Crypto Futures Research Dataset

What these rows include—and what they cannot tell you

A row joins a venue to a researched asset. Some rows contain an exact symbol returned in the targeted contract collection; others retain documentary references for an asset without a confirmed symbol. The recordScope field makes that distinction explicit. Counting rows as currently tradeable contracts would overstate the evidence.

The 41 assets include OP, while the editorial market cluster contains 40 guides. That is a difference in coverage, not an extra article or a missing country. The catalog also differs from the immutable Research studies: their observation windows, contract sets and execution samples are defined by their own editions. This download does not silently replace those studies.

Leverage ceilings, normalized tier counts, fee references, funding intervals, source states and individual check dates are included. Full order books, every margin-band value, personal VIP rates, account permissions and complete funding histories are not. For execution evidence or a size-dependent conclusion, follow the relevant study or tool rather than ranking these rows by their largest leverage number.

The historical field names liveProvider and adapterState describe the collection method recorded in the dataset. They do not enable live quotes: the site no longer requests market prices or funding rates while you use its tools. Coinbase’s retired INTX identifier does not describe its successor. The coverage matrix explains field availability separately.

Read one real BTC row before using the whole file

The different dates are intentional. A refreshed leverage endpoint does not recheck the fee document. A newer catalog edition does not make the stored spread current. Where a cap was converted to dollars using an observed price, that valuation belongs to the observation rather than being a permanent dollar limit.

For a non-confirmed row, a venue-level reference may provide context without matching the asset in every field. For example, a source URL referring to BTC does not prove TON order-book depth. Treat the source’s scope and the actual value together; a retrieved label alone is not enough to promote a missing observation into an asset-specific fact.

Null, zero and a retrieved document are different things

The JSON preserves missing numeric values as null. CSV represents the same absence as an empty cell. A numeric zero remains 0 in both formats, and booleans remain true or false. Avoid spreadsheet import settings that replace blanks with zero or interpret dates as local-time measurements.

One stored example is Aster–BTC, where makerFeePct is 0. This says the catalog contains a zero reference; it is not an entitlement to a current manual, API or regional promotion. By contrast, GMX–BTC has makerFeePct and takerFeePct set to null because those maker/taker fields do not contain its complete pool-based fee model. Null does not mean commission-free.

A tierCount of 0 means no normalized bands retained, not a maintenance requirement of zero. A null position limit is unknown, not unlimited. sourceCount belongs to the venue and repeats on its asset rows; summing it would multiply the same references. Source states describe retrieval history, not legality, account approval or proof of a completed trade.

Field dictionary: types, units and interpretation

Schema 2.0 retains the existing columns and adds recordScope, feeScope and adapterState. Fee percentages use percentage points, leverage uses multiples and spreads use basis points. Thus 0.05 in a fee column means 0.05%, while 5 bps describes the same fraction. The distinction matters before any cost calculation.

Current download or fixed edition: choose the right citation

The current endpoint can change when normalized records or supporting fields change. Use it to inspect the latest catalog, and save the response if you rely on a value. For a published comparison, the edition below fixes both the CSV and JSON bytes so another reader can retrieve the same records in the same order.

Field dates inside the fixed files remain their original dates. If a later correction is needed, it should receive a new edition and a correction note; the previous bytes are not overwritten. The manifest’s creation time identifies the freeze, while the source dates identify the supporting observations.

The compiled dataset is distributed under CC BY 4.0. Attribute the compilation, link the license and identify modifications. Linked exchange documents, screenshots and trademarks keep their owners’ terms; the dataset license does not relicense them or imply their endorsement.