cfbfastR

🏈 cfbfastR 3.0.0 is on CRAN — the biggest release in the package's history, and it lands for Week 1. • ESPN college football layer: 8 → 73 wrappers (players, teams, games, pbp, FPI, QBR, recruiting, futures) a Fox Sports layer (fox_cfb_*) • Rebuilt modular play-by-play EPA/WPA engine — cfbd_pbp_data_v2() and espn_cfb_pbp_v2() run one shared engine w/ penalty resolution, player ids, and output tiers • 48 new full-season data loaders covering every published release dataset: 27 ESPN datasets (2004 ), ratings/talent/recruiting/crosswalks, and 10 stats.ncaa.org datasets — including FCS play-by-play • 15 new CFBD endpoints (playoffs, coaches, ratings, player success), situational params everywhere, betting ATS records, per-game havoc • httr2 HTTP layer, native pipe throughout, session caching for ESPN catalogs Nothing deprecated — your old code keeps working, with three pbp sources to choose from (classic 2014 , ESPN 2004 , NCAA incl. FCS 2013 ). install.packages("cfbfastR") 📖 cfbfastR.sportsdataverse.org Every one of those loaders has a twin in sportsdataverse-py reading the same published files — the R and Python loader surfaces were built to mirror each other. One caveat on the EPA/WPA models if you build play-by-play yourself this weekend — next post. 👇

🏈 A good-news/bad-news note for anyone computing CFB EPA/WPA themselves this weekend. Worth two minutes before today's kickoffs. The good news. On GitHub, cfbfastR's expected points and win probability models now load from the shared cfb_model_artifacts bundle — literally the same artifact sportsdataverse-py scores with. R and Python finally return the same EPA for the same play, and one retrain updates both. That branch also adds: • cp / cpoe — completion probability and completion % over expected • vegas_wp — spread-aware win probability alongside the naive wp • xpass / pass_oe — expected pass rate and pass rate over expected • fourth-down and two-point decision surfaces, and create_qbr() • a fix for the failure that made epa_wpa = TRUE unusable for 2006–2013 The bad news, and it's mine to own. I rushed the CRAN submission and the updated models and parser work didn't make it in. So the CRAN build still computes the previous model generation. If you compile play-by-play yourself right now with the CRAN functions, your EPA/WPA won't match the numbers the loaders and sportsdataverse-py are producing. What to do the next couple weekends: ✅ Use the loaders. load_cfb_pbp() and load_espn_cfb_pbp() read published data that already carries (or will soon in load_cfb_pbp()'s case) the updated models, columns and outputs. This is the short path and it's the one I'd take. ⚠️ Hold off on compiling play-by-play from scratch until the next CRAN release — unless you specifically want the older model. 🛠️ Or take everything now: remotes::install_github("sportsdataverse/cfbfastR") The GitHub build is now numbered 3.0.0.9000, so packageVersion("cfbfastR") tells you which one you're on — until today both reported 3.0.0 while returning different EPA/WPA. All of it goes into the next CRAN submission. Sorry for the split — and enjoy Week 1. 🏈 📖 cfbfastR.sportsdataverse.org… And if you'd rather just watch the numbers than compute them — next post. 👇

Full game stats for Florida State vs Alabama can be found at Game On Paper (@gameonpaper): gameonpaper.com/cfb/game/401…