Polymarket baseball data: 98k markets behind the book

Polymarket prices baseball as moneyline, run-line, over/under, and a large player-prop catalog. Every question is a binary market with a live book you can read free from the official API.

Live is easy; history is the constraint. The official /book endpoint has no date parameter, so any study of how baseball prices moved — how the market priced a starter's outing, how the last inning repriced — needs depth that was recorded at the time. The PolyOrderbooks archive captures that depth at 250ms for every baseball series it lists.

The 700,000+ sports catalog splits into over/under lines (372k), props (58k), corners and cards (57k), exact score (56k), and map/set handicaps (34k). Baseball leans on the first three: total runs Over/Under, moneylines, and an enormous props shelf covering pitchers and hitters.

Whatever the contract, the stored record is the same: full bid/ask ladders with sizes, a captured_at timestamp, and the winning outcome once resolved — so a historical book is queryable exactly as it stood at any moment.

• Live ladder — official /book, free and keyless, current moment only (what the API serves).
• Historical depth — full L2 books at 250ms over REST, free Starter with 3 days, paid windows to 120 days.
• Discovery — /v1/series for MLB series; /v1/markets for the question list and slugs.
• Coverage — measured counts per sport on the sports data page.

Take a September MLB night game: the moneyline is liquid all evening, the run line waits until the starter exits, and a ninth-inning walk-off settles the winner in one pitch. The official API serves that live ladder free; the archive keeps the whole ninth available for replay after the park empties.

Baseball's inning structure maps onto the catalog's quarter-style markets: regular-season and postseason games, player props (hits, strikeouts, home runs), and in-game questions that close and settle on their own clock.

Three repeatable questions anchor the record: whether first-inning lines are systematically wider than final-score lines, whether strikeout props drift during the middle innings when a starter labors, and whether the run line reprices inside the ninth for a one-run game.

In practice the /v1/series slug for the MLB regular season narrows 98,000-plus markets to tonight, and the archive query for the ninth-inning window returns the full ladder per 250ms row, ready for replay.

The volatility clusters at inning breaks: the top of the inning and the first batter after an exit repricing are where the ladder actually moves.

Player props have wider spreads than moneylines, which makes them the honest window for spread analysis — granular quotes, not just the flat favorite.

Free Starter covers 3 days of 250ms; a weekend series study fits inside it, and Pro extends the window to 120 days.

Study one: test whether the run line widens structurally after the starter exits, using inning-boundary frames rather than the mid. One-sidedness on the venue means the executable side is the floor, not the average of the two quotes.

Study two: replay a ninth-inning walk-off from the archive and count the seconds between the pitch outcome and the book's final repricing; the sub-second reprise pattern from the crypto corpus is the comparison.

Study three: compare first-inning lines against final-score lines across a single week of regular-season games to see whether early books trade systematically wider, holding market breadth constant.

A week of games at 250ms on Starter's 3-day window is the smallest honest study; the Pro tier to 120 days stretches it to a full homestand and keeps resolved games queryable.

The fastest possible start is one inning: pick any midweek game from the current series list, pull the eighth-inning window from the archive, and count how many frames the book spent with one side dominating the executable price. That single exercise teaches the one-sided share, the grid, and the bound faster than reading any documentation page.

Then repeat the pull on a ninth-inning walk-off game from the same night and diff the two windows. The difference between a tight two-sided inning and a resolved one-sided ninth is the exact signature that makes baseball books readable on this venue.

Finally, turn the diff into a standing check: bookmark the series endpoint, record the one-hour pull once a week, and keep every resolved game queryable so the homestand becomes your first reproducible dataset at 250ms.

• Sports data
• Market data
• Order book data