Daily Pulse — Options Flow + Dark Pool, 8 September 2026

Daily Pulse · Tuesday, 8 September 2026 · EOD options flow & dark-pool research · AZTMM HLDGS LLC

8 September 2026 — EOD read. The tape opened at its high and closed at its low. SPY printed 769.70 in the first five minutes, spent seven hours grinding sideways in a two-dollar band, then broke in the last half hour to close at 765.96, down 0.55%. The single most important thing that happened is not in the price. The 270,729 far-out-of-the-money VIX calls this desk flagged on Friday came back as new open interest almost dollar for dollar, and the very next session reversed.

SPY opened at 769.07 and made the session high of 769.70 inside the first five minutes. The opening fifteen minutes bracketed 767.09 to 769.70, and the rest of the day never traded back above that ceiling. The low of 765.14 arrived in the closing bar, between 3:55 and 4:00 PM, on 3.9 million shares, and the close came in at 765.96. Every one of the fifteen largest same-ticker dark-pool prints in the regular session was contingent, each priced at 770.35 off a derivative leg rather than the tape, and the largest of them carried $193.6M of notional at 9:45 AM. None of those is a level where stock changed hands. Of the heaviest premium shelves, 767, 766 and 765 all sit inside the day’s range and the close landed between them.

Friday’s radar — OI confirm

Nine confirmations, one partial, no failures. That is the cleanest card this desk has published. The two VIX November call lines are the ones that matter: 271,674 contracts of volume produced 270,729 contracts of new open interest, taking the $31 strike from 462 to 128,234 and the $34 strike from 804 to 143,761. Ratios of 1.00 mean essentially every lot opened. Whoever paid $23.1M at the offer on Friday for calls struck at more than twice spot was not closing anything and was not trading around a position. They were putting one on. The single partial is MU, where 414 lots of October $1000 put volume produced 146 new contracts against a 3,640-contract base, which reads as a mix of opening and closing hands rather than a build.

What happened

The index did very little and the leadership did a great deal. SPY lost 0.55% and QQQ lost 0.08%, closing at 718.36 after trading as high as 721.89. NVDA is where the damage was. It opened at 233.11, which was 1.19% above Friday’s close, reached 233.71, and then sold off for the rest of the day to finish at 225.73. That is down 2.01% on the session and down 3.17% from its own opening print. A market that gaps its largest name higher and then spends seven hours selling it is not the same market that opened.

The aggregate options tape looked constructive and the index vehicles did not. Market-wide volume ran 36,576,634 calls against 26,362,908 puts for a put/call of 0.72, with $24.29B of call premium against $14.22B of put premium. Inside SPY the ratio inverted to 1.27, with 5,645,454 puts against 4,458,325 calls and net premium of −$109.4M. QQQ ran 1.15. NVDA carried a call-heavy 0.57 put/call and still finished with $416.6M of bearish premium against $314.3M bullish, for net premium of −$109.1M. Call volume without call premium is what a name being sold into strength looks like.

The tide is the clearest record of the turn. Net call-minus-put premium opened at +$1.7M, fell to −$76.9M by 10:10, then reversed hard through late morning to peak at +$147.2M at 11:40. It held positive through early afternoon, crossed back below zero at 2:45, and then decayed without interruption for the last seventy-five minutes to finish at −$151.0M. The closing print was the worst reading of the entire session. A full round trip of $298M in net premium inside one day, ending at the low, is a market changing its mind late rather than one that started the day defensive.

Volatility is the one number this edition cannot state cleanly. The vendor returned a VIX close of 15.00 against a session low of 15.22 and a high of 15.94, which is impossible, so the session range is published in place of a close and no day-over-day change is claimed.

Why it matters

Friday’s edition said the VIX lines were the story and that Tuesday’s open-interest update would tell us whether any of it was real. It was real. 270,729 contracts of new open interest in November calls struck at 31 and 34, while VIX itself spent today between 15.22 and 15.94, is a position that needs the index to roughly double before it pays. It has been on the books for one session and the first session it saw was a reversal.

