Ohio High School Football Rankings and Predictions
Frequently Asked Questions
Questions about Rankings
Do all teams start the season with equal ratings?
No. While that with would be ideal, the combination of 700+ teams, a ten-game regular season, and localized scheduling makes that unrealistic. Each team starts with a rating that is based on where they finished the previous season, but pulled back a bit toward their historical average over the last five seasons, and toward the average of teams in their OHSAA division. (This pulling back is called "regression toward the mean" in statistics.)
Are games against non-OHSAA opponents considered?
While they usually do count for OHSAA Harbin points, so they are factored into playoff predictions and probabilities, they are not considered in my rankings. This may change in a future season.
Why did my team's ranking drop after they won?
Underperforming relative to expectations (such as winning a close game when heavily favored) can cause a team's rating to drop, even in victory. Also, a key victory in a prior week could lose some of its luster if that team turns out to be overrated, perhaps causing a domino effect. Finally, a ranking (not rating) drop can happen when an unrelated team jumps over others after an impressive win.
Why is Team A ranked lower than Team B, even though A defeated B?
My rankings are intended to be predictive, so that each team would be expected to beat all teams ranked below them (on a neutral field.) Sometimes upsets happen, and they don't always flip teams' rankings. (After Appalachian State won at Michigan in 2007, would it have been reasonable to predict that the Mountaineers would win a rematch? I doubt it.) My rankings have a different goal than that of a retrodictive ranking system, which seeks to rank the teams in a way that is fairest, based on the results of already-played games. However, even at mid-season, it is not possible to make a ranking where each team is ranked ahead of all teams it has defeated, and behind all teams to which it has lost. "Violations" of game outcomes are inevitable in any rankings. (Jay Coleman's MinV is a good example of a high-quality retrodictive ranking (for major college football.)
Why does my team's projected record include a future loss, even though they are favored in every remaining game?
Projected records are based on the middle-of-the-road (median) scenario, so a team will be projected to win out only if there is at least a 50% chance that they will do so. This issue (and similar ones where the projected record does not match the individual game predictions) are common at mid-season, and can occur even with two weeks remaining.
Why does it say that my team has clinched a playoff berth (or home game), but Joe Eitel's site does not show this?
The short answer is that my projections are sometimes too early to call a team "in" or "out", and that Joe's are sometimes too late to do so, based on the limitations of our methods. My predictions are based on the outcomes of many (typically 10,000-25,000) simulations of the remainder of the regular season, computing the Harbin playoff points each time. If a team earned a playoff berth in every simulation, then I show them as "in" (or 100% chance), even though there may be a highly unlikely scenario in which they would not. Joe's numbers sometimes show a team not having clinched even if it would require two teams that play each other to both lose (or both win) that game. If Joe shows a team "in", then it is a certainty (barring score mistakes or forfeits), and if I show it, then the probability is at least 99.99%, but perhaps not yet guaranteed. However, I am not aware of any situation where I ever called a team "in" and they missed out, or vice-versa.
How the Fantastic 50 get started?
When I was a math teacher and head athletic trainer at Fuquay-Varina (NC) high school in the late 1990s, I became interested in the question of whether the top small schools could compete with the large schools. This led me to build a statewide computer ranking, and then predictions on individual games and later playoff teams and seeds. When I moved to Ohio in 2008, the Buckeye state's combination of heavy interest in high school football and a points-based playoff system made it an easy decision to start my Ohio site.
How can a student get involved in research related to sports statistics?
First, you can learn a ton by reading up on sports analytics research online. (For example, ESPN's Brian Burke wrote an excellent NFL blog for several years.) Next, come up with an interesting research question, and find a data set online to help you explore it, or build a data set yourself (maybe from teams at your school.) A high school math teacher may be able to help you with some tools to analyze data, and an AP Statistics class would be even better. Start small If you get the chance, attending a regional sports analytics conference (such as those in Boston, Pittsburgh, Virginia, and Iowa) will give you a chance to talk with more experienced folks in the field.
How does someone get into sports analytics as a career?
Just as for athletes and coaches, it's hard to make it to a professional team in sports analytics. A college degree in statistics or computer science (with proficiency in R, Python, and SQL) along with a related undergraduate research project, is a starting point. Many of those in the analytics departments or front offices of pro teams have an MBA (or another advanced degree) from a top university.