The Success Rate of a Talented Clemson Offense
A look into advanced metrics for defining offensive efficiency in college football
We all know that feeling. It’s 3rd down and 8. Your quarterback drops back, finds a receiver on a quick slant five yards down the field — but, the thing is you still need three more yards. Nothing makes a college football fan more upset than coming up short.
You understand that even if your team gains yards, the play was a failure. Had you had that result on first down it might have been a different story, but because of our understanding of the situation, the play leaves us pulling our hair out rather than clapping our hands.
This is exactly the intuition behind Success Rate. Success Rate is a beloved measure of efficiency in which a play is deemed a success if a team gains 50% of the needed yards for a first down on first down, 75% on second down, and 100% on third and fourth down. Using this understanding we can begin to understand the efficiency of offenses (and defenses) in college football. Success Rate has great predictive power, as well, when comparing two team’s rates within a game. It is helpful to understand that the average across FBS is typically about a 40% conversion rate.
Against Auburn last Saturday, the Clemson offense showed flashes of talent, but failed to show any semblance of a true rhythm besides two mid-game scoring drives. Thus, Clemson, in 66 plays, had just 27 “successful” plays for a success rate of 40.9% (13 of the successful plays came on the two scoring drives). While this is about average for college football, this falls short of the norm for the high-powered Clemson offense in the past few seasons — averaging 43.8%.
However, having just this simple success rate measure has limitations. It is one thing to put up a 56.4% success rate against Kent State and another to be successful against a high-caliber defense, like an Alabama. In order to adjust for some of the added factors, I decided to explore the distributions of Clemson’s success rate against a number of situations. For this analysis I used beta distributions to model the success rate for home and away games and games against ranked and unranked opponents using play-by-play data from 2011-present.
As an example, here are the distributions of Clemson’s success rate for home games against ranked (orange) and unranked (purple) opponents. Note that the unranked distribution is shifted to the right — a higher success rate — and more concentrated — this is due to more evidence to allow for a better estimate.
Against Auburn last week, the Clemson offense (40.9% success rate) performed slightly worse than it did when it faced Auburn last season (42.3%). Based on the success rate, it would seem as if Clemson’s offense fared better against Auburn last year. However, due to the fact that Auburn was a better team this season (and Clemson has surprisingly performed slightly better offensively against ranked opponents on the road than at home), we can estimate that the Kelly Bryant led offense had a more impressive performance than the Deshaun Watson led offense, at least in comparing these two games.
By comparing the success rate of each game to the cumulative distribution function (CDF), we can measure the likelihood of the typical Clemson offense achieving a success rate as high as the team in fact did perform. We should be somewhat careful with this analysis in comparing different distributions, but it is a helpful exercise to quantify the efficiency. Thus, against Auburn in 2016, there is a 76% chance that the typical Clemson offense would have a success rate greater than 42.3%, given the situation. Yet, the 2017 Auburn matchup would lead to just a 64% chance of doing better — still not a great offensive performance, but better than 2016 . Remember that lower percentages are better because they equate to the performance being more outstanding compared to the distribution.
This “Chance of Performing Better” (CPB) metric can allow us to look at recent Clemson history by utilizing the beta distributions to take into account the efficiency of Dabo Swinney offenses. Particularly, with ranked opponents like Louisville and Virginia Tech upcoming on the 2017 schedule, Clemson fans might be interested in the success the Tigers have had against ranked opponents. Here are the best 11 offensive performances against ranked opponents in the past 6 seasons as ranked by CPB:
Of these games only the 2015 National Championship game and the 2012 game against South Carolina were losses. Looking at the defensive numbers will help to better understand the final result. Of the top 10 games, 5 were post season performances — a remarkable thing to consistently be peaking at the right time of year.
Over time Clemson has had differing levels of offensive efficiency. In 2014, as the keys to the offense were eventually handed over from senior Cole Stoudt to freshman Deshaun Watson, the offense felt a dip.
Many have speculated this might be another transition year, as Kelly Bryant takes over at quarterback. However, he has appeared poised and has talent scattered around him, demonstrating why the Clemson offense has performed well in the opening two games. The big test will come in Louisville, Kentucky this Saturday for another primetime contest.
There are many avenues this analysis could continue down in the future. Adding more complexity to the factors that create the distributions — such as how good the opponent is rather than just ranked versus unranked — would allow for a more detailed ranking system, but would need to be done using a larger sample size, likely from all of college football. Also, spreading this analysis to look at the defensive side of the ball might provoke an understanding of how the results of game occurred.
Yet, this further analysis is for another day. Today we get to fret about coming up short on third and long. With two games in hand this season, the Clemson offense must continue to grow as new challenges arise on the schedule. College football fans know, at heart, what a successful play is but manufacturing them at a high rate is a hope of everyone on Saturday afternoons.
Data thanks to Seldom Used Reserve and Sports Reference. Please email me at aepowell95@gmail.com for further questions or follow me on twitter @aptigers12.
