Kalshi and Polymarket Historical Data: APIs, Downloads and Backtest Coverage
Retrieve Kalshi historical data and Polymarket price history. Compare candles, trades, archived books, API downloads, and Turbine backtest order-log exports.

Kalshi historical data is available through official live and historical API endpoints, separated by published cutoff timestamps. Polymarket offers official token price history, while order-book replay requires separately recorded books. These sources answer different questions: a price chart or trade print does not prove that an order of your size could have filled.
This guide explains retrieving source data and reviewing coverage. For the testing workflow, see Kalshi backtesting without Python or Polymarket testing options without coding.
Which historical data do you need?
| Data | What it describes | What it cannot establish alone |
|---|---|---|
| Candles or token price history | Prices summarized or sampled over time | Full historical depth or your order's queue position |
| Trade prints | Executed transactions | All resting or canceled orders |
| Recorded order books | Bid/ask levels and available sizes at recorded times | Future liquidity, order-arrival latency, or guaranteed fills |
| Market metadata and settlement | Contract identity, rules, dates, and outcome | A complete tradable price path |
| Turbine backtest order-log CSV | Simulated orders and fills produced by one strategy test | A raw venue-wide historical data download |
Record source, venue, market ID, requested and actual dates, timestamp units, sampling interval, and missing periods. Check access terms and licensing before redistributing a dataset. Third-party access does not imply complete coverage or permission to republish.
Kalshi historical data and the historical API
Kalshi publishes GET /historical/cutoff to identify which records have moved into historical storage. Read the cutoff for the data type you need; there is no single permanent three-month boundary for every endpoint.
Older markets, candles, and trades have dedicated historical endpoints. Private fills, orders, and positions are account-scoped and require the appropriate authentication. Follow the official historical-data guide and current API rate limits rather than assuming a full-history pull fits one request.
A read-only request you can run
This public request downloads the cutoff metadata and never places an order:
curl --fail --silent --show-error --max-time 20 \
'https://external-api.kalshi.com/trade-api/v2/historical/cutoff'On October 8, 2026, our request returned these fields:
{
"market_settled_ts": "2026-08-09T00:00:00Z",
"trades_created_ts": "2026-08-09T00:00:00Z",
"orders_updated_ts": "2026-09-24T00:00:00Z",
"market_positions_last_updated_ts": "2026-08-09T00:00:00Z"
}This is a dated metadata sample, not a promise of coverage for a particular market. Route the next request to the corresponding live or historical endpoint, follow pagination, and preserve the exact request with your output. Candlestick history and trade history are different datasets; neither is a complete historical book.
The Kalshi build guide explains how API development differs from Studio's no-code strategy workflow.
Polymarket historical data and data downloads
Polymarket's official price-history endpoint returns token price samples in a history array, with t timestamps and p prices. Use the token ID for the desired outcome, not just its human-readable market title.
The request below is a template: replace YOUR_TOKEN_ID with a real token ID obtained from the venue's market metadata. It downloads JSON price history, not archived order-book depth.
curl --fail --silent --show-error --max-time 20 --get \
'https://clob.polymarket.com/prices-history' \
--data-urlencode 'market=YOUR_TOKEN_ID' \
--data-urlencode 'interval=1d' \
--data-urlencode 'fidelity=1' \
--output polymarket-price-history.jsonAn empty response is not evidence of a complete period with no price movement. Check token identity, dates, endpoint limits, and source availability. Preserve the original JSON before converting samples to CSV.
For trades or settlement research, use the venue's relevant API or on-chain records and document how you identify unique transactions. Do not apply a blanket volume correction to every dataset: inspect the source schema and aggregation first. A current order-book response cannot recover a past book.
What historical data does Turbine use?
Turbine uses supported Kalshi candles and recorded PMXT order-book archives where available. Source coverage depends on the venue, market, window, and rule inputs. It does not promise every trade since launch, a complete subgraph mirror, or a raw-data download service.
Polymarket price history is useful for research, and a Paper Run can check a Polymarket strategy on new market data. See the Polymarket backtesting guide.
Kalshi results may report partial coverage and different actual dates from the request. A detected historical replay gap can fail the test with the affected market and UTC bounds. Read the Kalshi historical testing reference and compare the requested and admitted windows before reporting performance.
A completed test has a public run page with a downloadable order-log CSV of its simulated orders and fills, up to 10,000 events; see a recorded example. That file supports reviewing one strategy; it does not deliver an unrestricted raw historical venue dataset.
Frequently Asked Questions
Where can I find Kalshi historical data?
Start with Kalshi's official API and its historical-data guide. Retrieve the current cutoff metadata, choose live or historical endpoints for the needed record type, and follow pagination. Inspect the market and window rather than assuming complete coverage from an endpoint's existence.
Can I download Polymarket historical data?
The official price-history endpoint can download token price samples as JSON, which you can convert to CSV. That is price history, not a complete order-book archive. Use the correct token ID and document dates, fidelity, gaps, and access terms.
Are candles enough to backtest market making?
Candles alone cannot reconstruct resting-order depth or queue position. A quoting model needs additional assumptions or recorded books and still differs from live execution. Kalshi backtests in Turbine use a separate resting-order model.
Does Turbine export raw Kalshi or Polymarket history?
A completed strategy test offers its simulated order log as a CSV on the run's public page. It is not a venue-wide raw historical-data download. For source data, follow the relevant official API or archive provider's access terms.
What should I do when historical coverage is missing?
Retain the limitation, verify market identity and requested dates, and inspect the reported missing or clipped window. Do not invent fills or silently replace unavailable order books with price samples. Use a supported test or a separate paper workflow, with its assumptions disclosed.
