Kalshi · Book snapshots
Full depth snapshots per market, sorted market first; 5,000-row groups.
- Rate
- $80 / GiB
- Available daily files
- 187
- Sealed date bounds
- 2026-01-08 — 2026-10-16
- Listed archive size
- 17.63 GiB
The recorded market. In your hands.
Kalshi datasets, ready to download.
Buy recorded Kalshi yes/no order-book snapshots and level-change datasets in daily Parquet files. Keep market tickers, timestamps, and venue-native ladders in your research.
One-time file purchase · no API subscription · card or credits
ParquetColumnar files for your own stack
UTC daysChoose the dates you actually need
Exact quoteReview the price before paying
30 daysTo issue fresh download links
The archive catalogue
Keep both sides of the market.
Choose a dataset and date range. See the files and estimated price before creating an account.
Inventory snapshot ·
Sealed daily Parquet files
.parquetKalshi · Book snapshots
Full depth snapshots per market, sorted market first; 5,000-row groups.
187 available UTC days2026-01-08 — 2026-10-16
Latest 90 available daily files · dark blue = selected · light blue = available. One square per listed file.
Date range UTC
| UTC day | File size | Rows | Version |
|---|---|---|---|
| 2026-10-16 | 0.14 GiB155,493,553 bytes | 8,545,572 | v1 |
| 2026-10-15 | 0.14 GiB152,582,294 bytes | 8,546,708 | v1 |
| 2026-10-14 | 0.14 GiB149,096,458 bytes | 8,656,494 | v1 |
| 2026-10-13 | 0.13 GiB144,427,707 bytes | 7,825,572 | v2 |
| 2026-10-12 | 0.13 GiB134,272,906 bytes | 6,789,013 | v1 |
| 2026-10-11 | 0.09 GiB95,521,235 bytes | 4,976,325 | v1 |
| 2026-10-10 | 0.09 GiB92,806,158 bytes | 5,093,466 | v1 |
Know what you’re buying
Choose the observation
your model needs.
Recorded depth is different from a closing price. Book snapshots and level changes answer different research questions.
Kalshi · Level changes
Every price-level change (deltas) per market; 25,000-row groups.
- Rate
- $80 / GiB
- Available daily files
- 71
- Sealed date bounds
- 2026-08-07 — 2026-10-16
- Listed archive size
- 181.03 GiB
Yes/no ladders, with Kalshi identity intact.
Kalshi book files retain ticker, series, receive timestamps, and the yes_prices / yes_sizes and no_prices / no_sizes arrays. These are venue-native yes/no ladders; do not silently treat them as a normalized bid/ask table.
Level-change files include seq, kind, side, price, delta, and size. Replay needs an appropriate seed and gap handling. Updates are not executed trades, and a file cannot recover observations made before recording began.
Inspect before you commit
Know the format.
Test your workflow.
Explore clearly labelled synthetic rows, inspect the fields, and try a Python notebook. No account needed. Your quote lists the real daily files available for purchase.
Synthetic compatibility sample
8 demo rowsKalshi · Book snapshots
Try the file format and your own code before buying. These values are fabricated, with demo identifiers and a 1 January 2000 UTC clock. They are not captured market history or evidence of archive coverage.
| Time (UTC) | ticker | yes_prices | no_prices | yes_sizes |
|---|---|---|---|---|
| 00:00:00.025 | DEMO-MARKET-001 | [0.4, 0.39] | [0.56, 0.55] | [100, 50] |
| 00:00:01.025 | DEMO-MARKET-001 | [0.41, 0.4] | [0.55, 0.54] | [105, 50] |
| 00:00:02.025 | DEMO-MARKET-001 | [0.42, 0.41] | [0.54, 0.53] | [110, 50] |
| 00:00:03.025 | DEMO-MARKET-001 | [0.41, 0.4] | [0.55, 0.54] | [115, 50] |
| 00:00:04.025 | DEMO-MARKET-001 | [0.4, 0.39] | [0.56, 0.55] | [120, 50] |
| 00:00:05.025 | DEMO-MARKET-001 | [0.42, 0.41] | [0.54, 0.53] | [125, 50] |
| 00:00:06.025 | DEMO-MARKET-001 | [0.43, 0.42] | [0.53, 0.52] | [130, 50] |
| 00:00:07.025 | DEMO-MARKET-001 | [0.42, 0.41] | [0.54, 0.53] | [135, 50] |
Selected columns shown. Prices are fractions of $1; sizes use venue-native quantities. Kalshi YES and NO ladders are separate sides. Level-change files are updates, not a trade tape.
