Jump to content
  • Available Subscriptions

  • CBCR2 Preseason Power Rankings

    By Nathan Lawrence
    Published in 

     0

    Hi Y’all. This is Nathan, co-host of Chapel Bell Curve, UGA fan, and stats enjoyer. As part of our work at CBC, we’ve been lucky enough to work with Ross Rutledge, proprietor of r2sportsmetrics.com. In partnership with Ross (read: he did the hard work), we release our power ranking metric, CBCR2, each week. Throughout the year, I’ll be hopping on DC to give y’all updates on what our numbers are telling us. 

     Before I get into our preseason rankings, let’s define a few things. First, what is CBCR2? Second, how do we come up with these numbers? And third, what is this metric useful for? Our metric, as many in the industry do, rates each team based on points above average. Each CBCR2 entry will have a net number, a number for offense, and a number for defense. These are our estimations for what the team would score against a theoretical average team, how many points they would give up, and a net rating (the difference between the two) to represent the total quality of a team as Points Scored - Points Surrendered. This gives us a number that we can use to say how many points any given squad is better than the “average” team. (And by the way, we derive the average team from a composite of the previous few seasons) For convenience and smack talk purposes,  we also provide the national ranks for offense and defense in each entry. 

    So how do we come up with these rankings? In short, we let a computer decide. In long, or longer, we select hundreds of stat categories that matter to winning a football game. Then, we use machine learning to weight those stats by how much they impact a teams chance of winning. Finally, we derive our ranking using a linear model based on the weights we get, and apply that model to a team's returning production, talent level, and in-game results. 

    We output two models: preseason – which is focused on talent and returning production –  and in-season, which is derived entirely from play-by-play data for each team. As we enter each season, our preseason model is slowly phased out as we have more in-game results. In terms of what our in-season  model values, we generally care about three broad categories: 

    • Efficiency: How well can you move the ball down the field on a down-to-down basis? How well does your defense do in preventing sustained drives? This category includes things like opportunity rate, success rate, and EPA. 
    • Finishing Drives: When you have the ball inside your opponents 40, how likely are you to score?  How likely is your defense to surrender points in that same situation? We define “opportunities” as trips inside of the 40, because they are the most likely to result in points. The key metric here is points per opportunity. 
    • Explosive Plays: Does your defense give up homeruns? Does your offense score big plays? The key factor here is EPA (expected points added). 

    These categories are similar to Bill Connely’s five factors, but leave out turnovers and field position; those categories are generally represented in finishing drives and efficiency. If you’d like to learn more about the specifics of what we do, you can find an in-depth description of this process in Ross’ excellent primer here. 

    So what can you use these numbers for? Well, first, they are great for smack talk. It may not be literally true that UGA is, say, 5 points better than Bama in 2024. But it sure is fun to say that we are. Second, more practically, they give us an answer to the age-old question: “On a neutral field, which one of these teams is better?” Generally speaking, I use our numbers to give me a rough estimation of team quality. In the era of the 12 team playoff, I use CBCR2 to give myself tiers of teams. Who is fighting for a top-4 seed? Who has a shot at hosting a home playoff game? Who is fighting for the 12th spot? It’s not exact, but CBCR2 gives us more informed answers to these questions. Week to week, you can use these numbers to give yourself an approximate idea of each matchup. While it may not give you specifics on how the game will play out, it will give you a rough idea of the quality of each team involved in a given game. I use it to set my expectations not just for UGA’s games, but for games of teams whose rosters I’m not familiar with. 

    It’s important to note, however, that CBCR2 is not intended to give an accurate prediction of any given matchup. These aren’t numbers designed to help you bet, win your pick-em pool, or generally predict the specific score of a game. While we could use the numbers to give an implied predicted score, that’s not what they’re designed to do. So don’t bet the house on these, is what I’m saying. 

    With that warning given, let’s look at our preseason top 25. (You can find the full list at r2sportsmetrics here.)

    IMG_7033.thumb.jpeg.ab24cd7c9ef9624c3d24d2675ec82d9b.jpeg

    A couple of surprises here based on the eye test. First of all, CBCR2 loves Oregon. If you can get past the corpse-of-the-Pac-12 bias – a difficult task for some, it makes sense. Oregon’s complete dominance of the transfer portal means that this is a title-caliber roster. Two of the main components of our preseason model are team talent and returning production, and the Ducks have that in spades. 

    The second thing that stands out to me is Ohio State. Thought by many to be the best bet for a title game appearance, our numbers have them at #5, 6 full points between co-leaders Georgia and Oregon. I’m not sure I agree, but I think the Buckeyes aren’t necessarily getting the credit for their splashy defensive transfers from the numbers that the public and press are generally giving them. However, based on the digging we’ve done in the model, CBCR2 doesn’t think that the massive exodus of talent from the 2023 Buckeyes’ offense has been adequately replaced during the offseason. While CBCR2 doesn’t have eyes, it gives Will Howard a funny look based on his performance at Kansas State. You can take the boy out of Manhattan (Kansas), but you can’t take the Manhattan (Kansas) out of the boy, I guess.  

    As for the tiers of teams, our numbers see UGA and Oregon as a clear cut above the rest. There is a 5 point drop to our 3rd place team, Bama, who is grouped in tier 2 (or really, 1a) with Texas, and the aforementioned Buckeyes.  Fittingly enough for the first year of the new playoff, our next multi-point gap comes between #12 LSU and #13 Oklahoma. Inside the top 12, the big surprise to me is Penn State. This is probably just my bias, as I think James Franklin is a walking self-help book with a pair of bad sunglasses attached. Having said that, their inclusion does illustrate the pressure that Franklin and the Nittany Lions face this year. I think playoff or bust is a reasonable expectation for fans in State College. 

    Ultimately, preseason metrics are an exercise in rangefinding. It’s a near certainty that there will be a team in these rankings that makes me look foolish. That’s fine. The point of the exercise isn’t perfect accuracy, but to set our expectations for the coming season. If anything, a team over- or underperforming from this benchmark is a really compelling moment to know advanced stats. A lot of the time, what we consider to be “underdogs” are really just teams whose public perception hasn’t caught up with their performance. When we see a team truly outperform their numbers, we’re able to identify which squads are truly surprising. If there’s anything that makes college football compelling, it’s the surprises. 

    • Like 2

    User Feedback

    Recommended Comments

    There are no comments to display.


×
×
  • Create New...