Smooth Is Not the Same as Quiet

~14 min read. Part two asks why. Part three asks what the asset is.

Picture two flights.

The first is bumpy the entire way. Constant light chop, drinks sliding across the tray table, nobody sleeps. Four hours of low-grade misery. But nothing actually happens.

The second is glass-smooth for four hours — and then drops two hundred feet without warning, and everyone who unbuckled is now acquainted with the overhead locker.

Ask a statistician which flight was more turbulent and they'll say the first. Ask a passenger which one they'd rather not repeat, and they'll say the second.

Hold on to that, because it's the whole story of what institutional money has done to Bitcoin.

The two numbers nobody puts side by side

Bitcoin has calmed down. Its volatility is roughly a fifth lower than at the start of 2022, and it has fallen in every period since — no reversals, no false dawns. This is the number everyone quotes, and it's true.

Here's the one nobody quotes. The sudden-drop portion of that volatility is 71% higher than in the year the spot ETFs launched.

Same asset. Same stretch of time. Both facts sitting in the same dataset, pointing in opposite directions.

They're not contradictory. They describe two completely different things that happened for two completely different reasons — and squashing them into a single "volatility" number throws away the only part that matters.

Short version, before the evidence: the spot ETF stopped Bitcoin breaking. The options market then stopped it fidgeting — and handed the drops back through a different door. Two fixes, one relapse.

Three questions, not one

Whenever anyone says "volatility," they're usually blurring three questions together.

How big is a normal day? The standard measure. Useful, and on its own almost meaningless.

Does the movement arrive as chop, or as drops? A market can be volatile because it moves 2% most days, or because it does nothing for a month and then moves 20%. Identical averages, completely different experiences — and only one of them takes out your stop loss. Statisticians split total movement into a diffusion part, the ordinary continuous churn, and a jump part, the discontinuities.

And does the turbulence itself come and go? Long calm stretches broken by chaos is a different beast from steady, predictable bumpiness, even when the averages match. That one is measured by vol of vol — how much the volatility reading itself moves around.

The table below uses the proper names, with definitions underneath. The concepts don't survive being translated into everyday words without losing something, so it's worth two minutes with the footnotes.

Split Bitcoin's history into three eras — before the spot ETFs, the ETF's first year, and the period since the options and futures complex built out around it — and all three questions get different answers.

Most cells carry two numbers. The first covers every trading day. The second, in brackets, strips out the two events you'd expect to be responsible for the whole picture: the four days of the FTX collapse in November 2022, and a single day, 10 October 2025.

That second one is worth a moment. The China tariff announcement sent Bitcoin from an intraday high of $122,574 to a low of $104,782 — down 14.5% — and forced the closure of roughly $19 billion of leveraged positions across crypto. It remains the largest single-day liquidation the market has seen.

More on those shortly. They don't do what you'd think.

Source: Bloomberg Finance L.P.; Pandemonium calculations.

Main figures cover every trading day. Bracketed figures exclude the two big crash events — the four days of the FTX collapse (8–11 November 2022) and 10 October 2025 — to test whether those alone explain the pattern. The middle column has no brackets because neither event falls inside it.

What each row means

• Annualised volatility — the standard measure of how much the price moves over a year. The number everyone quotes.
• Jump volatility — the portion of that movement arriving as sudden discontinuities. The air pockets.
• Diffusion volatility — the portion arriving as ordinary continuous trading. The chop.
• Jump-to-total variation ratio — what share of all the movement came from jumps rather than ordinary trading.
• Top 1% variance share — what share of all the movement happened on the wildest 1% of days. In a 739-day stretch, that's about seven days.
• Excess kurtosis — how much more extreme the worst days are than a well-behaved market would produce. Not a percentage. Zero is the benchmark; higher means the outliers tower further above the ordinary days.
• Vol of vol — how much the volatility reading itself moves around. High means calm stretches punctuated by chaotic ones; low means steady, predictable bumpiness.

One thing not to do with these numbers: jump and diffusion volatility do not add up to total volatility. Volatilities never add — variances do. Before the ETF: √(32.5² + 46.0²) ≈ 56.3% — the 56.4% in the table, give or take rounding. Not 78.5%. The two components are pieces of the same whole, combined the way the sides of a right-angled triangle give the hypotenuse.

Read the second row and you get the comfortable story: things are improving, Bitcoin is growing up, next question.

Read the two rows beneath it and the comfortable story falls apart.

What the ETF actually did

Between the pre-ETF years and the ETF's first year, overall volatility fell about 11%.

Every bit of that came from the drops, which collapsed by 60%. The everyday chop didn't fall at all. It went up — to the highest level anywhere in the dataset.

Which means the ETF year, the one everyone remembers as the calm one, had the most turbulent ordinary days Bitcoin has ever had. What it didn't have was the floor giving way.

That's flight number one. Bumpy the whole way, and nothing broke.

And it's exactly what you'd expect once you look at who the ETF brought in. These are buyers placing steady, unlevered, size-limited orders. They're not borrowing to buy. They can't be margin-called. When the price falls they aren't forced to sell into it, because nobody is chasing them for collateral.

Which is the opposite of what came before. The old market ran on borrowed money. A price fall triggered margin calls, the margin calls forced sales, the forced sales pushed the price lower, and the lower price triggered the next round of margin calls. That loop is what people mean by a cascade, and it is the reason a bad afternoon in crypto used to turn into a catastrophic one. Buyers who haven't borrowed anything cannot start that loop, and there were suddenly a great many of them.

Buyers like that don't make a market move less. They stop it breaking. Not the same thing — and the ETF gets far too little credit for the second one. Wiping out six-tenths of Bitcoin's cascade risk is the single biggest change in the table.

