Sources, definitions and known limitations

This page is the contract behind every number on the dashboard. All sources are public and require no API keys, so anyone can reproduce the dataset from the open-source pipeline. Where a metric involves a judgment call, the call and its consequences are stated here.

Scope

The observatory tracks perpetual futures and dated futures on BTC, ETH and SOL, plus an “everything listed” series covering every derivative contract on the tracked venues — the altcoin tail and RWA, equity-index and commodity contracts included. Options are excluded in this version. Onshoremeans a CFTC-registered designated contract market: CME, Coinbase Derivatives Exchange (CDE), Kalshi, and Kraken’s US exchange — Bitnomial, the CFTC DCM/DCO acquired by Kraken’s parent in April 2026, whose perpetual contracts trade through Kraken Pro. Offshorecovers the major venues serving the global market outside US regulation: Binance, Bybit, OKX, Deribit, Hyperliquid, and Kraken Futures (Crypto Facilities, the separate non-US Kraken entity closed to US persons — distinct from Kraken’s onshore exchange above). Cboe wound down its digital-asset futures business and is not tracked. In CFTC reports, “LMX Labs, LLC” is Coinbase Derivatives’ former legal name; both names are merged into CDE.

Per-venue sources

Binance and Bybit geo-block US IP addresses. Live reads are made directly from the deployed application, which runs in an EU region; historical Binance data comes from Binance’s public archive CDN, which is not geo-restricted. No credentialed or private access is used anywhere.

Metric definitions

Venue-wide totals (“everything listed”). Volume sums every listed contract per venue from bulk public endpoints (Binance/Bybit/OKX/Kraken tickers, Deribit book summaries, Hyperliquid asset contexts, the Coinbase/Kalshi/Bitnomial catalogs, CME’s crypto suite on the delayed feed). Open interest likewise, with one exception: Binance has no bulk OI endpoint, so its venue-wide OI comes from CoinGecko’s Binance Futures page — the only non-direct source in the observatory. Onshore venue-wide OI adds the CFTC-reported XRP and BCH contracts to the majors. Venue-wide volume historyis reconstructed contract-by-contract from each venue’s daily candle feeds across its current catalog (instruments delisted during the year cannot be enumerated, so older months undercount modestly); venue-wide OI history cannot be reconstructed this way and accumulates from July 3, 2026.

Onshore share of volume. Onshore ÷ (onshore + offshore) daily USD notional, summed across tracked contracts per asset, presented as a 7-day trailing average. Offshore venue volume is the venue-reported 24-hour notional for the tracked perp (all listed futures for Deribit); CME volume is front-month standard + micro contracts × contract size × settlement price; CDE volume sums every listed contract × contract size × price; Kalshi is venue-reported notional.

Onshore share of open interest.Onshore OI uses the CFTC’s weekly Commitments of Traders reports (all CME and CDE crypto contracts, dollarized at the spot close on the report date, forward-filled between reports) plus Kalshi’s daily reported OI. The offshore denominator is a fixed set — Binance, Bybit and OKX, the three largest — because Deribit and Hyperliquid publish no OI history and letting venues drift in and out of the denominator would fabricate share moves. Their OI is still collected daily and reported in the dataset; the headline share is modestly overstated as a result (both are small relative to the trio).

Funding rates.All funding is normalized to an 8-hour-equivalent rate, then annualized (×3 ×365) for display. Daily values are the mean of the day’s funding periods. Each composite is an equal-weight mean of its group’s available venues — offshore: Binance, Bybit, OKX, Deribit, Hyperliquid, Kraken Futures; onshore: CDE, Kalshi, Kraken US. Hourly funders (Hyperliquid, CDE) are scaled by interval; 8-hour funders are used as published. Composites are displayed as 7-day trailing averages (a single 8-hour print, annualized, would otherwise dominate the chart); per-venue daily series remain in the raw dataset and CSVs.

CDE funding reconstruction.Coinbase Derivatives publishes no funding-rate history, but it publishes the mechanism: the hourly rate is the hour’s average perp–spot premium scaled down by 24, EMA-smoothed. Pre-July-2026 CDE funding is therefore reconstructed from hourly candle closes — hourly premium = (perp close − spot close) ÷ spot close, daily mean ÷ 3 = the 8-hour-equivalent rate — sampling hourly where Coinbase samples every three minutes, with prints more than 1% from spot discarded as stale. Validated against observed funding on overlapping days (same order of magnitude and sign). These rows carry source premium-reconstruction in the dataset and are estimates, not exchange-published rates; funding observed live at ingest always takes precedence.

