First 5 Innings methodology
How the First 5 Innings model works
Run line / moneyline / total for the first 5 innings, sized to remove bullpen risk from the equation.
Inputs
- • Starter rolling stats (ERA, FIP, xERA, pitch mix)
- • Lineup quality + handedness mix
- • Park run factor + umpire zone size
- • Weather
Outputs
- • Run-line / total / moneyline probability
- • Lean + matchup grade
- • Public-pick or research-only status
- • When available: the sportsbook line and price the model compared against
Approach
F5 isolates the starter-vs-starter portion of the game, removing bullpen variance. Inputs are starter form, lineup quality, park, weather, and umpire context.
The public version blends several signals into one probability. Park-aware and higher-confidence versions are monitored privately before any changes reach the site.
Training data
Multi-year first-5 outcomes from MLB StatsAPI box scores. Pitchers without sufficient history are handled more cautiously until there is enough data to trust the matchup.
Calibration
First 5 projections are checked against settled games and adjusted so confidence bands stay realistic.
Because F5 has a smaller sample than NRFI, we treat the record as directional and keep monitoring it before making bigger model changes.
How a pick becomes official
For a First 5 read to count publicly, the model needs a real market line and a saved pregame price so the result can be graded fairly later.
If the price or game context is incomplete, the model may still show the angle as research, but it does not count toward the official record.
Research variants tracked
Several experimental F5 versions run in the background. They are compared against the public model over time before any of their ideas are promoted.
Known biases and limits
- ! Sample size is small relative to NRFI. Treat F5 record as directional signal, not a verdict.
- ! Pitchers with no prior season history receive a more conservative treatment because the model has less to compare against.
Not for use as
- × Predicting one exact at-bat or one exact inning. These are probability models, not play-by-play fortune tellers.
- × Ignoring the sportsbook price. A model edge only matters if the available price is good enough.
- × Replacing late-breaking baseball judgment. Lineups, weather, scratches, and pitching changes can move after the page refreshes.