Football Expected Points Model. FSU Collapse. Clustered QBs. NCAAF Projections. xG Bayesian Modeling.
Issue Two Hundred
Welcome
Football is in the air :)
Issue Summary
The stunning collapse of the 2024 Florida State Seminoles highlights a new age of analytics, with an updated Expected Points model and a new Bayesian solution for xG that quantifies player skill. New tools are also at the forefront of projections, from a custom win total tool for top teams to a method for clustering college football's starting quarterbacks by performance.
Unexpected Quote of the Week
“Not so fast, my friend!”
Lee Corso
This Week's Lineup
A New Expected Points Model
Sports Info Solutions updated its Expected Points (EP) model for football, adding new variables—including time remaining in the half/game and the score differential in the fourth quarter—to more accurately account for changes in pace of play. The new model, which applies to both the NFL and college football, is better at predicting scoring outcomes in high-leverage situations, such as when a team is losing at the end of a game, an area where the previous model was notably deficient. sportsinfosolutions.com
A Numbers Nightmare: The Stunning Fall Of The 2024 Florida State Seminoles
The Florida State Seminoles football team, projected to win 9.5 games and ranked as the 13th most talented team in the FBS, had a historically catastrophic 2024 season, finishing with a shocking 2-10 record due to a dramatic drop-off in both their offense and defense. This unprecedented collapse, characterized by a lack of explosive passing plays and a porous offensive line, made FSU the biggest underachiever of the season based on preseason win totals and team talent rankings. via
at cfbnumbers.substack.comWhy Your xG Model Might Be Wrong: The Bayesian Solution to Accurate Scoring Predictions
By employing Bayesian hierarchical modeling, a new study demonstrates that while a player's position has minimal impact on their expected goals (xG) when controlling for shot features, individual player skill significantly affects goal-scoring probability. This approach allows analysts to quantify a player's finishing ability, providing a more accurate measure of performance than traditional xG models. via
at thexgfootballclub.substack.comClustering the Power Four: An analytical look at college football’s 2025 starting quarterbacks
Using a K-Means clustering algorithm, PFF has categorized the projected starting quarterbacks across the Power Four conferences into tiers based on performance metrics such as passing and rushing grades, accurate throw percentage, and Wins Above Average (WAA), with the largest group being unproven players who could be breakout candidates. pff.com
Using Our New Win Total Tool On 8 Projected Top Teams
In a new article,
uses a custom "win total tool" to project the 2025 season for eight of college football's top teams. Instead of a binary win/loss prediction, the tool assigns a percentage chance of winning each game to create a more nuanced win total. This method resulted in the author being higher on Clemson and lower on Texas, Georgia, and Penn State compared to current sportsbook odds. https://cfbnumbers.substack.comSelf-Promotion
I am building +EV Bets, which identifies positive expected value (+EV) in the major sports betting markets based on popular prediction models and live betting odds.
I just released a big update that vastly improves the user experience after logging in. Here’s a sneak peek:
New to +EV betting? This explainer is for you.
This RSS feed is published on https://unexpectedpointsadded.com/. You can also subscribe via email.