Understand all 9 fields
tickerString- Market or outcome identifier. Demo values here; retain source identifiers in purchased files.
seriesLowCardinality(String)- Venue-specific metadata; inspect the purchased file schema before processing.
base_assetLowCardinality(String)- Venue-specific metadata; inspect the purchased file schema before processing.
market_typeLowCardinality(String)- Venue-specific metadata; inspect the purchased file schema before processing.
recv_ts_msUInt64- Unix milliseconds; display in UTC.
yes_pricesArray(Float64)- Price as a fraction of a $1 contract. Ladders retain venue-native sides.
yes_sizesArray(Float64)- Venue-native quantity. Price and size arrays align by index.
no_pricesArray(Float64)- Price as a fraction of a $1 contract. Ladders retain venue-native sides.
no_sizesArray(Float64)- Venue-native quantity. Price and size arrays align by index.
Free sample
One fixed, synthetic example. Use it to test loading, column types and analysis code.
Your purchased files
Recorded observations for the products and UTC days in your quote. Review listed file sizes, row counts and coverage before paying. Archive versions can differ.
A file, then a query
Bring your own research stack.
Download a synthetic sample and inspect its fields locally. Use the same tools on purchased daily files, checking their schema first.
Read the download & format guide ↗-- DuckDB: run beside the downloaded sample
DESCRIBE SELECT *
FROM read_parquet('kalshi-book.parquet');
SELECT *
FROM read_parquet('kalshi-book.parquet')
LIMIT 2;From archive to analysis
Choose. Pay. Download.
A one-time purchase for a defined set of files. No sales call required.
- 1
Review your files
Pick the dataset and UTC dates. Create a free account and accept the data licence to get the exact quote.
- 2
Pay when you’re ready
Use hosted card checkout or existing credits. Creating a quote or payment link does not charge your card.
- 3
Keep the data locally
Get your files from the paid-order dashboard and emailed links. Refresh signed URLs for 30 days.
For code-driven research
Request by API.
Let a person pay.
Create the quote in code, send the hosted payment URL to the buyer, then retrieve the paid order. Payment is confirmed by Stripe before download access is granted.
Read the bulk purchase API ↗# After accepting the current data licence:
POST /v3/data/quotes
{ "items": [{
"product": "kalshi-book",
"from": "2026-10-16",
"to": "2026-10-16"
}] }
POST /v3/data/orders/checkout
{ "quote_id": "<quote.id>" }
→ payment_url # human completes checkout
GET /v3/data/orders
POST /v3/data/orders/<paid-order-id>/linksBefore you download
The practical details.
Do I need an API subscription to buy files?
No. Create a free DepthFeed account, accept the data licence, choose available files, and review the exact quote. Bulk file purchases are priced separately from API subscriptions. Pay by card or with existing credits.
What is inside a daily file?
A sealed Parquet file for one product and UTC day. Book products contain recorded full-depth observations. Level-change products contain observed price-level updates, not a trade tape. Availability and capture history vary by product and day.
Can I inspect a sample without signing up?
Yes. Each dataset has a free fixed synthetic Parquet sample, matching JSON and an example Python notebook. The eight fabricated rows let you inspect the schema and test your loader without accessing the historical archive. They do not demonstrate real market performance or coverage. Archive versions may differ.
How are files priced?
The estimate uses the sealed file bytes and the listed USD per GiB rate, with a $5 minimum per order. One GiB is 1,073,741,824 bytes. The signed-in quote confirms current availability and the exact subtotal; applicable tax is shown at checkout.
How do I receive my downloads?
After confirmed payment, your order appears in the bulk-data dashboard and download links are emailed to your account address. Signed links last one hour. You can issue fresh links from the dashboard or API for 30 days after the order. Save the files locally during that window.
Can my code request files and let a person pay?
Yes. Create a quote through the API, then request a hosted payment URL. Opening a quote or payment link does not charge a card. A person completes checkout, and you retrieve the paid order and signed download links through the API.
Are all calendar days and markets covered?
No. The catalogue lists the sealed days currently available for purchase. A first/last date does not guarantee an uninterrupted series, every market, or uniform intraday capture. Inspect the file list and samples before ordering; do not infer missing observations.
Can I redistribute files or request a refund?
The standard licence permits internal research and use; it does not permit redistributing the raw dataset. All orders are final and payments are non-refundable. Review the data licence and sample before buying. Contact DepthFeed about redistribution needs before placing an order.
Start with the files that fit
Less collecting.
More research.
Inspect a sample. Choose your dates. Know the price.
Choose your datasetAll orders are final. Internal-use licence · Terms · Need a different licence? Contact us