Beta Free full access during beta. All models, tools, and daily picks included.

Home Runs methodology

How the Home Runs model works

Per-game HR probability for each batter, surfaced as a research model. Rare-event modeling; public W-L is intentionally not the trust framing.

Research model · Status: Research model. Calibration active. Experimental challengers are tracked in the background. · Retrain: Weekly (Sunday overnight phase). Shadow variants graded daily. · Last reviewed: 2026-05-05

Inputs

  • Batter Statcast EV / LA / barrel rate / hard-hit rate
  • Handedness vs pitcher arsenal
  • Park HR factor
  • Weather (wind out / in / cross)
  • Lineup slot

Outputs

  • HR probability per batter per game
  • Confidence
  • Research-board status
  • When available: the sportsbook line and price the model compared against

Approach

Per-batter HR projection from Statcast contact quality (EV / LA / barrel rate), park HR factor, weather, opposing pitcher pitch mix, and matchup history.

Because HRs are rare, the trust framing is whether the probabilities stay realistic over many games, not whether any one long-shot swing happens tonight.

Training data

Multi-year batter PA-level data with Statcast batted-ball events. Handedness splits + pitch-type whiff rates are core inputs.

Calibration

Home run probabilities are checked against a large set of settled batter games, with special care because home runs are rare events.

Experimental versions run quietly in the background and need to beat the current model over time before they replace it.

How a pick becomes official

Home run reads stay research-first unless the matchup, lineup context, and sportsbook price are complete enough to evaluate cleanly.

Because home runs are rare, we are especially cautious about turning a high-probability angle into a public pick.

Known biases and limits

  • ! HR is intentionally excluded from the public picks page until its volume / trust framing is respectable enough.
  • ! Rare-event modeling means small probability differences can matter, but they are easy to overread without price and sample-size context.

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.