How Our Projections Are Built
Every number on Opening Line Analytics comes from Monte Carlo simulation over public data: Statcast, the MLB Stats API, and play-by-play records. This page explains each product’s inputs in plain language, and how we grade the results in public.
NRFI probabilities
Each game’s scoreless-first-inning probability comes from simulating both halves of the first inning 10,000 times. Per matchup we blend the batter’s quality against the pitcher’s handedness with the pitcher’s contact quality allowed, weight the top of the lineup most heavily (the leadoff spot matters far more than the sixth), and apply our own empirical first-inning park factors, built from thousands of real first innings rather than full-game park data.
Probabilities refresh throughout the day as official lineups are confirmed. Every graded call is published, win or lose, on the records page and in the daily archive.
Pitcher strikeout projections
For every plate appearance we estimate the strikeout chance from the pitcher’s strikeout rate against that batter’s side, the batter’s strikeout rate against the pitcher’s hand, and the league rate. Simulating 10,000 full starts turns those per-batter rates into a distribution: a median plus the 10th, 25th, 75th, and 90th percentile outcomes.
We publish the distribution rather than a single number because a starter with a median of 6 strikeouts and a wide spread is a very different proposition from one with the same median and a narrow spread.
Home Run Power Score
Every batter on the slate gets a 0 to 100 score from four weighted components: the batter’s power quality (40%), the opposing pitcher’s home run susceptibility (25%), the ballpark’s handedness-split home run factor (20%), and the day’s weather, including wind projected onto each stadium’s center-field bearing (15%).
The score is a daily ranking, not a calibrated probability: it says who is best positioned to homer today, relative to everyone else on the slate.
Hits & Bases
Expected hits and total bases per batter come from the same simulation approach: per-plate-appearance outcome rates rolled up over thousands of simulated games, published as full distributions with probabilities for common lines.
NFL player props
Every NFL game is simulated 10,000 times over play-by-play and usage data. Each player’s passing, rushing, and receiving yards are published as a projected median plus the p25 to p75 range, alongside anytime touchdown probabilities. The board refreshes weekly with a game-day pass for inactives.
How we grade ourselves
Records are graded automatically from official results and published whether they flatter us or not. NRFI keeps two records: one for calls where our probability beat the market’s implied odds, and one for pure directional accuracy. Nothing is retroactively edited, and losing days stay on the site next to winning ones.