Sample

Check it against your own data.

Real production rows, not a mock-up — the same file contents a paying customer receives. No signup, no sales call. Load it, compare it against whatever source you already trust, and decide from evidence.

58,030 rows
87 snapshots
674 instruments
underlying
ETH
date
2026-07-14
window
12:00–14:00 UTC
format
Parquet · zstd
size
2.3 MB

What you can check with it

That the schema is what we say

Every column on the schema page, with real values and real nulls. Confirm dtypes, units and nullability before you write a loader.

That the numbers reconcile

Mark, bid and ask IVs and the Greeks are the exchange's own published values, captured at the snapshot instant. If you hold an independent source for this window, they should line up row for row.

That nulls are honest

You'll find nulls in bid_iv on illiquid strikes. That's the exchange publishing no bid, stored as null rather than filled with a value nobody quoted.

That the cadence is real

Count distinct timestamps and check the spacing yourself. Then read coverage, where every gap over five minutes is listed with its UTC timestamps.

Loading it

pandas
import pandas as pd

df = pd.read_parquet("volar-sample-ETH-2026-07-14.parquet")
print(df.dtypes)
print(df.timestamp.nunique(), "snapshots")
DuckDB
import duckdb

duckdb.sql("""
  SELECT instrument, mark_iv, bid_iv, ask_iv, delta
  FROM 'volar-sample-ETH-2026-07-14.parquet'
  WHERE type = 'call'
  LIMIT 20
""").show()
One fixed window, deliberately. This is the same file for everyone and it never rotates — we publish metadata about the data freely, and sell the data itself. If you need a different date or a wider slice to complete an evaluation, ask and we'll usually say yes.
Read the schema See coverage Get an API key