Bitcoin’s $2.24 billion Friday options expiry teased a reversal then fell to $76k again

Bitcoin fell into Deribit’s Sept. 11 options expiry, rebounded during the first hour after settlement, then surrendered the move before a two-hour post-expiry window ended.

The sequence resembled only part of a pattern documented in a recent peer-reviewed study. Friday’s price path did not complete the reversal, while the study’s high at-the-money open-interest condition could not be matched with public data and its strongest negative-gamma regime was not corroborated.

Across Deribit’s BTC perpetual, Coinbase spot and Kraken spot, Bitcoin declined about 0.16% to 0.18% from 07:00 to 08:00 UTC. It then gained about 0.19% to 0.21% by 09:00.

The rebound did not hold. From 08:00 to 10:00 UTC, the Deribit perpetual slipped 0.014%, Coinbase fell 0.025% and Kraken lost 0.038%. Bitcoin therefore ended slightly below its expiry-time level on all three venues.

What Friday’s expiry showed

A PerpFinder market-data snapshot at 06:40 UTC placed the expiring Bitcoin options at about $2.24 billion, comprising roughly $1.40 billion of calls and $844 million of puts. Deribit’s official delivery price was $77,234.

Bitcoin fell as low as $76,000 into the Europe afternoon trading session before recovering back toward $77,500 as of press time.

Under Deribit’s settlement rules, the options expired at 08:00 UTC. The delivery price is a 30-minute time-weighted average of the exchange’s Bitcoin index from 07:30 to 08:00 UTC, sampled every four seconds.

Those mechanics created a timely comparison with research published in Finance Research Letters. The peer-reviewed study examined 1,059 Deribit expiry days from January 2021 through December 2023 using five-minute returns.

Its authors found a statistically significant tendency for Bitcoin to fall in the hour before expiry and reverse during the following two hours when at-the-money open interest ranked in the sample’s top decile.

The result was strongest when a reconstructed cumulative gamma proxy was negative. The paper did not infer market-maker positioning from total open interest or a put/call ratio. Its underlying methodology estimated a market-maker proxy from contract trading history and assumptions about which side initiated each trade.

The pattern, however, does not seem to be playing out in 2026 as three model-based dashboards, Optionly, CryptoGamma and ByKaranteli, did not show a negative cumulative gamma proxy near spot.