Before arguing about dealer-gamma signs, check how much of the book is even in the print

Two open-interest snapshots and one volume field, per contract, per session. Divide the absolute change in open interest by the day's volume and you get a floor on how one-sided that contract's trading was, because opens and closes inside the same contract net out before they reach an OI print ^[3]. Three hundred lots opened and three hundred closed reports zero ^[12]. Every one-sidedness number in the study is therefore understated by construction, and the large ones survive the handicap: a median contract-day floor of 0.50, with 22% of contract-days above 0.9 ^[7].

The bucket split is where the mechanism shows. A day's trading in a contract 22 or more days out is at least 58% one-sided, against 42% at one to five days ^[6]. The far book is inventory and the near book is turnover, so an open-interest gamma map is most faithful exactly where per-contract gamma is thinnest.

Now the internal cross-check. If 44% of tomorrow's opening 0DTE open interest was created today ^[8], and today's volume in those same 1DTE contracts is 1.06 times that opening open interest ^[9], then newly created open interest is about 41.5% of the volume that produced it ^[15]. That sits a little above the whole-book net figure, and it should: gross opens are clipped at zero, net change is not, and medians of ratios do not divide cleanly. Call it a sanity check rather than a second result.

Coverage is the part that bites. Nearly eight turns of opening open interest on the median 2026 0DTE session ^[10] means the morning's print accounts for roughly 12.5% of that day's activity in those contracts, leaving about 87% of it outside ^[17]. The author states the same thing as an eighth ^[11]. Published dealer-gamma figures start from yesterday's outstanding contracts multiplied by a convention about who holds which side ^[0], so the convention dispute is an argument about the 12.5%.

The residual is also growing. Going from 40% in 2024 to 52% in 2026 is 12 points, a 30% relative rise across three annual readings ^[16]. A method resting on the morning's open interest does not have to change to get worse.

The stated limits are worth keeping in front of the numbers. Nothing here signs a trade, the OI change shows that positions moved rather than who moved them, and it is one archive from one vendor's end-of-day chain ^[12]. Contracts expiring on the day have no next-session open interest and leave the statistics entirely, and 4.1% of rows carrying 3.8% of volume showed an absolute OI change larger than volume, which cannot happen and is attributed to snapshot timing; those were dropped ^[4]. Half the contract-days in the window never traded: 4.3 million of 8.6 million had volume ^[14]. Thin per contract, fat per session, which is the right shape for a claim about sessions.