Lighter Order Book Dataset

Lighter Historical Order Book Data

Lighter perpetual and spot order book events packaged for replay, latency analysis, and market microstructure research.

No credit card · The sample is one real hourly Parquet file and needs no API key

Is this the right dataset?

Choose the order book detail your research needs

This landing page is for coverage, schema, sample files, and downloads. Use the SDK docs for method parameters and the reconstruction tutorial for replay logic.

L1 / top of book

Best bid and best ask only. Derive this view after replaying the L2 snapshots and deltas; ticker statistics are not a substitute for quotes.

L2 / price-level depth

This dataset. It records aggregate quantity at each visible price level so you can measure spread, depth, and imbalance.

L3 / individual orders

Not provided. L2 cannot reveal queue position or how many individual orders make up a level unless the venue publishes an order count.

Sample

See the data before you sign up

The first rows of one real hour: BTC on Lighter, 2026-09-02 at 12:00 UTC. One row per bid or ask price level in a snapshot or update.

lighter/2026-09-02/12/BTC_orderbook.parquetDownload full file

received_time	event_time	transaction_time	symbol	event_type	first_update_id	final_update_id	prev_final_update_id	last_update_id	side	price	quantity	order_count
1788351287231700242	1788351287231700242	null	BTC	snapshot	null	953873	null	953873	bid	20000.0	0.10000	null
1788351287231700242	1788351287231700242	null	BTC	snapshot	null	953873	null	953873	ask	119000.0	0.00094	null
1788351287231700242	1788351287231700242	null	BTC	snapshot	null	953873	null	953873	bid	30000.0	0.05000	null
1788351287234708698	1788351287234708698	null	BTC	update	null	953880	null	953880	bid	76731.7	0.00521	null
1788351287234708698	1788351287234708698	null	BTC	update	null	953880	null	953880	ask	76751.6	0.00000	null
1788351287234708698	1788351287234708698	null	BTC	update	null	953880	null	953880	bid	76735.1	0.00260	null

Showing 6 of 603,855 rows in this hourTimestamps are raw integers; units are in the schema below

Wire format

From exchange message to Parquet row

Every row starts as a message on the exchange's WebSocket or REST API. Point at a line or a column to see where each value comes from and how we convert it. Our test suite replays these messages through the production pipeline, so the rows shown are the rows you download.

Lighter (Perpetuals & Spot)

When the collector subscribes, also after each reconnect, Lighter sends the order book as subscribed/order_book. Each snapshot becomes one snapshot row per level and starts a new nonce chain.

Exchange sends

WebSocket message · order_book/<index>received 2026-10-05 19:15:04.141358 UTC

{
  "channel": "order_book:1",
  "last_updated_at": 1791227703748044,
  "offset": 4577935,
  "order_book": {
    "code": 0,
    "asks": [
      {
        "price": "85635.5",
        "size": "0.03539"
      },
      {
        "price": "85635.7",
        "size": "0.01700"
      }
    ],
    "bids": [
      {
        "price": "85635.3",
        "size": "0.00020"
      },
      {
        "price": "85635.2",
        "size": "0.00020"
      }
    ],
    "offset": 4577935,
    "nonce": 23863347847,
    "last_updated_at": 1791227703748044,
    "begin_nonce": 0
  },
  "timestamp": 1791227703769,
  "type": "subscribed/order_book"
}

Shortened: Lighter sent 1821 bid levels and 1271 ask levels in this snapshot. The example shows the best two levels of each side.

You download

Parquet row · orderbook · 1 of 4 rows from this message

Point at a line in the message or a column in the row to trace where each value comes from. Click a column to pin it.

Our tests replay this message through the production pipeline

All Lighter (Perpetuals & Spot) mapping rules →

Conventions for every exchange are in the wire formats overview. Column types and units for order book data are in the schema below.

Use cases

What you can build with order book data

Spread, depth, and imbalance: Replay the book to measure bid/ask spread, depth within N basis points, and order book imbalance at any moment in the hour.
Execution and market-making backtests: Simulate fills against the liquidity that was actually resting on the book instead of assuming a mid-price fill.
Cross-venue liquidity research: Compare depth and spread across exchanges with one schema instead of maintaining a parser per venue.

Coverage

Order book coverage

Every venue is stored under its own exchange ID, so spot and derivatives never mix. Bars show each venue's history from its first hourly file to today, or to its last file for a venue that is no longer collected. Symbol counts are live from the symbols API.

Order book coverage by venue, exchange ID, market, history start, and live symbol count
Venue	Exchange ID	Market	History	Order book symbols
Lighter	lighter	Perpetuals & Spot	2025-12-3011 mo	252

Other exchanges

Order Book data on other exchanges

The same schema on every venue, so switching exchange is a one-argument change.

Binance Historical Order Book Datasince 2025-06-28

Bybit Historical Order Book Datasince 2025-06-28

OKX Historical Order Book Datasince 2025-06-28

Hyperliquid Historical Order Book Datasince 2025-09-29

Aster Historical Order Book Datasince 2025-09-29

Bitget Historical Order Book Datasince 2025-07-15

Kraken Historical Order Book Datasince 2025-06-28

BitMEX Historical Order Book Datasince 2025-09-29

Or compare every venue on the cross-exchange order book overview.

