Drew Pasteur's Ohio Fantastic 50

Ohio High School Football Rankings and Predictions

Follow the Fantastic 50 on Twitter for the latest updates.

Frequently asked questions

Updated 16-Aug-2023

How accurate are your game predictions?
In the first week of the season, I get "only" around 70% of the winners correct and my average miss on the score margin is about 17 points. By week #5, the percentage of correct picks is typically around 83-85% and the average margin miss is about 12 points. Those numbers stay about the same until the regional semifinals, when the games get very competitive, making it tougher to pick winners. (By comparison, good NFL systems pick about 65% of winners and miss by 11 points on average, but the NFL has fewer mismatches.)

How long does it take each year until the numbers become meaningful?
In the preseason, we can get a decent idea of what it will take for each team to achieve certain outcomes (e.g. likely to get a playoff berth if 6-4, borderline for a home game if 8-2, etc.) After about three or four weeks, the rankings and predictions are more reliable and we start to get a decent picture of the range of possibilities for the playoffs.

How are the starting rankings determined each year? Do you account for returning personnel, coaching changes, etc.
With 700+ teams in the state, I don't have the time or resources to keep up with roster or coaching changes at every school. CalPreps does some of this, but my starting ratings are based 75% on the prior season's final ratings and 25% on the season before that.

How are games involving out-of-state opponents treated?
Those games are not counted at all in my ratings/rankings. To include those games would require estimating the quality of those non-OHSAA teams, which then requires assessing their opponents, and so on. To do this well turns into a national ranking system like the one done by CalPreps. Because I do need to factor those games into Harbin point playoff predictions, I use CalPreps' ratings (converted into my scale) to estimate the chances of Ohio teams winning future games against out-of-state teams. I do not consider these as real predictions, so they are not included on my picks page or accuracy calculations, but they do show up on team pages.

How do you get "what-if" probabilities on playoff scenarios?
Having access to powerful scientific computing software, I can run many (thousands of) simulations of the remainder of the regular season, calculating the Harbin points and playoff teams/seeds/outcomes each time. Within a given sim, a team's rating may rise or fall based on their outcomes. (A team that pulls off a big upset is likely better than I gave them credit for.) We can then find "conditional" probabilities based on certain situations, such as a team winning all of their remaining games, which leads to a variety of interesting data.

Do teams get extra credit for running up big scores against a weak opponent?
Accounting for margin of victory is important in ranking football teams, but lopsided scores often have more to do with offensive styles & philosophy about rotating in young players than about the difference in team strengths. Any extra margin beyond 42 points is ignored, and there is a diminishing returns principle well before that. Beyond a 3 TD margin, each additional score is worth less than the previous one.

How large is home-field advantage in high school football, and do you account for teams playing at community stadiums or other off-campus locations?
What I have found to best fit the data in Ohio HSFB is crediting the home team with 1.5 points, which is a smaller number than typically assumed in collegiate or professional football. Research on pro sports shows that much of the observed home advantage comes from favorable officiating (disproprtionate numbers of certain types of calls), rather than travel or crowd noise, but extreme climates (Denver's altitude, winter in Green Bay, etc.) also matter. Regarding shared stadiums, I do not have detailed information to assess whether a particular stadium is a true home venue for the host team.

Why do prominent teams seem to "cover the spread" less than half the time?
In making score margins predictions, I am trying to balance two conflicting goals: to minimize the average error (point differential between predicted and actual scores) and to set a number that the favorite team has a 50/50 chance of outperforming. I lean more toward the first of these, without completely disregarding the second. In D1-D3 (large school) games, the favorite only "covers" about 45-47% of the time. That number rises to about 50% in games matching D4/D5 schools, and 53-55% in D6/D7 games. One other notable trend is that when a large school is an underdog against a team from 2+ divisions lower, they cover almost 60% of the time. This suggests that my algorithm may slightly inflate the ratings of top smaller schools.

How does team quality vary by division (school size), and how is this information used in computing the rankings?
Division level is not considered at all. Teams are ranked solely on their performance. The biggest gaps are between D1 and D2, and between D6 and D7, because very large or very small schools are fundamentally different in their ability to have quality depth. In recent years, average team ratings have approximately as follows: D1 140, D2 117, D3 111, D4 103, D5 96, D6 86, D7 71.

How long have you been doing Ohio rankings?
I moved to Ohio in summer 2008 and almost immediately behan publishing rankings and predictions. Much of that info was not archived, so today's history pages are based on a 2022 redo of 15 years of games using my current method. There likely are some minor differences from what was published at the time.

How did you get into high school football predictions?
After college, my first real job was as a teacher and head athletic trainer at a high school in my native NC. In the late 1990s, computer rankings became a part of selecting college football teams for the Bowl Championship Series, including the two-team playoff. I started playing around with computer ranking ideas, and the idea caught on in the NC HSFB community.

What is the math behind how the predictions work?
Over 20 years ago, I started out by using an Elo ranking. Eventually, I built my own extension of the Colley Matrix method, to account for margin of victory, game recency, carry-over between seasons, etc. My version was published in a 2010 book and you can find a copy here. Changes since then are relatively minor.

What is your educational background?
After graduating from a public high school in Raleigh, NC, I attended the University of Florida. Over five years, I earned an undergraduate degree in math, and a master's in math education, and took coursework in athletic training. After teaching high school for a few years, I went back to graduate school full-time at NC State, earning a PhD in applied mathematics in 2008. My specialization was in using math as a tool to study issues in biology and medicine. (Sports analytics barely existed as an academic field at that time.)

What kind of tools do you use to run the numbers?
If I were starting from scratch today, I would write the code in Python. However, I have used MATLAB/Octave for a long time, and have many thousands of lines of code, so I continue to adapt that. When everything runs properly, it can download scores, run rankings and simulations, build the web pages, and upload them to the site, as Joe Eitel posts scores overnight each Friday and Saturday.

What advice do you have for students who want to get into sports analytics?
Learn to use common tools such as R, Python, Git, and SQL, either through classes or by free online resources. Take some statistics and data science classes. Try some projects own your own, and write about what you find. Join a sports analytics club at a college or university. Recognize that it's hard to make a career out of this (though I have a couple of former students who have done so), but the same skills carry over to good jobs in business analytics, scientific research, etc.

Do you make a living off of running this website?
No, I am a professor at The College of Wooster, where I teach math, advise student research projects, and mentor student-athletes. The F50 project costs me a small amount of money annually for web hosting, etc., but I have chosen to keep the site available to all and ad-free.

Do you gamble on sporting events?
No, even if I didn't have moral/ethical concerns about sports betting, I know that the bookmakers are very, very good at what they do. In an academic research project on predicting NFL games, that was published in a top sports analytics journal, we found that we weren't good enough to consistently beat the oddsmakers. Very few people have the knowledge and skill to do so over the long term.

Where do I send corrections, questions, or ideas?
All of my score/schedule data comes from Joe Eitel, so if you email him, then I will also eventually get the info. Comments/questions regarding my work can be sent here or to my Twitter account @OhioF50.