Calmer bitcoin logs more three-sigma days in 2026 than in all of 2018

CoinDesk's tally is relative by design. Each day's price move was set against bitcoin's realized volatility over the previous 30 days, and any day that moved at least three times that amount, up or down, was counted ^[4]. A move that would have been ordinary in a turbulent 2018 month can qualify after a quiet one ^[4]. That construction lines up with the value-at-risk models CoinDesk describes, some of which rely heavily on recent price fluctuations, so a calm stretch makes an asset look less risky ^[9]. In our view the 10 days are close to a count of how often a risk band drawn from the prior month would have been broken at three sigma this year ^[1][4].

A bell curve keeps 99.7% of moves within three sigma ^[6]. The other 0.3% comes to about 1.1 days in a 365-day year ^[16]. On that basis 2018's eight such days were roughly seven times what a normal distribution predicts, and 2026's 10 so far are about nine times ^[19]. A model that scales its bad-day estimate to annualized volatility would have cut that estimate by about 45% between the two years ^[14], while the average three-sigma move fell about 30% ^[17].

The comparison across assets points the same way. Bitcoin's three-sigma count since 2024 is about 3.25 times Nvidia's at roughly the same 47% volatility ^[18]. A sizing rule that treats the two as equivalent on volatility alone gives them the same allowance for very different tail records ^[7].

"Bitcoin still goes through long quiet stretches followed by sharp repricings, and that hasn't changed. The market has matured, with more institutions, ETFs and much deeper liquidity, so the average day is calmer. But the shocks haven't gone away: macro, leverage, positioning," said Nicolas Quatravaux, head of EMEA at Paradigm, a liquidity network for crypto derivatives traders ^[8].

"Standard VaR measures do not properly assess the full tail risk, and this is one of the main reasons industry has been moving towards Expected Shortfall and similar measures, that do take tail risk into account," said Luuk Strijers, chief executive of the crypto options exchange Deribit ^[10]. VaR estimates a loss threshold without saying how severe losses get beyond it. Expected shortfall looks at how bad the worst days are ^[11].

The problem is smaller if most institutional books already run expected shortfall, the direction Strijers describes; the oversizing risk would then sit with simpler volatility-targeting setups ^[10][9]. It is also smaller if drawdowns keep coming in shallower. CryptoBriefing put this cycle's fall at roughly 53-54%, from about $126,200 in October 2025 to near $58,000 by mid-2026 ^[13], against the 73% bitcoin lost in 2018 ^[12].

We think the exposure is concentrated in books that size on a short window, because a 30-day window is what CoinDesk's count is scored against ^[4]. Holders who size on drawdown have a better cycle to point to ^[13]. CoinDesk's analysis does not include data on allocators' position sizes or flows, so its warning about oversized allocations rests on how the models are built ^[9]. Disclosures showing bitcoin position sizes held flat as 30-day volatility fell would undercut that warning.