Dealer gamma regime conditions whether opening price shocks continue or revert into the close

Dealer positioning is supposed to change how prices behave, not just how much they move. When dealers are long gamma, their hedging leans against moves and an opening shock should fade; when they are short it should extend. Does the morning's dealer positioning regime condition whether an opening price shock continues or reverts into the close?

Tests the CONDITIONING mechanism; dealer gamma positioning modulating intraday price dynamics; not a flow-level signal. Pooled across 9 mega-cap symbols with symbol intercepts. Standing limitations: gamma quantities carry the then-unresolved OI-vintage label (exposure-surface OI vintage not pinned, measured in v0 forensics); GEX aggregates all strikes/expirations with multi-leg contamination uncorrectable on options-surface-exposure. A null does not falsify strike-local gamma, dealer-inventory-signed, or volatility-conditioned variants.

Nine liquid names (two index ETFs and seven mega-caps) pooled with symbol intercepts over seven months of sessions. The shock is the log return across the opening half hour (09:30–10:00 ET). The positioning regime is mean signed dealer gamma exposure over the same window, scaled by its own trailing 20-session mean absolute level using strictly prior sessions (at least 15 required, so warmup sessions drop). The signal is the product of the two, and both main effects enter as controls, so the registered quantity is the interaction alone. The outcome is the log return from 10:01 to the close. OLS with session-clustered errors; the permutation null reassigns each session's positioning block to another session in the same calendar tercile while shock and outcome stay paired; 999 draws; preregistered negative direction.

Nothing, and cleanly. The interaction came in at 0.031 with t = 0.46 and a permutation p of 0.65 across 125 sessions and nine names, with the point estimate on the opposite side of the registered direction as well. Positioning regime does not condition shock continuation or reversion at this horizon. One diagnostic was recorded rather than promoted: in the same regression the positioning LEVEL carried a t of 2.03 against the return outcome. That was not the registered claim and it was not treated as a result. It was logged as a candidate, and the magnitude-space version of it became the registered claim of HYP-010, which found dealer positioning damping realized volatility. This was also the lab's first interaction claim. Two more have followed, both also null, so interaction claims stand at 0-for-3.

The options surface behind positioning quantities covers only strikes that actually traded, not the full open-interest ladder. Statements about total or net positioning describe the traded subset of the book, not the whole book.

The open-interest snapshot behind positioning quantities was established as point-in-time valid by a prospective paired probe: open-settled, unchanged intraday and post-close, with no archive restatement. Results published before that verdict were produced under the then-unresolved vintage caveat. A weekly sentinel re-gates all positioning research automatically if the measured semantics ever change.

Formation windows close roughly a minute before outcome windows open, and measured feed publication latency consumes most of that slack. Live availability of each signal at the formation boundary is therefore tight rather than generous.

The positioning measure aggregates every strike and expiration, and a material share of the underlying premium comes from multi-leg structures whose attribution cannot be corrected on this surface. A null on the aggregate interaction therefore does not falsify strike-local, dealer-inventory-signed, or volatility-conditioned versions of the same idea; those are different quantities and would need their own registrations.

Like every result the lab has published so far, this is a discovery-substrate result: preregistered, but not yet confirmed on data the research process has never seen.