Pro-level metrics mapped to college equivalents. What each stat measures, what data exists, and where the gaps are.
Percentage of batted balls meeting specific exit velocity (≥98 mph) and launch angle (26–30° range, expanding with higher EV) thresholds. A barreled ball has a minimum .500 batting average and 1.500 slugging percentage historically.
MLB Statcast (Baseball Savant)
Not available in college. TrackMan data at select venues captures exit velocity but isn't aggregated publicly.
MLB barrel rate is available per batter and pitcher. College proxy: ISO (isolated power) and HR/FB ratio serve as rough power indicators.
Barrel rate requires pitch-level tracking infrastructure. College venues that lack TrackMan or Hawk-Eye cannot generate this metric.
Completion Percentage Over Expected. Measures how often a quarterback completes passes above or below what the model expects given the difficulty of each throw (distance, coverage, pressure).
NFL Next Gen Stats (AWS)
PFF provides college CPOE via their grading system. Limited public availability — requires PFF subscription.
NFL: available via Next Gen Stats. CFB: PFF grades include adjusted completion metrics but raw CPOE is behind a paywall.
College CPOE models use different tracking data (All-22 film grading vs. NFL's player-tracking chips). Direct NFL-to-CFB CPOE comparisons are not valid.
Expected Points Added. Measures how much each play increases or decreases a team's expected points, based on down, distance, yard line, and score. A 3rd-and-1 conversion from the 50 adds more expected points than a 1st-and-10 run for 3 yards.
NFL play-by-play data (nflfastR, ESPN)
Available via cfbfastR and collegefootballdata.com. Same methodology applied to college play-by-play data.
Both NFL and CFB EPA are publicly available. BSI uses EPA/play for team and player evaluation in football content.
EPA doesn't account for personnel, scheme, or defensive alignment. Garbage-time plays inflate offensive EPA and deflate defensive EPA.
The average number of points a team would score from a given game state (down, distance, yard line) based on historical play-by-play data. First-and-10 from the opponent's 20 has higher expected points than first-and-10 from your own 20.
NFL play-by-play models
Same methodology, different calibration data. College expected points tables are built from college play-by-play, not NFL data.
Publicly available via nflfastR (NFL) and cfbfastR (CFB).
Expected points assumes league-average teams. A first-and-goal from the 1 for Alabama is worth more expected points than the same situation for a bottom-tier program, but the model treats them equally.
A statistical model that evaluates pitch quality based on velocity, movement (horizontal/vertical break), release point, spin rate, and location. Often expressed as Stuff+, Location+, or Pitching+.
MLB Statcast, various model providers (FanGraphs Stuff+)
Limited. Programs with TrackMan have internal pitch models, but no public standard exists. Rapsodo data is sometimes shared by individual pitchers.
MLB pitch models are fully public. College: BSI tracks K/9, BB/9, HR/9, and WHIP as proxy metrics for pitch quality.
Without movement data, college pitch models rely on outcome-based metrics rather than stuff-based ones. This conflates pitcher ability with defensive support and park effects.
A composite metric (typically from PFF) that grades quarterback passing performance on a 0-100 scale. Incorporates accuracy, decision-making, pocket presence, and throw difficulty.
PFF (Pro Football Focus)
PFF grades college quarterbacks on the same scale. Available behind PFF's paywall.
PFF grades are available for NFL and CFB. BSI uses publicly available stats (completion %, yards/attempt, TD/INT ratio) as proxies when PFF data is not accessible.
PFF grades involve subjective film evaluation. Two graders can disagree on the same play. The grades are useful directionally but shouldn't be treated as objective measurements.
The x, y, z coordinates where a pitcher releases the ball. Consistency of release point correlates with deception and command. Measured by Hawk-Eye or TrackMan cameras.
MLB Statcast (Baseball Savant)
Available only at venues with TrackMan/Rapsodo. Not publicly aggregated for college pitchers.
MLB: full release point data per pitch. College: not available through public sources.
Scouting reports from college coaches reference release point qualitatively, but quantitative data stays internal to programs.
Revolutions per minute (RPM) of a pitched ball. Higher spin on fastballs creates "rise" effect; higher spin on breaking balls creates sharper break. Context-dependent — spin axis matters as much as raw RPM.
MLB Statcast (Baseball Savant)
Available at TrackMan-equipped college venues but not publicly aggregated. Some programs share spin data in recruiting materials.
MLB: full spin data per pitch. College: sporadic. BSI does not currently ingest college spin data.
Spin rate without spin axis and movement data is misleading. A 2,500 RPM fastball with bad spin axis performs worse than a 2,200 RPM ball with efficient spin.
A model that estimates each team's probability of winning at any point during a game, given the current game state (score, time, possession, field position, etc.).
ESPN, FiveThirtyEight (historical), various implementations
Same concept applies across sports. College-specific models must account for wider talent gaps and higher variance than professional leagues.
BSI is building a win probability model documented in the Models hub. See the methodology page for inputs, assumptions, and validation status.
Win probability models assume teams play at their average level for the remainder of the game. They don't account for in-game adjustments, coaching decisions, or momentum.
Expected Weighted On-Base Average. Uses exit velocity and launch angle to estimate the expected value of a batted ball, removing defense and luck from the outcome.
MLB Statcast (Baseball Savant)
No direct NCAA equivalent — college baseball lacks Statcast tracking at most venues. TrackMan data exists at some programs but is not publicly available.
MLB xwOBA is fully available via Statcast. College proxy: use wOBA from traditional stats (hits, walks, HBP weighted by run value).
Without exit velocity data, college xwOBA cannot be calculated. BSI uses traditional wOBA as the best available approximation.