Coming soon — 2026 season

OnBase plays every MLB game out before it happens, pitch by pitch, in the real weather and ballpark of each field. Then it shows where its number and the book's disagree. The scorecard's public, wins and losses both.

One email at launch. No spam, no daily picks.

50,000+
Simulations per game
2016–now
Historical pitch & market data
5 books
De-vigged per market
100%
Calls graded publicly

How the engine works. Not a promise about results.

What's inside

Built like a trading desk

Fair probabilities, the reasoning behind them, and a public ledger that keeps the count honest.

The Edge Board

Every market on the slate, our fair probability next to the book's price. Sorted by edge. The number is right there; the reasoning is one click in.

  • Fair price vs. book
  • Closing-line value
  • Confidence on each call
MarketModelBookEdge
A. Judge TB Over 1.558.5%+121+7.5
J. Soto Hits Over 0.568.9%-132+5.9
S. Ohtani Runs Over 0.557.1%+104+5.0

Calibrated, and checkable

When OnBase says 62%, it lands near 62% over time, and we post the math. Brier score, calibration error, and CLV on every graded call, updated as results come in.

  • Public accuracy ledger
  • Losses stay up
  • Graded on CLV
Predicted → realizedECE 2.2%
55%
54%
60%
61%
65%
64%
70%
71%

Ballparks, modeled for real

Ball flight runs through tonight's air, with temperature, humidity, altitude, and wind resolved at each field. Coors in dry air behaves differently, so we price it differently.

  • Per-batter park factors
  • Wind by field
  • Physics, not flat multipliers

Get on the board early

People on the early list get in before the public launch.

One email at launch. No spam, no daily picks.