Published nulls

Trading ideas we’ve tested and rejected.

Most research that gets published is research that worked. That selection is what makes published findings hard to trust: you cannot see the tests that were run and quietly dropped.

So the lab publishes the other side. Everything below was registered before it was run and reported whatever it found: nulls, results too unstable to stand on, and questions that a measurement problem blocked before a test was ever spent. Each one narrowed the search.

HYP-009NULLEfficiency Results

Options premium bursts predict volatility expansion over the next 15 minutes

A burst of options premium had already failed to predict which way price would move. That leaves the weaker and more plausible version of the same intuition: perhaps a large slug of premium does not say where price is going, but does say that something is happening, and volatility is about to pick up. Direction is the…

Verdict: t = 0.75 · p = 0.47
What we learned: Bursts do not mark volatility expansion. The mean pair difference was 0.006 log units, roughly half a per cent of relative volatility, with t = 0.75 and a permutation p of 0.47 across 3,319 matched pairs in 139 sessions. Correctly signed, far too small to matter, and well inside the permutation distribution. Read with its parent, this is the more useful null of the pair. Reusing the same frozen event definition and…

Read the full experiment

HYP-005NULLOptions Flow → Direction

Options premium bursts drift in the burst direction over the next 15 minutes

The opening-window family fixed the clock and asked whether the tape predicts the day. This asks the question practitioners actually pose when they watch a feed: when an unusually large slug of options premium hits at some arbitrary moment, does price follow it over the next quarter hour? A yes would have meant that…

Verdict: t = 0.69 · p = 0.51
What we learned: Nothing follows the burst. The mean signed post-event return was 0.32 basis points with t = 0.69 and a permutation p of 0.51 over 3,341 events in 140 sessions. The point estimate is correctly signed and economically negligible, and 3,341 events is a sample large enough that a drift worth trading would have been visible. This is the strongest of the directional nulls in one specific sense: it frees the test from the…

Read the full experiment

HYP-004NULLDealer Positioning & Volatility

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…

Verdict: t = 0.46 · p = 0.65
What we learned: 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…

Read the full experiment

HYP-003NULLOptions Flow → Direction

Cross-sectional opening options pressure predicts relative 09:41-to-close returns across mega-cap names

The two pooled tests of opening options pressure asked whether the tape predicts an index proxy's own return, and both came back null. This asks a structurally different question of the same signal: on a given morning, do the names with the most buy-side options pressure outperform the names with the most sell-side…

Verdict: t = 0.25 · p = 0.80
What we learned: No ranking information. The mean top-two-minus-bottom-two spread was 2.8 basis points per session with t = 0.25 and a permutation p of 0.80 over 140 complete sessions, and the rank correlation between signal and outcome was approximately zero. Differencing out the market did not uncover a relative effect that the pooled tests had masked, which is the specific alternative this design existed to test. Two things came…

Read the full experiment

HYP-002NULLOptions Flow → Direction

Aggressor-partitioned opening options pressure predicts 09:41-to-close underlying return

The aggregate opening-pressure null left one obvious escape route open. Aggregate net premium pools trades initiated by buyers with trades initiated by sellers, so genuine directional pressure could have washed out in the aggregation rather than been absent from the market. This experiment closes that route. It…

Verdict: t = -0.56 · p = 0.61
What we learned: The escape route is closed. The coefficient came in at −0.0048 with t = −0.56 and a permutation p of 0.61 over 140 sessions: indistinguishable from zero, and negative where the registration expected positive. Both symbols and both halves of the sample sit near zero, so there is no subsample doing the work and no instability to appeal to. This matters more than a second null usually would, because it removes the most…

Read the full experiment

HYP-001NULLOptions Flow → Direction

Opening 10-minute aggregate options pressure predicts 09:41-to-close underlying return

The simplest version of the oldest question about options flow: when the opening tape shows more premium going one way than the other, does the underlying follow? A yes would have meant that aggregate options pressure is a directional signal readable in the first ten minutes and usable for the remainder of the…

Verdict: t = 1.57 · p = 0.13
What we learned: No. The pooled coefficient came in at 0.0018 with t = 1.57 and a permutation p of 0.13 over 140 sessions: correctly signed, small (a fully one-sided opening tape maps to under twenty basis points of subsequent return), and comfortably inside what reshuffling the signal produces by chance. Aggregate opening options pressure carries no rest-of-session directional information at this granularity. The result did two…

Read the full experiment

HYP-008INSTRUMENT FAILUREDealer Positioning & Volatility

Flow followed by open-interest growth as persistent positioning

Before asking whether options flow followed by open-interest growth identifies persistent positioning, a precondition had to hold: can the open-interest input be shown to be point-in-time valid? Does a session's snapshot reflect what was actually knowable during that session?

Verdict: retired before any test was run
What we learned: That an honest ledger needs an instrument-failure disposition. Re-pull stability and daily updates were confirmed, but the vintage identity is unrecoverable from archives alone, and an earlier live probe was indeterminate. Using the snapshot intraday without pinning its vintage would be silent lookahead, so the experiment was retired unrun, no statistical test was spent, and every positioning-derived result in the…

Read the full experiment

Get the nulls too

One email per batch of experiments: the results that worked and the ones that didn't.