Okay, so this week’s all about the CFP rankings, right? Thought I’d share how I tackled it. It’s messy, it’s probably not the “right” way, but it’s my way.
First, I grabbed all the team data I could find. I’m talkin’ wins, losses, strength of schedule – the whole shebang. Scraped it from different sports sites, some were easy, some were a total pain. Spent a good chunk of time just cleaning up the data, getting it into a format I could actually use.
Then, I started messing around with different ranking formulas. Tried a simple win-loss record thing, which was garbage. Then I tried incorporating strength of schedule, which helped, but still didn’t feel right. I figured that beating a team that’s also good should be weighted more heavily, you know?
So, I dug into some statistical analysis stuff I vaguely remembered from college. Played around with different weighting factors for wins against ranked opponents. This involved a lot of trial and error. Like, a lot. I was just throwing numbers at the wall to see what stuck. I used a spreadsheet at first, then switched to a Python script because I was getting tired of manually recalculating everything.
The biggest headache was figuring out how to deal with early-season games. Like, a win in week one against a team that later turns out to be terrible shouldn’t have the same weight as a win in week twelve against a top-ten team. I ended up using a decaying average for strength of schedule. Basically, recent games mattered more.
After that, I started comparing my rankings to the “official” CFP rankings. And, uh, let’s just say they weren’t exactly the same. Which wasn’t surprising. But it gave me a baseline to see where my formula was going wrong. I tweaked the weighting factors some more, ran the script again, and compared the results. Rinse and repeat.
It wasn’t just about getting the same top four teams. I also looked at the overall distribution. Was my formula consistently overrating or underrating certain conferences? Were there any glaring outliers? This helped me to identify biases in my approach.
Honestly, it was a frustrating process at times. But I learned a lot about data manipulation, statistical analysis, and the inherent subjectivity of rankings. And hey, even if my rankings aren’t perfect, at least I can say I tried. Plus, it gives me something to yell about with my buddies when the actual CFP rankings come out.
I’ll probably keep tweaking the formula over the next few weeks. It’s a fun little side project, and a good way to keep my coding skills sharp. Maybe next year I’ll even try to build a whole web app for it. But for now, I’m just happy to have a semi-functional ranking system that I built myself.
