Speculators net-short VIX futures is the textbook fragility gauge. Twenty-two years of positioning data shows it stayed quiet before the two unwinds that mattered most.
The cleanest explanation for why some volatility spikes turn violent is crowding. When too many traders are short volatility, a shock forces them all to cover at the same time, and that cover-buying becomes the spike. It is the mechanism behind the famous unwinds. So the obvious question is whether you can watch the crowd build before it breaks.
There is a free, public, weekly read on exactly that: the CFTC Commitments of Traders report, which breaks down how speculators are positioned in VIX futures. We pulled all 22 years of it, about 1,090 weekly readings back to 2004, measured how net-short speculators were as a share of open interest, and asked a simple question. When the short-vol crowd is heaviest, is the next spike bigger?
| Speculator positioning | Median VIX spike, next 8 wks | ...when VIX started below 20 |
|---|---|---|
| Most net-short | +45% | +54% |
| Middle | +36% | +40% |
| Least short (net-long) | +23% | +26% |
That is a real pattern, and it points the right way. When the short-vol crowd was heaviest, the spike that followed was roughly double the size of weeks when speculators were net-long, and the effect was even cleaner in calm tape (VIX below 20). The catch is that the relationship is loose: across all weeks the rank correlation is only about minus 0.23, a faint tilt rather than a reliable trigger.
A fragility gauge has one job, to light up before the violent unwinds. This one did not. Here is where speculator positioning ranked just before each famous blow-up, on a scale where 0 is the most net-short reading in history and 100 the most net-long.
| Just before the unwind | Short-vol crowding rank | Did it warn? |
|---|---|---|
| Volmageddon, Feb 2018 | 63rd percentile | No |
| COVID crash, Mar 2020 | 8th percentile | Yes |
| Yen carry unwind, Aug 2024 | 56th percentile | No |
Before Volmageddon and before the August 2024 unwind, speculators were not crowded short at all. They were actually a touch less short than their historical average. The only event the gauge flagged was March 2020, and that was an outside shock, a pandemic, not a positioning blow-up. On the two cases that were genuinely about a crowded short-vol or carry trade detonating, the VIX-futures read was silent.
This is the trap with positioning data: the crowding that matters is not always in the instrument you are watching. The right idea, that crowded shorts break violently, was pointed at the wrong meter.
The two episodes it whiffed on were not driven by speculators in VIX futures at all.
Speculators short VIX futures are only one of several crowded short-vol and carry trades at any moment, and in the violent episodes they were not the one that broke. The gauge is not measuring nothing. It is measuring the wrong slice of the crowd.
Twenty-two years of free, weekly positioning data, every reading counted, the spikes that followed measured against the crowd.
The most net-short fifth of weeks saw roughly double the forward spike of net-long weeks (+45% vs +23%). The direction is exactly the fragility thesis.
The rank correlation is only about minus 0.23. Useful as background color, far too noisy to act on as a standalone warning.
Speculators sat near their average, not crowded short, before both Volmageddon and the yen unwind. It only flagged 2020, an outside shock.
2018 lived in short-vol ETP exposure, 2024 in yen carry. VIX-futures positioning is one slice of the short-vol crowd, and not the slice that broke.
Worth stressing: the COT data itself is excellent, free, clean, and current to within a week. The problem is not the data. It is that VIX-futures positioning is the wrong window onto the crowd that actually detonates.
The idea was sound. The execution pointed at the wrong meter.
Crowded shorts really do break violently. But the crowd that breaks is not always in the contract you are watching. Before you trust a positioning gauge, ask whether the last few disasters even lived in that instrument.
"On average, more crowded means bigger spikes" can be true and still useless if the relationship is loose and goes quiet exactly when it counts. A fragility gauge is judged on the tail events, not the average week.
The two trades that did blow up are both trackable: short-vol ETP exposure can be rebuilt from shares-outstanding data, and yen carry crowding sits in the same free COT report, just under yen futures. Those are the windows worth opening next.
We test popular signals the honest way: every instance counted, every result measured against a plain baseline. See what else held up, and what did not.
Browse Reality Check
Comments
Loading comments…