them in the comments!