dataforge-labs/equity-perp-price-discovery

Equity and pre-IPO perpetual price discovery

What a price for Apple, Nvidia, OpenAI or Anthropic does when no cash market is open.

Perpetual futures on real-world assets trade around the clock, including the roughly 60% of the week when US equities are shut. The venue publishes an index history but no mark history, so the price it actually margins against, and the gap between that and the index, exist only if they were recorded as they happened.

Two of these names have no public market at any hour.

Contents

name	one row is
e23_perp_mark_index	one instrument at one moment: mark, index, and the basis between them

Reading it

market_type separates the population: equity, etf, commodity, fx, pre_ipo, and crypto. is_rwa marks the real-world-asset subset in one flag.

basis_bps is (mark - index) / index in basis points, signed, and null rather than zero where either leg is missing. Basis points make an instrument priced at 38 comparable to one priced at 78,000.

The crypto rows are the control, and they are included for that reason. Crypto perpetuals have no closed hours, so they show what this basis looks like when the underlying never stops. Compare an equity against them rather than against zero.

The interesting window is when the cash market is shut. One Saturday sample had the whole equity book live, with marks moving while their indices barely did.

Before you build on this

• The mark is the venue's own, used for margining. It is not necessarily a traded price, and a few of these instruments' books are in the separate crypto-execution-costs repo.
• How the index is derived for an equity while its cash market is closed is the venue's business and is not modelled here. It was observed to move slightly even on a Saturday, so treat it as the venue's reference rather than a last cash close.
• pre_ipo names have no public reference market at all, at any hour. Their basis is measured against a construction of the venue's own, so read it as internal consistency, not as a premium to a market price.
• One venue. This is where one book put the price, not a consensus.
• Instruments are listed and delisted over time, so count distinct instrument_name per day rather than assuming a fixed universe.

Partitions are parquet, one file per collection window, under dataset/YYYY/MM/. Every dataset here carries a FIXED 7-day sample WINDOW starting at its own first day of collection, together spanning 2026-08-29 to 2026-09-04, so you can check schema, coverage and quality before asking for more. It does not advance, so there is nothing to gain by re-downloading it. The full history is held privately, available on request.

from huggingface_hub import snapshot_download
import pandas as pd, glob

path = snapshot_download("dataforge-labs/equity-perp-price-discovery", repo_type="dataset",
                         allow_patterns="e23_perp_mark_index/**")
df = pd.concat(map(pd.read_parquet,
                   glob.glob(f"{path}/e23_perp_mark_index/**/*.parquet", recursive=True)))

Coverage

e0_run_manifest lists every collection window with its poll counts and failure counts, and is published in full rather than windowed. Gaps between windows are real, cannot be filled in afterwards, and nothing here is interpolated.

License and contact

ODC-BY: use it freely, credit "DataForge (dataforge-labs)". Questions and requests for the full history via the discussions tab.