Latest Rankings: Conference Championship 2021 Rankings
What is the J-Press Ranking System for College Football?
The J-Press system -Jones Probability Rating and Elo Scoring System- is a method devised to rank college football teams, although in theory it could be expanded to other sports.
The College Football Playoff system has been around for several years now, and it does seem to be a fair job pairing off teams that the public wants to see in culminating matches. As a personal project, though, I wanted to develop a rating system for college football that would numerically rank teams, in a similar way to the computers of the BCS era. The computers in the BCS time were never fully consequential, with the human polls and voters always able to get their way via supermajority, and the computers merely provided some false sense of legitimacy to the system.
The goal with the J-Press rankings is merely recreational, although if you choose to use them for other purposes, feel free to do so.
Benefits of the J-Press Rankings
There are a few useful features built-in to the mathematics behind J-Press:
- Strength of Schedule: This is heavily factored into the rankings. Facing many weak opponents to try and increase your ranking will not work very effectively. Teams are rewarded for facing harder opponents. Likewise, you can be effectively penalized by playing many lesser opponents.
- Game Importance: The relative importance of the game is factored in, with high-ranked games counting more and games of little importance counting much less.
- Wins and Losses: The biggest factor determining rankings are wins and losses. A team that wins will gain ranking, while a team that loses will lose ranking.
- Margin of victory: While it is a small factor in the rankings, margin of victory is factored in, but running up a very high score will not help your ranking at all.
- Home field advantage, crowd size, excitement factors, injuries, and other factors are not taken into account at all.
- Only Division I FBS NCAA teams count significantly in the rankings. Playing other teams does very little to affect your ranking.
- Past year’s results: Preseason rankings are determined by the past year’s results, so teams that had a great year last year will start off with a slight advantage.
The mathematical details of the J-Press Ranking system are explained in this post. In short, it uses an Elo ranking system -similar to what Chess uses- that is modified for use with College Football.
You can also see the full Python code for the J-Press Ranking system in this post.
How Should the P-Rankings Be Interpreted?
Each week or preseason when the Top 25 is released, a P-Rating is given to each team. These 25 P-Ratings all add up to 1, so they form a probability distribution -hence the “P” in P-Rating. While this doesn’t translate exactly into reality, a P-Rating should be interpreted as this: If all teams were competing in one large competition for a top prize, what is the probability of that team winning?
Thus, if we consider the “top prize” the National Championship, a team with a P-Rating of 0.15 could be loosely interpreted as having a 15% chance of winning. In practice, though, the P-Ratings are best used to compare teams. Thus, a team with a P-Rating of 0.15 has a much better chance of winning the National Championship than a team with a P-Rating of 0.003.
In general, a P-Rating of 0.10 or higher is considered “high”, a P-Rating of 0.20 or higher is considered “very high”, and any value beyond 0.30 is considered extremely impressive. Remember, all 25 teams add to 1, so a value of 0.30 guarantees being in the top 2 or 3 positions.
How Accurate is the J-Press Ranking System?
A comparative predictive validity analysis was retroactively done using the final top 10 in the J-Press Rankings each year, going back to the beginning of the CFB Playoff system in the 2014-2015 year. An overall correlation of 0.68 was found, which is considered satisfactory. In general, a correlation of 0.8 or higher would’ve been desired, but it is expected and even wanted in the J-Press system to highlight and highly rank some of the lesser-known teams that human rankers rarely do. For example, in the 2021 preseason rankings, Coastal Carolina is ranked #8 in the J-Press rankings but only #12 in the CFB Playoff rankings, despite a stellar season that is arguably better than many higher-ranked teams. Therefore, the 0.68 correlation is deemed to be a good indicator of performance, agreeing with the CFB Playoff results often but disagreeing enough to offer a flavorful commentary.
Are the Preseason Rankings Accurate?
Yes and no. The preseason rankings use the previous years’ results to generate a “first guess” of the upcoming year. You’ll probably notice a few teams you’ll laugh at and say “no way”. This is, after all, a computer program with no human input into the output, so it can’t see all the nuances that we can. Therefore, I’d recommend you take the preseason rankings with a serious grain of salt. As the season begins and wins and losses can start to be factored in, the numbers will look much more reasonable. Wins and losses are the main contributor to J-Press Rankings.
If you are using these rankings for anything serious, I’d recommend that you wait until Week 6 or 8 of the season before putting a ton of weight on the rankings. The entire J-Press system is designed to pick a national champion -the PRating is essentially “the chance that this team should be the national champion”, so it makes sense to wait until about the halfway point in the season before really sinking your teeth into the results. Despite this, I will still post the rankings for weeks 1-7 anyway, for record-keeping purposes. Just know the inherent limitations in early season predictive ability.
Archive of Past J-Press Rankings
As the system progresses, past results will be shown here. This system first went live in the preseason of the 2021 CFB season.