Polymarket Sports Data API: Markets, Live Books and Historical Depth

The core object chain

• Select a league and list the active Polymarket sports markets.
• Resolve the event, condition and outcome-token identifiers.
• Query the latest full bid/ask ladder for a selected token.
• Pull recorded book history for replay within the plan window.
• Join the market to a game record, score and settlement when available.
• Use source freshness fields before labeling any displayed price live.

Prediction-market prices are not conventional odds feeds

A Polymarket outcome token trades on a central limit order book. Its displayed probability depends on bid, ask and available size, while a sportsbook feed represents a different market structure. Comparing them can be useful, but the join should keep source, timestamp, rules and outcome mapping explicit.

Do not force every contract into a moneyline schema. Props, tournament winners and conditional outcomes need stable market text and identifiers before they can be compared safely.

DepthFeed sports endpoints

The DepthFeed sports API exposes an overview, league and market discovery, latest and historical books, games, linked markets, scores, players, injuries and sportsbook-line context. The same account and API key cover crypto and sports REST access.

Paid plans can subscribe to concrete Polymarket sports book channels on the main WebSocket. Historical sports depth is forward-captured, so every page and API response should state the available coverage rather than implying history before collection began.

Research uses

Model the sports market taxonomy

A sports catalog should distinguish league, game, market family, proposition, outcome and market window. Moneyline, spread-like propositions, totals, player props, tournament winners and conditional markets should not be forced into one label. Preserve the exact contract text and official identifiers beside any normalized taxonomy.

Build mappings conservatively. Two markets that mention the same teams may settle on regulation time, include overtime, use different handicap values or resolve under different cancellation rules. A verified link should carry the matching method and source so downstream analysis can distinguish an exact join from a text-based candidate.

Synchronize games, markets and books

Store scheduled game time, market open and close times, book observation time and score update time as separate fields. They describe different systems and can arrive out of order. At any decision point, use the latest preceding value from each source and show its age.

When a game is postponed, abandoned or corrected, do not rewrite the historical decision record. Append status changes and let the official contract determine settlement. This makes it possible to reproduce what a monitor displayed before and after the correction.

Sports order-book metrics

Cross-source probability comparison

Convert each source into a comparable representation only after retaining its native quote, timestamp, rules and margin structure. A sportsbook price may include vigorish; a prediction-market quote contains a bid-ask spread and finite displayed size. Comparing a book midpoint with an unadjusted sportsbook implied probability can create an artificial difference.

Define outcome mappings and remove margin with a disclosed method when appropriate. Then align by as-of time, reject stale observations, and test whether the apparent gap survives executable prices and size. The comparison is a research signal, not proof of arbitrage.

Polymarket sports API quality checklist

• Retain event, condition and outcome-token identifiers with the exact contract text.
• Expose bid, ask, sizes and observation freshness rather than one unlabeled probability.
• State the earliest historical-depth date and leave pre-capture history unavailable.
• Version game-to-market mappings and distinguish exact from inferred links.
• Apply contract-specific settlement and cancellation rules.
• Use strict as-of joins for scores, injuries, lines and book observations.
• Report gaps, stale sources and unsupported market families explicitly.

Point-in-time sports research

A reproducible sports study needs the state that was knowable at the decision time, not the final cleaned record. Store book observations, score changes, game status, injury updates and comparison lines as timestamped events or snapshots. Join the latest preceding record and impose a maximum acceptable age for every source.

Late corrections should remain visible as corrections. If a game time, participant or score changes, do not rewrite the earlier research row. Version mappings and statuses so an analyst can explain why a signal appeared with the information then available.

Historical order-book depth begins when capture begins; it cannot be reconstructed from later trades or candles. Coverage reports therefore need eligible-market counts, observed-market counts, decision coverage and explicit gap reasons. That denominator is essential when comparing two models or two providers.