What sits alongside it is more interesting than the size. The named leveraged, inverse and volatility ETFs printed almost nothing again: $1.26M of TQQQ September $75 calls on the flagged-events screen, nothing at all in the inverse basket, nothing in the volatility ETFs, and no dark-pool print in any of them above the $100M threshold for a fourth straight session. Light activity in the inverse names reads as an absence of conventional hedging rather than a directional short. Protection continues to be bought in specific far-out-of-the-money places instead of through the usual vehicles, and today added a $8.2M SPXW September 7500 put line and a $5.4M SPX 7675 put paired with a call at the same strike.

The regime read is unchanged in label and softer in level. MPI stands at 66 on the 4 September bar against 68 on the 3 September bar published Friday, and the label remains Bull · early. A two-point slip is not a flip and is not scored as one. The Ledger scores the label, not the number.

What to watch into Wednesday’s session

• SPY 765.14 is today’s low and the level to watch. A close below it would extend the reversal into a second session.
• SPY 769.70 is today’s high and the opening-range ceiling the tape failed at all day.
• NVDA 224.85 is today’s low and 233.71 the high. The gap between them is 3.9% of the name in one session.
• QQQ 715.57 and 721.89 are the equivalent brackets.
• Net call-minus-put premium closed at −$151M. A positive close would end the reversal on the flow side.
• The SPY October $767 call line carried five separate repeated-hit alerts for $11.5M against open interest of 5,903. Tomorrow’s open-interest update settles whether it was opened.

Names on our radar

Every row below is frozen at publication and scored mechanically in the Accountability Ledger at +5 and +21 sessions — misses stay on the page. Options prints are checked against the next morning’s open-interest update in the following session’s post.

The set-up

Three things happened on the same day and they point the same way. A position in far upside volatility confirmed as genuinely new. The largest name in the index gapped up and was sold for seven hours. And net premium made a $298M round trip to finish at the session low. None of that is a forecast, and one session is not a trend. What it does mean is that the flow and the price agreed today for the first time in a week, and they agreed late in the day rather than at the open. Wednesday’s open-interest update tells us whether the SPY $767 call line was anyone opening a position on the other side of it.

The boards — data appendix

Today’s board was substantially larger than Friday’s on every measure. Notional rose from $18.09B to $28.43B, prints from 55 to 93, distinct names from 36 to 52, and the ETF share of notional from 53.1% to 68.6%. One methodology note belongs with the comparison: today’s count additionally removes four rows the vendor flagged as late or out-of-sequence re-reports of an earlier execution, which Friday’s count did not.

Premium stacked almost entirely at the money. The heaviest strike is 767, carrying $121.8M of call premium against $97.0M of put premium for a net of +$24.8M, and it sits one dollar above the 765.96 close. The nearest shelf on the other side is 766 at −$17.9M net, one cent below the close. Those two strikes are adjacent and pull in opposite directions, which is what a market with no conviction about direction looks like from the strike ladder. The next meaningful put shelf sits at 760, at −$14.8M.

Boards are computed on the complete universe of dark-pool prints of $100M or more for the session: 93 prints across 52 distinct names, $28.43B in total notional, paginated by descending premium until a page returned fewer than fifty rows. 103 raw rows were returned, 6 exact vendor duplicates were dropped, and 4 late or out-of-sequence re-reports of an earlier execution at the same tape timestamp were dropped. No canceled prints appeared.

Relative size

A notional ranking hides the unusual. $835M of MSFT is an ordinary session for MSFT. The table below ranks prints by share count as a fraction of that name’s own 30-day average volume, from a $10M screen rather than the $100M board above.

The two screens disagree and both readings stand. GNK’s 1,522,000-share print is 2.5 times that name’s entire 30-day average daily volume, and at $40.7M it never reaches the $100M board. Not one of the five names above appears in the $100M universe. The biggest-print figure in the tile row and the Participation board comes from the $100M screen; the relative-size figures come from the $10M screen, and neither is a correction of the other.

