What is modeled

Every currency has a real range (genuine Deribit tape, no reconstruction) and, for BTC only, a modeled range covering the historical gap. Modeled months are a bonus layer on top of the real archive, not a substitute for it.

Currency	Real coverage	Modeled coverage
BTC	2021-06-01 → 2024-10-01, 2026-04-21 → ongoing	2024-10-02 → 2026-04-20
ETH	2026-04-21 → ongoing	none
SOL	2026-04-27 → ongoing	none

Per-field provenance within a modeled window:

Field	Provenance
mark_iv anchors	Real Deribit trade prints — the only ground truth used.
volume	Computed from the trade tape (not modeled).
open_interest	NULL — unavailable for this window at any defensible price. Disclosed, not estimated.
bid / ask	Modeled spread (calibrated spread model), not observed quotes.
greeks	Deribit-native computation, applied to the modeled surface.

ETH/SOL historical samples (pre-live, sparse — not continuous coverage):

Currency	Range	Note
ETH	2024-01-01 → 2026-04-01	first-of-month days, 10-minute aggregated bars
SOL	2024-04-01 → 2026-04-01	first-of-month days, 10-minute aggregated bars

ETH/SOL historical samples cover first-of-month days only: 10-minute mean-aggregated bars derived from licensed third-party historical data, within the vendor's derived-data terms (≥10-minute aggregation). Field means are independent time-averages, so instantaneous ordering invariants (bid_iv < mark_iv < ask_iv) need not hold within a bar — disclosed, never clamped.

Method

The reconstruction runs in five steps: (1) recover trade-print anchors from the tape; (2) recover the forward via put-call parity; (3) invert Black-76 off that forward to get anchor IVs; (4) fit a global eSSVI surface with a calendar penalty, subject to the complete Gatheral–Jacquier no-arbitrage bounds; (5) apply a calibrated spread model to produce bid/ask around the fitted mark.

Mingone, A. (2022), "No arbitrage global parametrization for the eSSVI volatility surface", arXiv:2204.00312 — eSSVI parameterization and calendar-spread conditions. ^†
Gatheral, J. & Jacquier, A. (2014), "Arbitrage-free SVI volatility surfaces", Corollary 4.1 — the butterfly / no-arbitrage bound the fit is constrained to.
Alexander, C., Chen, D. & Imeraj, A. (2023), "Crypto quanto and inverse options", Mathematical Finance 33(4), 1005–1043 — the put-call duality underlying the forward recovery step. Deribit's index is non-tradable and settlement uses a 30-minute TWAP, so this inversion is approximate, not exact.

^† The Hendriks–Martini calendar conditions used here are sufficient (not the full Mingone necessary-and-sufficient set), per Pasquazzi (2023), "eSSVI Surface Calibration", arXiv:2304.02106.

Temporal interpolation between anchors carries no theorem claim — eSSVI's guarantees are static, per-snapshot. Every emitted snapshot is numerically re-verified against the no-arbitrage bounds after fitting, and the rejection / repair / carried-dropped counters below are the published record of what that re-verification found.

Accuracy

wing m1t0 median_abs_iv 0.0414 > 0.0400
wing m1t0 p90_abs_iv 0.1605 > 0.1000
wing m2t0 median_abs_iv 0.0938 > 0.0400
wing m2t0 p90_abs_iv 0.3386 > 0.1000
wing m2t1 median_abs_iv 0.0499 > 0.0400
wing m2t1 p90_abs_iv 0.2114 > 0.1000
wing m2t2 p90_abs_iv 0.1424 > 0.1000

Core (Δ 0.25–0.75, 7–90d) — blocks ship

Metric	Measured	95% CI	Bar	Status
band_containment	0.4001	[0.3427, 0.4457]	0.75	FAIL
median_abs_iv	0.0108	[0.0075, 0.0159]	0.015	PASS
p90_abs_iv	0.0358	[0.0248, 0.0505]	0.04	PASS
spread_width_median_rel_err	0.3933	[0.3383, 0.4461]	0.4	PASS
recall	1.0000	[1.0000, 1.0000]	0.95	PASS

Wing (everything outside core) — reported, never blocks ship

Metric	Measured	95% CI	Bar	Status
band_containment	0.3412	[0.3006, 0.3787]	0.5	FAIL
median_abs_iv	0.0272	[0.0220, 0.0341]	0.04	PASS
p90_abs_iv	0.1223	[0.0961, 0.1594]	0.1	FAIL
spread_width_median_rel_err	0.4671	[0.4217, 0.5116]	0.6	PASS

Wing 9-cell grid (mark_iv, moneyness × tenor)

Cell	median_abs_err	median CI	p90_abs_err	p90 CI	n_days	n_rows
m0t0	0.0217	[0.0193, 0.0242]	0.0832	[0.0701, 0.0962]	9	1470279
m0t1	0.0090	[0.0067, 0.0123]	0.0278	[0.0191, 0.0415]	9	376244
m0t2	0.0270	[0.0196, 0.0377]	0.0785	[0.0573, 0.1012]	9	330811
m1t0	0.0414	[0.0369, 0.0463]	0.1605	[0.1305, 0.2035]	9	640562
m1t1	0.0163	[0.0119, 0.0220]	0.0552	[0.0356, 0.0842]	9	1421400
m1t2	0.0280	[0.0216, 0.0367]	0.0782	[0.0584, 0.0993]	9	584716
m2t0	0.0938	[0.0499, 0.1504]	0.3386	[0.1289, 0.6536]	8	59396
m2t1	0.0499	[0.0364, 0.0683]	0.2114	[0.1572, 0.2801]	9	1424130
m2t2	0.0379	[0.0309, 0.0456]	0.1424	[0.1127, 0.1788]	9	1781720

Worst days (by core median |IV error|)

Day	Core median \|IV err\|	Wing median \|IV err\|	Recall
2024-12-01	0.0297	0.0505	1.00
2025-02-01	0.0126	0.0338	1.00
2025-06-01	0.0103	0.0280	1.00
2026-02-01	0.0085	0.0213	1.00
2025-08-01	0.0079	0.0258	1.00

Engine counters (summed over scored days)

snapshots_emitted: 12900
snapshots_rejected: 20
snapshots_repaired: 18
anchors_used: 25827
carried_dropped: 71

Every in-gap oracle day scored above is a post-expiry-roll 1st-of-month snapshot. Mid-month accuracy is not measured by this protocol — treat it as an extrapolation from these day-boundary results, not an equivalent guarantee.

open_interest is unavailable for the entire modeled window at any defensible price and is never estimated or backfilled. It is returned as null, not zero.

Access

Modeled files are opt-in: pass ?include_modeled=true to GET /v1/files/{ccy}/bulk — the only endpoint that accepts this parameter — to include modeled days in the bulk manifest (each such entry is flagged modeled: true). By default the manifest lists real coverage only.

On GET /v1/chains/{ccy}/snapshots and GET /v1/chains/{ccy}/instrument/{symbol}, ?source= filters the response body by the exact provenance tag carried in each row's source column — for example modeled_surface, live_ws, or live_rest. An unknown tag returns 400 invalid_source; there is no real/modeled shorthand.

Modeled months are not counted in coverage claims and carry no charge — they are a bonus layer on top of the real archive, priced at zero, disclosed everywhere they're served.

How the gap is reconstructed