When Is The Crypto Front Week Actually Cheap?

Every options trader has felt the tilt. That particular shape where the front week prices above the quarter and the market stops pretending risk is spread evenly across the horizon. Textbooks call it backwardation; desks call it a warning. You buy the front week if you must and you do not sell it lightly.

We measured that instinct across twelve months and four coins and the curve turns out to say very little about it. Across 1,396 coin-days the front week was worth its price only 31.2% of the time and the shape of the curve barely moved that number: 28.8% with the curve on its inverted side, 34.5% on its contango side. The seller was paid on roughly seven days in ten, whatever the curve looked like. Traders already fluent in term structure and realized volatility can jump straight to "The shape barely moves the odds".

Updated 23 September 2026: extended to twelve months of Deribit data and measured at a fixed seven-day tenor. The figures below are this edition's.

In this note we will first show what the curve is and how a clean seven-day price is read off it. Then we ask one question of every day in the archive, sorted by the shape of the curve. Finally we give the shape two chances to prove itself. It takes neither.

First, what is a term structure?

An option expiring on Friday and one expiring in three months rarely carry the same volatility price. Plot the at-the-money implied volatility of every listed expiry against its time to expiry and you get the term structure, the market's price list for uncertainty at different horizons.

Figure 1: Three real Bitcoin days. Each dot is a listed expiry; each ring is the fixed 7-day or 60-day point interpolated between them. Rising left to right is contango, the normal state; falling is backwardation.

Most days it slopes up. On 15 August Bitcoin charged 26.2% for the next seven days and 36.2% for sixty, the normal state, called contango. The far dates cost more because more can happen in two months than in one week.

Sometimes it flips. On 6 February the same book charged 73.9% for the week ahead and 56.9% for sixty days. That is backwardation: the market saying the danger sits in the days directly in front of you.

There is a practical detail here that decides every number below. Listed expiries almost never sit exactly seven days out. Deribit's nearest weekly can be hours away or nearly a week away. An option quoted two days from expiry is not a seven-day option. So we read a fixed seven-day tenor, $T = 7$ days, by interpolating in total variance between the two listed expiries either side of it:

$$\sigma_{T}^{2}\, T = \sigma_{1}^{2}\, T_{1} + \frac{T - T_{1}}{T_{2} - T_{1}} \left( \sigma_{2}^{2}\, T_{2} - \sigma_{1}^{2}\, T_{1} \right)$$

where $\sigma_1$ and $\sigma_2$ are the at-the-money implied volatilities of the listed expiries at $T_1 < T < T_2$ days and $\sigma_T$ is the fixed-tenor implied we use. The same rule gives the 60-day point. A day with no expiry on one side of the target is dropped rather than extrapolated.

That is what the shape is. What it means is where the field splits. Amberdata, working through four years of Bitcoin, found the curve in contango almost 77.5% of the time and treats the flip as a reaction to shocks. Brett Friedman (2025), writing for OptionMetrics, found the VIX futures curve backwardated on 18.8% of days since 2006, concentrated in the financial crisis and the pandemic. An experienced options trader once caught one forming ahead of the ETF deadline, before anything had happened at all. Three readings of the same shape and the measurement that follows adds a fourth.

What the market charged against what it got?

Two numbers run through everything below and they are easy to mix up.

Implied volatility is a price, what the market charges today for the week ahead. Realized volatility is a measurement, how much the coin actually moved over that same week, computed after the fact from Deribit's index every ten minutes:

$$\text{RV}_{t \to t+7} = \sqrt{\frac{365}{7} \sum_{i=1}^{n} \left(\ln \frac{S_i}{S_{i-1}}\right)^2}$$

where $S_i$ is the index price at the $i$-th ten-minute mark after the reading and $n$ is the number of ten-minute returns in the seven days that follow, 1,008 when none are missing. Crypto annualizes on 365 days because the market never closes.

Figure 2: Top: the price of the next seven days against what those seven days then delivered. Bottom: the 60-day implied minus the 7-day, red wherever the curve is inverted.

One is the bill, the other is what you got for it. The whole note is one question asked 1,396 times: was the bill worth paying? A day counts as worth its price when the volatility that showed up over the next seven days came in above the implied the option was carrying that morning. Look at the top panel and the answer is already visible: the coral line sits above the silver one far more often than below it and the silver line only wins in bursts.

