The Three Opens

Three prices exist on every session before most traders have taken a trade: the 8:30 level (where price stood when the data hit, or just an hour before cash on quiet days), the 9:30 cash open, and the 10:00 price — where the first half-hour of real trade left you.

We measured every ordering of those three prices across 1,401 NQ sessions (Dec 2020 → Jul 2026), tick-derived 1-minute bars, front month stitched by daily volume winner, everything in percent of price. We went in looking for a directional signal. That’s not what the data handed back.

The base rate first

10:00 → close drifts green on 55% of sessions, median +0.08%. That’s the house drift. Any claim that can’t beat it is decoration. Median favorable and adverse travel after 10:00 is ±0.50% in every configuration — the tape rotates roughly six times more than it drifts. Hold that number; it’s the drop.

Finding 1 — mixed states are magnet days

Classify each session by where the 10:00 price sits against both earlier levels:

Aligned: the anchors get revisited ~6 in 10 — barely above base. Mixed — price straddling the anchors, the morning’s directional attempt half-undone — and the revisit rate jumps ~17 points. On quiet mornings the strongest cell tags the 8:30 level 88% of the time (n=86).

A mixed 10:00 state means the auction rejected the morning’s attempt, and the anchors themselves become the day’s most probable destinations. Rotation regime: the market pays whoever expects the levels to trade again and punishes whoever needs escape.

Finding 2 — only the clean bull stack tilts direction

Six orderings are possible at 10:00. Every single one closes green at the median — that’s the index long-bias, quantified. Only one clears the base rate with margin: 10:00 > 9:30 > 8:30, the clean bullish staircase — 61% green, median +0.159%, n=356. Its mirror image, the clean bear stack, gives shorts nothing: n=323, and 52% of those sessions still close green. Downside continuation in this index dies by lunch.

Best t-stat in the table: +1.4. Say it plainly: a tilt, not a law.

Finding 3 — 8:30 impulses persist up, die down

Bucket the 8:30 → 9:30 reaction, measure the rest of the cash session:

A strong up-reaction keeps paying. The same-size down-reaction pays shorts nothing. And on high-impulse mornings specifically (top-quintile 8:30 range — the CPI/PPI cohort), the weakest cell of the whole study shows up: 10:00 below the open but above the data anchor → 42% win, median −0.107% (n=40, thin sample, flagged as such). Post-data limbo is where longs bleed.

What this buys you — and what it doesn’t

Buys you: a regime classification, decided by 10:00, from three prices on every chart. Bull stack → the only measured directional tilt. Mixed → ~80% odds the anchors trade again; levels are targets, not launch pads. Bear stack → no measured short edge exists. Knowing which hope to delete before it costs anything is the product.

Does not buy you: an entry. No row fires a trade. Best directional t-stat +1.4. No costs modeled. The revisit stats say nothing about when the touch comes or how much heat precedes it. The window is a bull half-decade. Anyone selling you the bull stack as a system is selling drift plus survivorship.

We went digging for a signal. We found a map instead.

How we’ll try to kill it

1. The noise null — any random path revisits old prices. Shuffled-returns test (Drop 01 method): the finding survives only if the 62-vs-80 aligned/mixed split beats structureless paths of identical volatility. This stat does not graduate until that runs.
2. Out-of-sample decade — replicate on 2009–2020: two real bear markets.
3. Regime splits — 2022 alone; vol terciles. Does the bear-stack null survive a bear year, or is it sample?
4. A real calendar — swap the impulse proxy for actual CPI/PPI/NFP dates.
5. Time-to-touch / heat-to-touch — the EV test. Hit rate is not edge until the heat is priced.
6. Liquidity at the anchors — a year of recorded depth: do resting-size shelves form at these prices, or is the revisit pure path behavior?

Whichever way those tests go, they ship. Dead stats teach.