CME basis. (Front-month close − spot close) ÷ spot, annualized by days to expiry, where expiry is approximated as the last Friday of the contract month and the series rolls to the next month inside ten days of expiry (annualizing over a near-zero window would amplify close-timing noise). Spot reference is Coinbase Exchange; displayed as a 7-day average. This is comparable to annualized perp funding as a cost-of-carry measure.

Execution cost.From full order-book snapshots taken at ingest: one-way cost versus mid, in basis points, of a simulated market order of $100K and $1M notional walking the book, plus quoted spread and resting depth within ±50 bps of mid. The dashboard presents each group’s bestbook (the venue a cost-sensitive trader would route to) and the group’s summed depth; per-venue snapshots are in the raw dataset. Contract-unit books (CDE, Kalshi) are converted to implied coin prices via contract size. The headline dollar figure is the best book’s $1M BTC fill cost converted from basis points (1 bp on $1M = $100).

Headline summary.The “What the data shows” box is generated from the data on every build. The share comparison uses the 7-day-average onshore share today versus 90 days earlier; “genuine migration” means onshore 7-day-average volume rose over that window, as opposed to the share rising only because offshore volume fell faster. The new-entrant share is Kalshi plus Kraken US volume as a fraction of onshore volume over the trailing 30 days. The US funding premium is the onshore composite minus the offshore composite (7-day means, annualized): positive means a perp long pays more to hold onshore.

The listing gap

Each venue’s live contract catalog is enumerated daily (Binance exchangeInfo, Bybit/OKX/Kraken tickers, Deribit book summaries, Hyperliquid asset metadata, the Coinbase, Kalshi and Bitnomial product catalogs, and a curated CME list). Symbols are normalized to one underlying per asset (1000PEPE/kPEPE→PEPE, XBT→BTC) and classified using Binance’s own listing metadata — underlying type COIN, EQUITY, KR_EQUITY, PREMARKET, COMMODITY or INDEX — with curated fallbacks for what Binance doesn’t list (Kraken’s X-suffixed tokenized equities, FX pairs, CDE’s commodity and sector-index roots). An underlying counts as “missing onshore” when no CFTC-regulated venue lists any derivative on it. Classification of symbols listed nowhere on Binance falls back to heuristics and defaults to crypto; the full inventory is exported for audit.

Known limitations

• Coinbase Derivatives volume history comes from daily candles for every currently-listed contract (the perp-style contract, most of CDE volume, is complete from its July 2025 listing). Coinbase stops serving a contract’s candles once it expires, so dated contracts that expired during the past year are missing from older months — a modest understatement. CDE funding has no public history and is observed from July 2026. Its open interest history is complete via CFTC reports.
• CME volume counts front-month standard and micro contracts only (back months excluded), and the public delayed feed carries no history for SOL futures — CME SOL volume accumulates from July 2026. CME options are excluded entirely.
• Offshore volume tracks the dominant USDT-margined perp per venue (plus Deribit’s full futures complex); coin-margined and exotic-margin contracts are excluded, which understates offshore totals somewhat — a conservative bias for the onshore-share thesis.
• Offshore OI history retention (verified against each API): Bybit serves a year+, OKX 180 days, Binance only 30 days — the binding constraint on the OI-share window, which grows one day per day as the pipeline accumulates snapshots. Deribit exposes no OI-history endpoint, and Hyperliquid’s only historical OI archive is a requester-pays S3 bucket, i.e. not freely public.
• Order-book snapshots are point-in-time, once daily; they are indicative of relative depth, not a volume-weighted average. Hyperliquid’s API returns 20 aggregated levels, which can understate its depth at the $1M tier. CME depth requires a paid feed and Kraken US publishes no free depth feed, so the onshore group’s books are CDE and Kalshi only.
• Kraken US (Bitnomial) publishes prices in scaled exchange points, so its open interest is dollarized with the observatory’s spot reference; its volume and OI history begin at observatory launch, and it is below the CFTC’s weekly reporting threshold as of mid-2026.
• The 24-hour volumes read at ingest are rolling windows attributed to the UTC ingest date; venue UTC-day candles are used for history where available, so a small boundary mismatch exists on the most recent day.

Citing this data

Every chart has a CSV download and every underlying dataset is a flat JSON file under /data. Cite as “US Crypto Derivatives Market Structure Observatory” with the dataset date. The pipeline and site are open source under the MIT license.