Then the options market showed up

From the ETF's first year to now, everything flips.

The everyday chop fell hard, down about 19%. The sudden drops rose 71%. And excess kurtosis rose roughly fivefold, from 0.74 to 3.68 — the worst days pulled much further away from the ordinary ones.

Also not mysterious, once you see who is on the other side of all those options.

Dealers hedge their option books by trading Bitcoin itself, and the hedge has to be re-adjusted every time the price moves. Which way they adjust depends entirely on whether they are net buyers or net sellers of the options.

A dealer who owns options trades against the move. As the price rises his hedge needs less Bitcoin, so he sells a little; as it falls he buys a little. Thousands of small trades, all leaning back against wherever the market is trying to go. That is a cushion.

A dealer who has sold options does the exact opposite. Now a rising price forces him to buy and a falling price forces him to sell — chasing the market rather than resisting it. That is an accelerant.

Which one dominates depends on the customer flow, and crypto's has been unusually one-sided. Miners, corporate treasuries and yield products have been persistent sellers of calls and puts, harvesting premium. Somebody has to buy what they are selling, and that somebody is the dealer. For much of this period, then, dealers have been sitting on the cushion side of the trade — which is a large part of why ordinary days have got quieter.

But a cushion built this way has two weaknesses.

It only exists near the strike prices where the options actually sit. Move far enough away from them and there is nothing there — the market walks off the edge of its own support.

And it vanishes on expiry, when a very large block of contracts stops existing on the same scheduled day. The hedging that went with them unwinds at once, and the thing that had been absorbing moves simply isn't there any more.

Then there is the third case, which is the dangerous one. Premium-selling dries up in a frightening tape, and buyers of protection take over. The dealers end up short the options instead of long them, and the same hedging machinery that was absorbing moves starts amplifying them: the price falls, the dealer must sell to stay hedged, the selling pushes it lower, and the next adjustment is larger than the last.

Calmer Tuesdays. Livelier Fridays — and the last Friday of the quarter most of all, when something like 40% of all open contracts stop existing in a single morning.

I'd offer that as the best available explanation rather than a proven one. The data shows the drops came back; it doesn't say why. That same stretch also contains a 54% peak-to-trough drawdown and a change of Federal Reserve chair. Either leaves similar fingerprints. The way to settle it is to date every drop and check whether they cluster around expiries.

"Isn't this just FTX and last October?"

The obvious objection. Two famous disasters, one in each era, doing all the work. Take them out and the whole thing evaporates.

It doesn't. Take them out and every measure of extreme risk gets worse.

Not by a rounding error, either. Strip the FTX days and pre-ETF excess kurtosis rises by a quarter, from 6.11 to 7.71 — the era looks more extreme without its most famous crash, not less. Strip last October and today's drop share climbs from 23.8% to 25.8%.

This sounds like a glitch. It isn't, and the reason is the actual finding.

Those crash days were enormous in absolute terms — but not enormous relative to how wild those periods already were. They were bad days in a market full of bad days. Removing them tightens up the ordinary middle of the distribution without touching the genuine outliers, so the outliers stand out more against a calmer backdrop.

And the single most telling number in the table: before the ETF, the share of movement arriving as drops barely moves when FTX is removed — 33.3% to 33.4%. That era's cliff-edge behaviour was never about FTX. It was spread across three solid years. It was just how the market worked.

One number that looks like a typo

Remove the largest single-day liquidation event in crypto's history and today's overall volatility goes up, from 45.2% to 45.5%.

That should be impossible. Delete the worst day, get a calmer number.

Except the method that separates drops from chop sets its own bar for what counts as a drop, and it sets that bar using the data. One gigantic move raises the bar for everything else — and hides the smaller drops behind it. Take it out, the bar falls, and a batch of genuine drops that had been sitting quietly underneath get counted properly.

Which cuts in a direction worth noticing. The headline numbers understate how jumpy today's market is, not the reverse.

The column I wouldn't lean on

The middle one survives this test for an unimpressive reason — nothing was removed from it, because nothing happened in it. That isn't robustness. It's an untaken exam.

It's also only 295 days against 739 and 679 for the others, with no blow-up, no cascade and no macro shock anywhere in it. You can't measure how often disasters happen in a window that contains none. Some of that suspiciously well-behaved column is real and some of it is a short, lucky stretch, and I wouldn't hang anything on the exact figures.

What survives regardless: the ordering across the three eras holds on every single measure, with or without the crashes.

What this leaves open

Two things changed, and they aren't the same thing.

Ordinary days got calmer. More participants, deeper markets, proper hedging tools. That's real maturing, and it's what everyone predicted.

The drops didn't go away. They collapsed under the spot ETF, then came back.

Why they came back is the honest gap in this analysis. The options market is the most likely explanation. But the worst day of the current era was October 2025, when $19 billion of borrowed positions were force-closed in a matter of hours. That is the old cascade, not the expiry-driven kind described above. So the machinery behind the gaps may have changed, or the old machinery may simply still be running.

What hasn't changed is the part that matters. When the price falls hard, nothing steps in to catch it.

Which leaves a question this data can't answer.

Bitcoin now has the deepest, most professional, most heavily intermediated set of buyers it has ever had. Those buyers demonstrably smooth out ordinary trading. So why do they do so little on the days that actually hurt?

The answer isn't in the volatility numbers. It's in who those buyers are — and, more to the point, whether anything on earth obliges them to show up when the price is falling.

That's part two.

Part two: why nobody is obliged to own Bitcoin, and what that does to the money.