Every coin is ranked against its own history

Raw volatility points do not travel between coins. XRP's curve swings far wider than Bitcoin's, so a spread that marks a crisis on one book is an ordinary Tuesday on the other. Each day is therefore placed by rank within its own coin's distribution.

Figure 3: Bitcoin's 369 daily term spreads. The deciles used below are cut from this distribution, per coin, which is why they land on a different number for every book.

Bitcoin's tenth percentile falls at -4.34 points. XRP's falls at -10.80. Same rank, different number. That is the point of ranking at all. The rank itself is simple:

$$r_{c}(t) = \frac{1 + \#\{\,s \in D_c : \text{spread}_c(s) < \text{spread}_c(t)\,\}}{|D_c|}$$

where $D_c$ is the set of sampled days for coin $c$ and $\text{spread}_c$ is its 60-day minus 7-day implied volatility. A rank of 0.1 or below puts a day in the most inverted decile, above 0.9 in the steepest contango.

The shape barely moves the odds

Here is every day in the archive, sorted into deciles by where its curve sat, against how often the front week turned out to be worth its price.

Figure 4: Share of days the next seven days delivered more volatility than the 7-day option charged, by term-structure decile. No decile reaches a coin flip.

Term-structure decile	Days	Front week worth its price	Median realized / implied
1, most inverted	136	27.9%	0.858
5, the weakest	140	24.3%	0.903
10, steepest contango	140	39.3%	0.904
Inverted side, deciles 1 to 3	416	28.8%	0.891
Contango side, deciles 8 to 10	420	34.5%	0.927
All days	1,396	31.2%	0.911

Two things come out of that table and the first is the one to keep. The front week was dear on almost every kind of day. The median week delivered about 91% of the volatility the option charged and no decile, inverted or calm, came near a coin flip. Amberdata's four-year Bitcoin study found 30-day implied above 30-day realized nearly 70% of the time; on seven-day options across four books we measure 68.8%, which is the same premium seen through a shorter window.

The second is how little the shape adds. From the most inverted third of days to the calmest third, the share moves by 5.7 percentage points, from 28.8% to 34.5%. The deciles in between wander up and down. They do not climb.

The warning the curve gives is real but it is a warning about the weather. Inverted days are the volatile days: the median week after one delivered 63.1% realized volatility, against 43.6% after a day in contango. The market knew and charged for it: 72.0% against 45.9% at the median. That is what returns the same verdict on both kinds of day. The front week was dear on the calm days and dear on the rough ones.

A sailor would recognise the distinction. The wind tells you how hard the boat will heel; the tide tells you whether you make the harbour. The curve is the wind. Whether the front week pays its buyer is the tide. It ran against the buyer in every decile.

Two checks, neither of which rescues the shape

A gap of a few points on overlapping data invites two questions: is it the price level in disguise and is it bigger than chance?

The first is the price level. Inverted days carry high front implied volatility (the two correlate at -0.67 in this sample) and the premium is known to vary with the level of implied. So the same comparison runs inside thirds of implied level, like with like.

Figure 5: Top: the measure within each band of implied level, flat in all three. Bottom: the contango-minus-inverted gap on seven independent subsamples, split three against four.

Held at a comparable price, the shape changes almost nothing. Same verdict. Among mid-implied days the front week was worth its price 33.3% of the time with the curve inverted, 28.2% in the middle and 34.4% in contango. The other two bands are missing a corner for a reason: low implied almost never comes with an inverted curve and high implied almost never with contango.

The second is chance and overlap makes it the harder question. A seven-day forward window on daily readings means two consecutive rows share six days of outcome, so 1,396 rows are nothing like 1,396 independent draws. Taking every seventh day gives seven disjoint subsamples. Three lean toward contango and four the other way, with a median gap of -1.7 percentage points. Resampling whole calendar weeks, 55 of them, puts a 90% interval on the pooled gap from -7.5 to +18.3 percentage points, what includes zero comfortably. We cannot tell the gap apart from chance.

Bitcoin and ETH lean one way, by under 10 points

Figure 6: The same measure per coin. Bitcoin and ETH lean toward contango by just under 10 percentage points; XRP by less; SOL the other way.

