--- url: "https://predictiontoolkit.com/tools/probalytics/" publisher: "PredictionToolkit" dateModified: "2026-11-03" --- [← Historical data]() ![Probalytics logo]() Dataset # Probalytics Probalytics stores prediction-market history for execution and microstructure research. It normalizes Polymarket and Kalshi markets, fills and outcomes, and records book states as they change rather than reducing the archive to coarse candles. [**4.0**/ 5Editorial rating]() [Editorial assessment]() [Explore Probalytics]() www.probalytics.io **Available on:** API **Pricing:** [See plans and costs ↓]() [Screenshots]() [Explore alternatives]() [Historical data]() [Data APIs]() +1 more [SDKs & libraries]() ## Screenshots and interface Product preview1 / 2 The provider’s unified schema and data-access overview. ## What is Probalytics? Full bid and ask ladders, millisecond-level timestamps and continuity metadata support detailed replay work. ClickHouse SQL filters and joins near storage; REST and a Python SDK integrate the data with research code, while Parquet exports support reusable local datasets. Public samples and typed table schemas allow inspection before signup. Market and fill history can begin earlier than order books. Polymarket book coverage starts in November 2025 and Kalshi in May 2026; past books cannot simply be recovered from venue APIs. Match the venue, dataset and capture window to the intended experiment, then supply explicit size and latency assumptions for any execution model. ### Who it is for - For quants modeling historical fills against depth and researchers joining market, fill and outcome records. - For developers needing bulk history they can retain in a local analytical workflow. ## Features and coverage - **State-change book history** Query full recorded bid and ask ladders with millisecond timestamps, using state and continuity fields to interpret the sequence. - **Normalized identifiers** Join market, fill and outcome tables through normalized identifiers while retaining the platform identifier to keep venues distinct. - **SQL, code and bulk exports** Use ClickHouse SQL, REST, the Python SDK or plan-eligible Parquet exports. Inspect public rows and typed schemas before committing to a dataset. ## Workflows with Probalytics ### Model historical execution Reconstruct the book for the selected market and outcome in time order. Estimate hypothetical fills only after specifying the order size and delay; the archived quotes alone do not describe how that order would have executed. ### Build a cross-venue panel Join markets, fills and outcomes in SQL and retain venue identifiers when comparing observations. ### Check the capture window Inspect venue start dates, continuity and missing periods before measuring a strategy or requesting a large export. ## Pricing and total cost Paid plans offer a three-day trial limited to the latest seven days across venues and access methods. Historical full-depth books require Orderbook or Custom rather than Starter or Pro alone. ### Starter $39/month - Rolling 30-day market and trade history; REST and SQL - 100 queries/hour and 100,000 result rows ### Pro $129/month - Full available market and trade history; 1,000 queries/hour - Up to 1 billion result rows and Parquet exports ### Orderbook $249/month - Pro features plus historical full-depth books ### Custom Contact sales - Custom retention, tables, limits and support ## Getting started **Before you start:** Historical books cannot simply be backfilled from venues. Inspect missing capture periods and continuity rather than assuming complete coverage. 1. Inspect sample rows and check the coverage table for the required venue and dataset. 2. Choose a trial and access method, then run a narrowly filtered query. 3. Validate identifiers and continuity before expanding the query or exporting bulk history. Book coverage starts in November 2025 for Polymarket and May 2026 for Kalshi; earlier market metadata does not imply earlier book history. [Read dataset documentation]() ## Useful links - [View data plans]() ### What to check - Historical books cannot simply be backfilled from venues. Inspect missing capture periods and continuity rather than assuming complete coverage. - A recorded quote does not by itself establish a realizable fill; a realistic backtest needs an execution model. A closer look ## Ratings & reviews Editorial rating **4.0**/ 5 0 user reviews PredictionToolkit editorial ### Our take on Probalytics Full-depth history, precise timing and SQL, SDK and export routes form a well-specified research product. Its usefulness for replay depends on matching venue start dates, capture completeness and trial restrictions to the period being studied. Based on documented features, setup and access terms. [How we rate]() ### User reviews 0 The main score switches to user ratings after 5 published reviews. **Be the first to share your experience** A specific example helps the next person choose the right tool. ## Related tools [Browse category]() ![Assymetrix prediction-market API overview and supported platforms]() ![Assymetrix logo]()[Assymetrix]() Data API**3.5**Editorial A live multi-venue data portal for canonical markets and history, alongside a separate consumer-platform preview. - Access Web · API Compare current and historical API access. ![Convex Lake endpoints, authentication and fields]() ![Convex Lake logo]()[Convex Lake]() Dataset**3.5**Editorial Retrieve historical prediction and options datasets for depth research, model testing and repeatable batch ingestion. - Access API · MCP Also in Historical Data and Archives. ![PolyOrderbooks historical L2 overview and provider book example]() ![PolyOrderbooks logo]()[PolyOrderbooks]() Dataset**3.5**Editorial Query historical books at 250 ms intervals or replay supported crypto strategies with depth-based fills. - Access Web · API · MCP Compare another order-book archive. ## Frequently asked questions about Probalytics ### Does Starter include historical books? No. Historical full-depth books require Orderbook or Custom. Starter covers rolling market and trade history. ### What does the trial cover? Three days of trial access to the latest seven days across venues and access methods. In this profile