Schema · 13 columns

Normalized L2 order book schema

Every row represents one bid or ask price level inside a snapshot or incremental update. Rows carry exchange timing, sequence fields when available, price, quantity, and event type.

Columns in every order book Parquet file
Column	Type	Description
received_time	INT64	Unix timestamp (nanoseconds) when our system received the event
event_time	INT64	Unix timestamp when the exchange generated the event (the time unit changes with the exchange)
transaction_time	INT64nullable	Exchange-specific transaction timestamp (if available)
symbol	STRING	Trading pair symbol (for example, "BTCUSDT")
event_type	STRING	Type of order book event ("snapshot" or "update")
first_update_id	INT64nullable	First update ID in the update sequence
final_update_id	INT64nullable	Final update ID in the update sequence
prev_final_update_id	INT64nullable	Previous final update ID (to find gaps)
last_update_id	INT64nullable	Last update ID that the exchange processed
side	STRING	Order book side ("bid" or "ask")
price	STRING	Price level (stored as string for precision)
quantity	STRING	Quantity at the price level (stored as string for precision). The unit is the base asset on most venues. OKX derivatives use contracts, and BitMEX uses raw instrument units (refer to the venue notes)
order_count	INT64nullable	Number of orders at this price level (if the exchange gives it)

Quickstart

From pip install to DataFrame in a minute

The example loads BTC from Lighter for 2026-09-02. Change the symbol, exchange, or dates to pull any other slice of the archive.

Install the Python SDK
pip install cryptohftdata
Create a free API key
Sign up with email or Google, then generate a key from your dashboard. No credit card.
Create account
Call get_orderbook()
Paste the snippet, replace YOUR_API_KEY, and you get a pandas DataFrame with the columns listed above.

Read order book SDK docs Reconstruct an L2 book Explore in the browser

lighter_orderbook_data.py

1import cryptohftdata as chd
2
3client = chd.CryptoHFTDataClient(api_key="YOUR_API_KEY")
4
5df = client.get_orderbook(
6    symbol="BTC",
7    exchange=chd.exchanges.LIGHTER,
8    start_date="2026-09-02",
9    end_date="2026-09-02",
10)
11
12print(df.head())
13print(f"Order book rows: {len(df):,}")

REST · no API key needed for single files

curl --fail --location \
  "https://api.cryptohftdata.com/download?file=lighter/2026-09-02/12/BTC_orderbook.parquet" \
  -o BTC_orderbook_12.parquet

Caveats

Know this before you backtest

Exchange feeds differ in subtle ways. These are the details that matter for accurate backtests, replay pipelines, and features. Missing hours and pending corrections for every venue are listed on Known Gaps & Corrections.

L2 order book data is aggregated by price level, not individual L3 orders.
Files contain snapshots plus incremental updates. Build state from a complete snapshot, apply later events in stored order, and reset whenever another snapshot appears.
Sequence fields vary by venue. If available continuity IDs show a gap, discard the local book and wait for a newer snapshot rather than replaying a state you cannot verify.
Quantities are in the venue's unit: base asset on most venues, contracts on OKX derivatives (multiply by the instrument's ctVal), and raw per-instrument units on BitMEX.
Hour-start checkpoint snapshots exist for Binance, Aster, OKX, Bybit, Bitget, Lighter, Kraken Futures and Hyperliquid, not for Kraken spot or BitMEX. A checkpoint is written only when the book is proven current, so some hours start without one; continue from the previous hour's book or wait for the next snapshot.
In some periods only part of the Lighter market universe was published (for example 2026-09-22 16:02 to 2026-09-23 about 00:30 UTC). These periods cannot be recovered.
This page starts on 2025-12-30 based on the earliest stored hourly files for its covered exchange IDs and grows as new hourly files are published.

FAQ

Frequently asked questions

Is Lighter order book data free to download?

Yes. A free CryptoHFTData account with no credit card gives API access to Lighter order book data and every other dataset. Individual hourly files can also be downloaded without an API key on the rate-limited free tier.

How far back does Lighter order book data go?

The earliest hourly files start on 2025-12-30, and new files are added every hour. Start dates differ by venue; the coverage table on this page lists each one.

What format are the files in?

Each file holds one symbol for one UTC hour, stored as Zstd-compressed Parquet under {exchange_id}/{YYYY-MM-DD}/{HH}/{SYMBOL}_orderbook.parquet. Parquet reads directly into pandas, Polars, DuckDB, and Arrow.

How do I load Lighter order book data in Python?

Install the SDK with pip install cryptohftdata, create a client with your API key, and call client.get_orderbook() with a symbol, exchange, and date range. It returns a pandas DataFrame.

Is this L2 or L3 order book data?

L2. Each row is the aggregate quantity at one price level, recorded in snapshots and incremental updates. Individual orders (L3) and queue position are not included.

Start using order book data today.

One free API key covers every exchange and every data type. $0.00.

Get free API key