Loren Maxwell, sports rankings expert
Today’s interviewee is Loren Maxwell, the founder of the Georgia High School Football Historians Association and the Maxwell Ratings published at AJC Varsity and frequently cited in GHSF Daily. In this interview, Maxwell gives his thoughts on the Georgia High School Association’s Post Season Rankings model that helps seed and select playoff teams in classes 3A, 2A, A Division I and 3A-A Private. Starting in 2026-27, the GHSA will use the PSR for all classifications and no longer use region finish to guarantee playoff berths or help seed except for region champions getting top-16 seeds. Maxwell believes a math model is a better way to pick and seed teams, but he considers the PSR to be an inadequate model.
1. What do you think of using computer rankings to select and seed playoff teams as opposed to the traditional GHSA way of using region finish, taking the top four from each region? “I applaud the GHSA for wanting to take a step toward greater fairness. Every year, deserving teams are left out of the playoffs simply because of the region they play in, and mathematically half of all classifications every year will have the top two teams seeded on the same side of the bracket, eliminating the chance they’ll meet in the championship game. It'll also have a positive effect in the regular season since now there will be an incentive for scheduling and defeating strong non-region opponents in games that are ignored by the region standings.
“But while moving to a mathematical model can help alleviate those problems, the design of the model is key. It must be transparent, equitable and rigorous. If those criteria are met, a mathematical model will easily outperform region standings in fairly selecting and seeding playoff teams. But if those criteria are not met, it will simply invite a new set of problems, and the outcome may be even worse.”
2. Is the GHSA's ranking model adequate? If not, what are the shortcomings of the PSR or any RPI-type method? “The specific concern with the PSR, like most RPI-type models, is that it lacks rigor, which is one of the criteria the mathematical model should have. At the end of each regular season, we can usually find several teams ranked higher or lower than others despite a clear, easily traceable chain of victories and defeats that contradict those rankings. Of course, contradictions will occur with any model, but the PSR has only a limited view of those relationships. A rigorous model, by contrast, considers the entire network of games and minimizes those inconsistencies to produce a more accurate set of rankings.
“That limited view also opens the RPI-type models to exploitation. Because they rely solely on simple win-loss records and opponent win-loss records, the ratings can be manipulated through scheduling, either avoiding strong opponents or padding with weaker ones. A rigorous model cannot be exploited in that way. A Class 4A team that loses to Buford by one point would be punished under the RPI-type model, but the rigorous model would reward them for performing well against one of the state’s highest-rated teams in spite of the loss.
“However, the larger concern isn’t with PSR’s specific flaws; it's with the illusion of precision these models can create. They only give the appearance of rigor while actually lacking it. Relying on these models to rate teams is only tricking ourselves into thinking it's rigorous because there's a method behind it and a number at the end of it. A truly rigorous model will also have a method behind it and a number at the end, but it will produce results that make sense as well.
3. What would you suggest as an alternative? “I recommend what’s known as the Bradley-Terry (B-T) model. The B-T model is a well-established statistical method used to analyze a set of pairwise comparisons, such as the outcomes of high school football games. The variation I've developed is called the Extended Standings, which I’ve recommended to the GHSA. The Extended Standings extrapolate the on-field results of games actually played to produce a set of hypothetical standings showing what the results might look like in a round-robin schedule. In other words, what would the standings look like if every GHSA team faced every other GHSA team during the regular season?
“This model is equitable, rewarding teams that perform well against stronger opponents, and it even accounts for home-field advantage. While it’s admittedly more complex than the PSR, it’s also more rigorous, and the results are demonstrably better. But regardless of which model is used, transparency is the key to offsetting complexity. The Bradley-Terry model is well documented and has been applied to sports many times, including as one of the early components of the BCS system.
“If the GHSA adopts such a model, it should publish the exact formula and data so that coaches, administrators and fans can replicate the results for themselves.”
4. GHSA coaches have a tough time understanding the PSR/RPI. Is it realistic that they will adopt or accept a model that is much harder to understand? “From what I see, the resistance to the PSR isn't from its complexity, which is fairly low, but from producing counterintuitive rankings with no satisfying explanations. All RPI-type systems will ultimately find resistance here because their results are poor, and that's because rigor was sacrificed for simplicity. A more rigorous model would produce results that would raise fewer questions and provide better explanations, even if someone doesn't know or understand specifically what's under the hood. In other words, the best accepted system is not the one easiest to explain; it's the one hardest to dispute. When coaches review the playoff bracket and it looks correct, I'm confident they'll accept the model that produced it. I think that's what they really want.”
