Open model ledger.

A public record of Posterior's resolved MLB model entries: model probability, no-vig consensus, market gap, grade, outcome, and model version. The public ledger is the receipt. It is not the real-time feed.

Posterior publishes resolved and voided model entries after settlement. Live entries, pending rows, book-by-book movement, and full signal attribution remain inside the subscriber almanac and API.

Public entries

no data yet

Resolved

0 voided

Live slate

Gated

subscriber + API

STRONG win rate

—

Download

picks.csv

Resolved ledger →

0 rows · won, lost, and voided entries only

subscriber + API

Full live feed →

Current slate, pending rows, book movement, and full signal attribution

Optional filter: ?season=2026 · public release delay: 24 hours after settlement

Schema

Simplified audit schema, no nested data, parses in pandas.read_csv.

Column	Notes
predictionId	Stable public ID for the model entry
predictedAt	ISO timestamp when the model entry was filed
gameDate	ISO date (YYYY-MM-DD) the game was scheduled for
season	Year integer
playerName	Player, pitcher, team, or market subject
teamName	Team at publish time when applicable
market	Public market label such as 1+ hits or strikeout over
grade	STRONG \| LEAN \| WATCH \| SKIP
status	won \| lost \| voided
actualHits	Resolved hit count when the market is hit-based
modelProbability	Calibrated model probability, 0-1
consensusProbability	Delayed no-vig market consensus probability, 0-1
marketGap	modelProbability minus consensusProbability
entryDecimal	Decimal odds captured at entry
closingDecimal	Closing decimal odds when available
closingMovement	Resolved closing movement metric when available
modelVersion	Model version label used for the entry
calibrationVersion	Calibration version label used for the entry
releasedAt	Timestamp when this entry became eligible for the public ledger

How to use

A 30-second example

import pandas as pd

df = pd.read_csv(
    "https://posterior.pro/data/picks.csv",
    comment="#",
)

resolved = df[df.status.isin(["won", "lost"])].copy()
resolved["actually_won"] = (resolved.status == "won").astype(int)
resolved["bucket"] = (resolved.modelProbability * 10).astype(int) / 10

calibration = resolved.groupby("bucket").agg(
    n=("actually_won", "count"),
    predicted=("modelProbability", "mean"),
    actual=("actually_won", "mean"),
)
print(calibration)

strong = resolved[resolved.grade == "STRONG"]
print(f"STRONG hit rate: {strong.actually_won.mean():.3f}")
print(f"STRONG sample size: {len(strong)}")

License + citation

Public ledger

The delayed resolved ledger is free to use with attribution to posterior.pro. Do not imply Posterior endorses downstream analysis.

Subscriber and API exports are governed by the Terms of Service. They may not be resold, scraped, mirrored, or republished in bulk without written permission.

BibTeX

@dataset{posterior_public_ledger_2026,
  author = {{Posterior Research Desk}},
  title  = {Posterior MLB Public Model Ledger},
  year   = {2026},
  url    = {https://posterior.pro/data/picks.csv},
  note   = {Delayed resolved ledger}
}