Cluster watch

Same name, same price, multiple prints. That is the fingerprint of one program working an order. Top three by aggregate notional, of twelve clusters detected at ±0.1%:

Also clustered: SPYM 2 prints, $1.85B at 90.15; MSFT 4 prints, $835M at 493.95; META 3 prints, $791M at 613.48; AVGO 3 prints, $720M at 368.56; MU 2 prints, $526M at 1000.26. Every one of the twelve clusters lands between 4:00 and 4:45 PM ET. That timing is consistent with closing-auction and basket work rather than twelve independent decisions, and it is why the ETF share of notional reached 68.6% today.

Notable data points

1. $28.43B in dark-pool prints of $100M or more, across 93 prints and 52 names, with 6 exact duplicates and 4 late re-reports removed from 103 raw rows.
2. Nine of ten flagged contracts confirmed in open interest, one partial, none failed.
3. The two VIX November call lines converted 271,674 contracts of Friday volume into 270,729 contracts of new open interest, ratios of 1.00 on both.
4. Net call-minus-put premium ran from −$76.9M at 10:10 to +$147.2M at 11:40 to −$151.0M at the close, a $298M round trip ending at the session low.
5. NVDA closed 3.17% below its own opening print of 233.11, and 2.01% below Friday’s close.
6. Market-wide put/call 0.72 (26,362,908 puts / 36,576,634 calls), while SPY’s own options ran 1.27 and QQQ’s 1.15.
7. All fifteen of the largest same-ticker SPY dark-pool prints in the regular session carried a contingent or derivative-priced sale condition, each at 770.35.
8. GNK’s 1,522,000-share print was 247.7% of that name’s 30-day average daily volume, on $40.7M of notional.
9. Inverse and volatility-linked ETFs printed $0 above the $100M threshold for a fourth consecutive session.

Method note

Flow and dark-pool data sourced from Unusual Whales. MPI score and regime classifier are our internal composite; daily synthesis is AI-assisted from those inputs.

Volume-over-OI activity is unconfirmed as new positioning until the next morning’s open-interest update; confirmations are published in the following session’s post. Relative size is a print’s share count divided by that name’s 30-day average volume as reported with the print, computed by AZTMM; it is not any vendor’s relative-size or relative-strength score. Option premium by strike is aggregated across all expiries. The session map is drawn from 5-minute bars; the opening range is the first fifteen minutes; dark-pool prints marked contingent carry the vendor’s contingent or derivative-priced sale condition. Board baselines are trailing-session averages computed from our own archived pulls; derived indicators are computed by AZTMM from licensed market data.

Degradations disclosed for this edition. The options-flow screen returned the fifty most recent alerts above $1M rather than every qualifying alert, covering 2:13 PM to 4:15 PM ET, so the sample is recency-weighted rather than ranked by premium across the full session. Bid and ask side percentages were absent on all SPX and SPXW rows, so no side skew is claimed for them. The vendor’s VIX close of 15.00 sits below the 15.22 session low, so the 15.22 to 15.94 session range is published in its place and no day-over-day VIX change is claimed. The MPI strip is dated 4 September because the pipeline finalizes each day’s bar overnight, and the SPY spot carried in that record (770.19) is the prior close rather than today’s, so the strip labels are split accordingly. Three figures this edition normally carries, the session map, the strike-shelf ladder and the multi-session boards strip, were built and validated but are not published here. The publishing interface accepts a post only as a single payload, and today’s full set of figures exceeded that limit. Nothing was dropped for want of data. Every measured value from those three figures appears in the prose: the session path and its contingent prints above, the heaviest strikes and the nearest shelf on the other side in the appendix, and the session-over-session board comparison in words alongside it.

This is research, not advice. Nothing here is a recommendation to buy, sell, or hold any security.

AZTMM HLDGS LLC is not a registered broker-dealer, investment adviser, or FINRA member. All content is retrospective research published for general circulation, not personalized advice and not trade signals. Options involve substantial risk, including losses that may exceed the initial investment. Full disclaimer.

New here? Start Here · Pulse Lab · Accountability Ledger · Trading Academy · Subscribe for daily updates