KalshiEye/kalshi-order-book-sample

Kalshi order book sample: one full day

One full UTC day (October 13, 2026) of Kalshi order book data for seven series, from the Kalshi Eye archive. It has every order book change at every price level, every trade, and every settlement. The files are the same daily Parquet files the archive publishes, with the same columns and folder layout, so code written against this sample runs unchanged on the full archive.

Series	What it is	Markets	Trades	Book changes (L2)
KXBTC15M	Bitcoin up or down, every 15 minutes	96	2,572,003	46,216,436
KXETH15M	Ether up or down, every 15 minutes	96	428,126	14,808,006
KXNFLGAME	NFL game winners	58	431,595	1,890,009
KXMLBGAME	MLB game winners	14	98,835	1,016,063
KXHIGHNY	Highest temperature in New York City	18	10,698	1,048,480
KXINXU	S&P 500 above or below, by hour	660	6,226	4,204,176
KXFEDDECISION	Fed rate decision	55	567	265,218

Also included for the whole day: the full Kalshi market list (178,700 markets, point in time), the series list, every settlement (119,997 markets), the recorder's incident log, and bitcoin and ether spot prices from Coinbase, Bitstamp and Kraken.

Datasets

Each one is a config in the viewer above and a folder under kalshieye-data/. Every column is described in the data dictionary.

Folder	What it holds
book_l2	Every price level and every change to the YES and NO bid books. Each file opens with a snapshot of every book at 00:00 UTC.
book_snap_1s	The top ten YES bid and ask levels at the end of every second in which the book changed.
bbo	Best YES bid and ask with sizes, one row each time any of the four changes.
bbo_1m	One-minute bars of the YES mid, with the closing bid and ask, trade count and volume.
trades	Every trade: YES price, size, and the taker's side.
settlements	Every market that settled that day: result, payout, and the value it settled on.
markets	Every market open at any time that day, with strikes, times, tick size, rules and fees.
series	Every Kalshi series: title, category, frequency, fees and settlement sources.
incidents	Every hour the recorder was not connected, missed sequence numbers or dropped messages.
spot	Best bid, ask and last for BTC and ETH on Coinbase, Bitstamp and Kraken, every update.

Quick start

Query it in place with DuckDB, without downloading anything:

# pip install duckdb
import duckdb

duckdb.sql("SET TimeZone = 'UTC'")  # show timestamps in UTC, as stored
duckdb.sql("""
  SELECT ticker, minute, yes_bid_close, yes_ask_close, trade_count
  FROM 'hf://datasets/KalshiEye/kalshi-order-book-sample/kalshieye-data/bbo_1m/*/series=KXNFLGAME/*.parquet'
  ORDER BY minute
  LIMIT 10
""").show()

Or download it all (845 MB, 41 files). The folder layout matches what Kalshi Eye's downloads produce, so the queries in How to backtest Kalshi markets run on it as written:

hf download KalshiEye/kalshi-order-book-sample --repo-type dataset --local-dir .

With the datasets library, pick a config:

from datasets import load_dataset

trades = load_dataset("KalshiEye/kalshi-order-book-sample", "trades", split="train")

Runnable examples that read this sample in place are in KalshiEye/kalshieye-examples: the busiest markets, spreads and depth, rebuilding the full order book from book_l2, a backtest, and a calibration chart.

Reading the data

• Prices are dollars between 0 and 1, sizes are contracts, and timestamps are nanoseconds UTC. ts_recv is when the recorder received a message and ts_venue is when Kalshi sent it. Order rows by ts_recv.
• Kalshi's books are bids only. A NO bid at price p is an offer to sell YES at 1 − p. bbo, bbo_1m and book_snap_1s already show asks in YES terms; book_l2 has the raw YES and NO bid books.
• In book_l2, snapshot rows set a level and delta rows add a signed change to it. Drop a level when its size reaches zero or below.
• Kalshi's taker fee per contract is fee_multiplier × 0.07 × P × (1 − P), rounded up to the cent per order. The multiplier is in markets and series.
• incidents lists 7 hours on this day where recording was incomplete. Leave those hours out of any test.
• SHA256SUMS has a checksum for every file. Each one matches the archive's published manifest for the day.

The full archive

Kalshi Eye has recorded every Kalshi market since October 12, 2026, and Polymarket, Limitless and Chainlink prices since October 15, 2026. Each day is published the next morning. See Kalshi historical order book data, data coverage and pricing. A free account downloads every dataset for the free sample days, and the free plan downloads the last 30 days of minute bars, trades, settlements and market lists.

License

CC BY-NC 4.0: free for research and non-commercial use, with credit to Kalshi Eye (kalshieye.com). For commercial use, see the plans or contact us.

Kalshi Eye is independent and is not affiliated with, endorsed by, or sponsored by Kalshi.

Citation

@misc{kalshieye_sample_2026,
  title        = {Kalshi order book sample: one full day},
  author       = {{Kalshi Eye}},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/KalshiEye/kalshi-order-book-sample}},
  note         = {Recorded October 13, 2026. Full archive at https://kalshieye.com/}
}