Coin	Curve inverted	Curve in contango	Difference, percentage points
BTC	26.4%	36.0%	+9.7
ETH	32.7%	42.3%	+9.6
XRP	24.5%	29.3%	+4.8
SOL	31.6%	29.3%	-2.3

The two deepest books lean the same way and by the same amount: Bitcoin and ETH both find the front week worth its price about 10 percentage points more often in contango. Interesting. Not yet a finding. Each state holds roughly 110 days per coin, drawn from a handful of long inverted episodes and that is a lean worth watching. XRP leans by half as much and SOL runs slightly the other way. Not one coin reaches a coin flip in either state.

How we measured it?

Every figure here is built from Deribit's option marks and index as Cayø Largo records them every ten minutes, one reading per coin per day at 10:50 UTC, from 5 September 2025 to 16 September 2026: BTC and ETH on 369 days each, SOL and XRP on 329. The per-expiry implied volatilities are the same quotes served by the volatility surface endpoint and the realized leg reads the index served by the spot rates endpoint.

At-the-money means the options at the two strikes nearest each expiry's own forward price, calls and puts both, so normally four quotes averaged. Expiries under half a day are excluded, because a mark hours from settlement prices the last hours' pinning rather than a view on volatility. The fixed 7-day and 60-day tenors are interpolated as in the formula above and never extrapolated.

The realized leg runs on Deribit's index at ten-minute cadence. Isolated single prints that sit far from the surrounding hour are removed before any return is taken, judged against the local median and the local spread rather than a fixed threshold, so a move the whole hour shares is kept and a print its neighbours disagree with is not. A forward week needs 90% of its ten-minute returns present or the day is skipped.

The reading is pinned to 10:50 UTC because an unpinned daily sample confuses the intraday volatility rhythm for a daily one and 10:50 sits clear of Deribit's 08:00 UTC settlement, when the front of the curve rolls to a new contract. The same curve is drawn live, every ten minutes, on the Term Structure page. Every headline figure is a share or a median rather than a mean, because one extreme week moves an average and barely moves a rank.

What this does not say?

Like any measurement taken from a single archive, this one rests on assumptions and has its limits and four of them deserve a sentence or two each.

The first is time. The archive covers twelve months and one market cycle: Bitcoin ran from a peak near 126,000 USD in October 2025 down to 58,000 USD in June and July 2026, then back above 86,000 USD by September. These are rates measured inside that stretch, not a law of crypto volatility.

The second is what "worth its price" means. The comparison is realized against implied, in volatility points. A delta-hedged straddle pays roughly in proportion to that gap but it also pays the spread, the hedging cost and the path. None of that is modelled here. What the premium actually paid a seller and when is the subject of our note on when to sell crypto volatility.

The third is the lean on Bitcoin and ETH. Ten percentage points on two books is not nothing but on twelve months of overlapping windows it cannot be separated from chance. Another year of archive would settle it one way or the other.

The fourth is coverage. AVAX and TRX are left out, because their options history in the archive starts in January 2026, too short to rank into deciles alongside the others.

What a desk would watch?

The useful question when the curve inverts is not how loudly it is warning. It is what the front week costs against what follows. Over twelve months of Cayø Largo's archive the answer barely depends on the shape: the seven-day option charged more than the week delivered on about seven days in ten, inverted or calm. The shape tells you what kind of week the market expects. It does not tell you the week is cheap. For how long inversions persist once they form, see how long crypto backwardation lasts; for the implied-against-realized frame this note assumes, see the variance risk premium guide.

References

• 2023-01-24, Amberdata, Bitcoin Options: Finding edge in four years of volatility regimes. In the ATM term structure section, measures contango at almost 77.5% of the time across four years and describes backwardation as a reaction to volatility shocks. Later in the same piece, finds 30-day BTC implied above 30-day realized nearly 70% of the time since April 2019.
• 2024-01-10, Imran Lakha, Beware The BTC Vol Reset. Records Bitcoin's term structure in backwardation ahead of the spot ETF deadline, with weekly options resisting decay until after the event: an inversion formed in anticipation rather than in reaction.
• 2025-03-20, Brett Friedman, OptionMetrics, VIX Futures Term Structure: A Warning Sign?. Finds the VIX futures curve backwardated on 18.8% of daily observations since February 2006, concentrated in the financial crisis and the pandemic.