Two sessions account for all of Bitcoin's Wall Street-hours loss since Sept. 21

Compounded, the declines on Sept. 30 and Oct. 2 add up to a loss of about 4.46% ^[16], more than the 3.24% the whole US-hours window lost ^[1]. On CryptoSlate's numbers, the slide is two sharp days on top of a run that was slightly positive (or rather, positive if the session starts at 9:30 a.m.). Bitcoin still fell in eight of the 13 cash sessions ^[3].

Take out the two outlier days and move the start time, and the gain disappears. A window opening at 9 a.m. still loses 0.16%, and one opening at 10 a.m. loses 1.68% ^[7]. The 9:30 window shows a gain because it leaves out a pre-open half hour that lost money and keeps an opening half hour that made money. Dividing one window's compounded result by another's isolates those edges. Across the 11 remaining sessions, 9:00 to 9:30 a.m. lost about 1.4% ^[19] and 9:30 to 10 a.m. gained about 3.0% ^[20].

The timing holds on a second venue. Coinbase dollar prices fell about 4.85% from 9 a.m. to the close and 5.36% from 10 a.m. over the same period ^[8], while the Binance figures are quoted against USDT ^[9]. Glassnode said this week that Bitcoin's net gains since Sept. 21 have come largely outside the US session, reversing an earlier stretch when American hours contributed more to the advance ^[10]. The sellers' identity is a separate question. Neither exchange reveals whether they are US institutions, retail traders, market makers or global investors trading during American business hours ^[9].

If US institutions were selling, US spot ETF flows are where it should show up, and the flows point both ways. The funds lost $148.7 million on Sept. 30 ^[11], matching the first big drop, then turned buyers on Oct. 2 ^[12]. Outflows were larger on Oct. 7, at $484.9 million ^[13].

The numbers fit more than one reading. US demand is weakening and the two sessions were its loudest days. Or the two days were one-offs and the session evens out from here. Or the sellers are global traders and market makers who are active in New York hours and never pass through an ETF. I think the evidence establishes the timing and leaves the seller unidentified. Two venues quoting in two currencies agree on when Bitcoin falls, while ETF flows were mixed on the days that decide the result. The strongest case for the institutional reading is the Coinbase discount, a sustained negative premium on a venue closely associated with US investors ^[15].

That view is wrong if CryptoSlate's own test is met: Bitcoin keeps losing ground through Wall Street hours while the Coinbase discount stays deeply negative and ETF outflows build ^[21].