About
This website is supposed to be for the College Football Nerds!
I think there is more of us than we would think. When I started thinking about what I wanted this site to be, I really just wanted a data aggregation layer. When I would gamble on games, I spent so much time obtaining disparate information across the web. I thought, someone has to have an overlay for when Alabama travels to Baton Rouge right? No. So I started building, and I wanted it to resemble the best data aggregator (besides Anduril ofc!), Palantir.
CFB Palantir is a data aggregation layer for college football betting. Built and run by a single operator. Please reach out if you want to contribute!
More from the author: joebarresi.dev.
Methodology and credits
All data comes from CollegeFootballData.com. The advanced stats on game pages are computed here from its play-by-play, drives and box scores, once a day after games finish. The definitions are other people's work, credited below. We use CFBD's expected-points model rather than building our own, so our numbers are close to, but won't exactly match, sites that use a different model.
How the numbers work
- As of kickoff. A game page shows each team's numbers through the week before the game, so a finished game never counts its own result. A team with no games yet this season shows last season's final numbers, labeled as such.
- FBS games only. Every team stat and rank counts FBS vs. FBS games. Ranks are among all FBS teams that have played, with ties shown as “T-#12”. Green is good for the unit it describes: a defense ranked #1 is the best defense.
- EPA. Expected points added per play, from CFBD's PPA. Counts scrimmage plays only: runs, passes, sacks and scrambles, not kneels, spikes or special teams.
- Success rate. A play is a success if it gains 50% of the yards needed on 1st down, 70% on 2nd, or 100% on 3rd and 4th. Bill Connelly's definition.
- Explosive plays. A pass worth at least 2.4 EPA or a run worth at least 1.8, the thresholds GameOnPaper uses.
- Havoc. The share of plays with a tackle for loss, sack, interception, forced fumble or pass breakup, as defined in sportsdataverse-py.
- Available yards %. The share of the yards in front of each drive that the offense actually gained. From bcftoys.com, by way of GameOnPaper.
- Adjusted EPA. EPA per play corrected for the opponents faced, from one ridge regression over the season's plays, refit each week on games played so far. Garbage time is left out using Connelly's score margins. A port of the method in sportsdataverse-py (MIT License, © 2021 Saiem Gilani).
- Strength of schedule. The average adjusted strength of FBS opponents already played, and of those still to play. Built on the adjusted EPA above.
- Negative plays. Sacks, interceptions, fumbles lost, turnovers and penalties per game, from CFBD box scores. “Committed” is what a team gave up; “forced” is what it took from opponents.
Credits
- GameOnPaper, by Akshay Easwaran and Saiem Gilani: the spider chart and matchup table forms, and many of the definitions above.
- Bill Connelly: success rate, garbage-time thresholds, and the radar chart form.
- sportsdataverse-py (MIT License): adjusted EPA and the havoc definition.
- bcftoys.com: available yards %.
- CollegeFootballData.com: every piece of data on this site.