Introducing Relative YPP Analysis
If any of you are familiar with institutional investing, hedge fund managers exchange ideas all the time to challenge each other’s theses, sharpen their analysis, and sometimes raise their profile in the industry by making their research public. Whether it’s through one-on one phone calls, lively idea dinners, or the Super Bowl of hedge fund presentations known as the Ira Sohn Conference, the currency of credibility on Wall St is a great idea.
Like many in this community, I came across Graham’s analysis of UGA games last fall and was blown away at the combination of analytics and film study. Graham’s weekly write-ups were as compelling and thorough as anything I cam across in my time on Wall St. He grounded his game theses in rigorous statistical analysis and film study, always presented the risks to his perspective, and made clear how strong the conviction he had in his views. Beyond being chicken soup for our souls to ease our worries on the eve of each week's games, his analysis prepared us to view the game as if we were coordinators calling plays against the opposition. Given my love of UGA football, I reached out to him this past summer and asked if there was any way to collaborate, and he was gracious enough to take my call and flexible enough to give me the choice of what to write about on my own.
So here is my first contribution to Dawg Central. I hope you enjoy it.
As we enter the stretch drive of the college football season, hopefully, the disproportionate focus on Kirk Ferentz and Iowa will give way to actual relevant discussions about contending teams. As we all know, the CFP selection committee releases its first rankings Monday night, but quite frankly, its purpose is to stimulate discussion and generate viewership at a time when that ranking is essentially meaningless. Of greater importance to me is how contending teams fare over the remaining 5 weeks and the regular season and championship week. Looking at their relative performance against their opponents provides insights about contending teams that should help us form our expectations for the next five weeks.
As a longtime fan of sports analytics, the emergence of YPP analysis as a preferred metrics is particularly gratifying because it has elevated the discussion significantly. My goal is to contribute to the disruption of CFP’s “talking heads” grip over the narrative of the sport by through innovative fact-based analysis.
What is relative analysis?
It is analyzing a team’s performance against its opponents relative to the how its opponents’ perform against their other opponents. To be clear, net YPP analysis is great, but we know not all yards are equal and the level of competition matters. Thus, we must control for the level of competition. I do that by excluding all non-P5 games from analysis, as they often skew the data significantly and only analyze performances involving P5/ND/BYU. Then, I calculate how teams performed against their competition relative to the other teams their opponents played on both offense and defense.
Why exclude P5 teams?
Simply put, they skew data. There is perhaps no better example of this than UF’s offensive YPP. At 7.2, Florida’s offense has the 5th highest YPP in college football. Wait what? Yes, you read that correctly. When we exclude all non P5 games games, its offensive YPP is 6.3, which is 23rd in the country. To be clear, I am certain Jimbo Fisher is aware that games against G5 still count, but if you are losing to App State, your team is almost certainly not relevant in the playoff discussion. But let’s all agree that any data should be filtered to exclude non-P5 results.
Why look at things on a relative basis?
With game level data, we have the tools to answer the question how a team’s performance (offensively and defensively) against an opponent compared to the other teams its opponents faced. In short, instead of YPP, we can calculate a relative YPP%, which is defined as OYPP Gained/Avg DYPP allowed (or DYPP surrendered/Average. In the case of UGA, our offense has gained 25% more per play than our opponents usually surrender, and our defense has allowed 15% less than our opponents have gained in their other P5 games, which translates to 41% Net YPP%. On a combined basis, UGA is 41% better than the teams our opponents have faced. Taken together, UGA generates a Relative Net YPP of 2.43, which means our Net YPP is 2.43 ypp better than our opponents other opponents.
I analyzed the top 19 P5 teams in the coaches poll plus future UGA opponents Florida and Mississippi State. Our goal is to answer a series of questions.
How have contending teams done against their P5 opponents on an absolute and relative basis?
When we limit teams to P5 games (and BYU/ND), UGA and Ohio State are significantly above the pack both on Net YPP and Relative Net YPP basis, which should come as no surprise to followers of the sport. We know that few, if any of college football’s contending teams have faced multiple quality teams. Thus, most contending teams relative Net YPP is slightly lower than their Net YPP. Interestingly, Tennessee’s Relative Net YPP is slightly higher than its Net YPP because its relative Net YPP against Alabama was an astounding 3.17, which reflects the fact that they had 3 explosive TD plays.
What is the profile of the average team they faced?
What might be somewhat surprising to fellow UGA fans is the fact that UGA’s opponents have performed quite badly in their P5 games. On average, despite wins over two ranked teams, the average net YPP of UGA’s five P5 opponents is -0.74, which is the second lowest among the group. While some of that is attributable to Vanderbilt’s -3.29 net YPP, three of UGA’s other four opponents (Auburn, South Carolina, Missouri) have negative Net YPP against P5 teams, with Oregon being the lone P5 opponent with a positive net YPP. Vandy has played three of top 7 teams on a YPP basis in UGA, Alabama, and Mississippi, and we would expect their Net YPP to improve a bit over the remainder of the season.
Who are the most effective teams on a relative Net YPP%?
If we combined relative offense and defense, UGA and Ohio State have excelled at both offense and defense and are separated from the pack.
How have they done it?
On the chart below, the Y-axis charts represents teams offensive ypp gained relative to its opponents average ypp surrendered (higher is better) and the X-axis represents the defensive ypp allowed relative to its opponents ypp gained (right is better). The upper right quadrant represents teams that have outperformed on offense and defense. The upper left quadrant represents teams that have outperformed on offense but underperformed on defense. The bottom right quadrant represents teams that have underperformed on offense but outperformed on defense. The bottom left quadrant is teams that have underperformed on both, which in this universe is Miss State.
What else does it tell us?
· Among the top tier, UGA, Ohio State, Michigan, and Alabama fare about how we would have expected.
· Michigan’s defense is bit more vulnerable than conventionally thought as Iowa and Penn State performances cancel each other out, and they are left with okay performances against Maryland and Indiana.
· Tennessee’s defense isn’t “terrible”, but it leaves the Vols offense little room for slippage if they are going to seriously contend.
· If we had done this chart in prior eras, Bama would be much further out on the X-axis, but Kirby Smart is in Athens, Jeremy Pruitt is on leave, and Pete Golding is in Tuscaloosa, which explain Bama’s decline in defense since 2017.
· Clemson just isn’t very good at anything.
· Oregon has recovered well, particularly given the burden it is carrying from its game against us.
· Illinois defense is well suited to stop terribly bad offenses.
Note: edited 11.29 for typos and grammatical errors.
Edited by MDC-NYC