NCAA Football Rankings and Predictions
We’re going to take what all the experts and computer models are saying about college football and make sense of it.
We’ll break everything down.
- What college football experts have been right the most?
- What are the best computer models for ranking teams?
- What do the best expert models and predictions have in common?
- What do they disagree on?
- Where are the best places for you to take action?
Before we start, it might be worth noting that these are rankings and predictions. What’s the difference?
NCAA Football Rankings uses past data like wins, losses, strength of schedule, and margin of victory to order teams from best to worst (think a top 25 rating)
NCAA Football Predictions guesses the future outcomes of games, instead of evaluating the past (like rankings did). Often times, this starts by using NCAA Football Rankingsto compare two teams, but then uses one or many other factors. The most common factor is home field advantage, but some can be more complicated and include injuries, how far a team has to travel, how many rest days players get, etc.
I’m not the expert, and I don’t design computer models
I want to be straightforward here.
I’m learning here just like you. Week to week.
People much smarter than me are analyzing film, have inside boosters and coaches giving them intel, and are developing computer models.
I’m just taking what they’re giving, and finding an edge in all the information.
It’s a learning process. Let’s start by breaking down some of the models that rank NCAA football.
A few ranking and prediction models we’re currently looking at
Let’s start out with a brief report on a few of our favorite NCAA ranking predictions.
Football Power Index
The Football Power Index is developed by ESPN. It feels generic putting a big brand as a leading predictor, but they also have big money to put behind it. The Football Power Index takes a ton of data like past season results, players lost and gained, coach, and recent games stats (total offensive yardage, defensive points allowed, red zone efficiency, and more) and crunches it using statistical software. There’s more that goes into this one than most, but after all, it’s ESPN. But does more = better?
What is the FPI number? FPI is the college football team’s predicted margin of victory (positive FPI’s) or loss (negative FPI’s) against an average opponent on a neutral field.
How does FPI predict games? The FPI number is the base number used to calculate the prediction margin of victory for individual games. It’s combined with other information specific to that game: the opponent, home or away, days of rest, and more.
Massey Rating Football Ratings
Massey is an OG in sports ratings and predictions. He started in 1995 while studying mathematics in college. In 1999, he was called to use his rankings are part of the BCS (NCAA Football’s ranking system from 1998 to 2013). The BCS used Massey’s and other polls and rankings to come up with their final college football team rank. But Massey’s predictions have stood out on their own with a lot of success compared to other ranking systems throughout the years. Let’s hope he never gives up – definitely some valuable data here.
How do Massey Football Rankings Work? Massey rankings primarily take into account wins and loss and margin of victory, but the margin of victory has diminishing returns–in other words, you get a bigger “jump” in ranking by winning by 10 points instead of 3 (7 points more) than you would by winning by 17 points instead of 10 (still 7 points more). This is simpler (fewer variables like motivation, weather, etc) than some systems (like ESPN’s), but there is still complicated mathematics underlying it. Strength of schedule is “naturally” included, meaning it is not a separate rating, but because so many teams play each other the mathematics naturally creates a strength of schedule.
Massey Ratings early in the season factor in “preseason information,” and the influence of this information decreases with time. The preseason information is mostly past season results, and doesn’t include things like starts retained or lost. In other words, it really is a rough starting point, and the system may be more valuable later.
Dokter Entropy Ratings
Jon Dokter’s Entropy Ratings have been one of the names you consistently see ranked highly on end-of-the-season accuracy. Jon made rankings from 2005, was highlighted by many major sources, and used to provide paid consulting to create ratings. Today, he still shares his Entropy system.
How do Entropy Ratings work? One major part of the Entropy Ratings is there aren’t a ton of inputs. In other words, it doesn’t take the approach of Football Power Index of including things like motivation, red zone efficiency, etc. The key components are wins and margin of victory. But another critical part is margin of victory becomes less important for bigger blowouts.
Entropy ratings use preseason information–mostly how a team did last year–to start the year off, but it’s value decreases as more data from the current season becomes available. Entropy is an older ranking system, but it’s stood the test of time.
NCAA Football Predictions
We’ve described NCAA football ranking predictions. In other words, what computer models are useful in determining the rank of NCAA football teams.
Naturally, this is going to help you predict who will win a game. But it’s not that easy. If Alabama is ranked #5 and Ohio State is ranked #2, does that mean throw your tax refund on Ohio State?
I wouldn’t. At least without more information.
Game predictions take another layer of mathematical modeling. You need to include other elements, and just like the computer ranking models, takes more information.
The most obvious piece of information in the game prediction model would be home-field advantage. That’s worth a lot.
Some models may only take home-field into account. Another model may take 20 other factors into account. What could these be?
- key player injuries
- motivation: how much does the team need to win?
- style of play
- historic coaching head-to-head
- and many more
You can see how this would be fun to tinker with to create your own models.
Throughout the football season, we’re going to seek out some of the best experts to see who they’re picking to win the big games. Just like the computer models, early in the season, we’ll use some of the models that were most successful in the past. As the season goes on, we’ll start to include those that are killing the game predictions this year.