Reading the term structure around expiries

Across four Deribit expiry cycles in BTC and ETH between July and October 2026, the 30-day leg of the term structure is close to static while the front end accounts for almost all of the movement. This note follows one cycle, the 25 September 2026 quarterly, at five-minute resolution from two weeks before settlement through the 08:00 UTC fix.

Empirical observations from VolaLabs surface data. Not investment advice or a recommendation to transact.

Summary: the 30-day at-the-money vol holds a narrow band through expiry week. The 3-day rises above it roughly four days before settlement, then declines by 16 vol points in BTC and 22 in ETH into the 08:00 fix.

01Front-end behaviour into settlement

Over the fortnight shown, the 30-day at-the-money vol is close to flat: 34% to 38% in BTC and 48% to 52% in ETH. On this evidence the back of the curve does not respond to the approaching expiry.

The 3-day behaves differently. It enters the period above its own recent range, firms through the run-up, and crosses above the 30-day around T−4, at 40% in BTC and 55% in ETH. The curve is inverted, front above back, for one to two sessions. Into the 08:00 settlement the 3-day declines to approximately 24% in BTC and 33% in ETH, 16 and 22 vol points below the T−4 peak and below the 30-day band it had crossed.

Expressed as the 3d−30d slope, the series starts near zero, peaks at +9 vol points in BTC and +13 in ETH at T−4, and reaches its low for the fortnight at settlement, at −14 and −20 vol points. Magnitudes vary by cycle. In one of the four, the ETH front rose some 21 points above its baseline. In the most recent, the pre-expiry hump was negligible.

02Sampling frequency and the daily mark

At five-minute resolution the 3-day series moves materially within sessions. Intraday ranges of 9 to 15 vol points are typical during expiry week, against 5 to 6 points at baseline. The ETH 3-day covered 15 vol points on T−3 and 9 to 11 points on settlement day.

Sampling the same series once a day at a fixed time produces a considerably smoother path. Each daily observation sits close to the local trend, so the daily series understates the intraday range by a factor of two to three through expiry week. Because the quarterly settles at 08:00 UTC, a mark taken at that time is recorded during the decline itself, and the final observation is not representative of the levels available over the preceding session.

03Implications

Two features should be treated separately. The pre-expiry inversion peaks around T−4 and is largely unwound before settlement, and it appears in three of the four cycles in the sample. The settlement-day decline appears in all four and is the more consistent of the two. The 30-day leg does not participate in either. For marking and risk measurement, the implication is that a single daily observation taken at the settlement time is drawn from the least representative point of the cycle.

Data and method. One cycle is shown in full because averaging across cycles removes the timing, which is the subject of the note. The sample is four Deribit cycles between 26 July and 30 October 2026: the 31 July, 28 August and 30 October monthlies and the 25 September quarterly, at five-minute resolution and at-the-money only, where the data is most robust. Both figures plot BTC and are drawn between measured anchor points, so the shape and the labelled levels are the reliable content rather than any individual point on the line. Four cycles is sufficient to identify a pattern and test its consistency. It is not sufficient to separate quarterly from monthly behaviour.