Daily Report — 09/04/26 · “The Pin That Survived the Print”
August payrolls printed +162k against +55k expected, with July revised up from a loss to a gain, and the S&P closed down 0.38% at 7,718, inside its daily band, sitting exactly on the strike where the week's expiring options had their largest negative gamma node. That was the day: a hot number the tape refused to sell, because there was nothing left to squeeze it into and the expiration held it in a fifty-point pocket. Underneath the pin the money moved hard in one direction. Memory and semiconductor hardware took the largest intraday dark-pool inflow in the census, MU up 6.1% and SNDK up 11.9% on real prints rather than the closing cross, while software, health care, payments and discretionary sold by price, breadth ran 254 up to 326 down, and every mega-cap “buy” on the tape turned out to be the 16:00 auction. Gold, silver and the miners fell on the hike odds and the structural line held anyway because the dollar did not rally. Then the weekend arrived with three Iranian tankers struck after a carrier was targeted, crude already up nearly 10% on the week, and Monday closed. Tuesday opens on a gap the tape has not priced. This edition reads Friday, prices Tuesday against the levels, and, because Savino has published nothing for the month, builds the framework's own September projection the way his charts are read: timing, direction and shape from the calendar and the dealer book, magnitude from the expected-move boards and nothing else.
The one thing that happened: a hot print, and a tape that was pinned rather than bought
The number was the bear trigger the Thursday report named. +162k against +55k, the prior month revised from -23k to +21k, wages in line, participation up. The rates layer moved the way a hot print moves it: September hike bets jumped on the wire, the ten-year closed at 4.78%, the highest weekly close since October 2023, and TLT stayed under its quarterly floor. The equity index did almost nothing. The S&P opened firm at 7,731, bled to 7,708 at midday, tested its daily lower band at 7,705.74 twice and did not close under it, and finished at 7,718.60 on a 0.50% SPY range, the second-slowest tape of the week.
The reason is on the expiration map, not in the narrative. The zero-day gamma panels at the close show the largest negative nodes sitting exactly at the closing price: SPX -15,000 at 7,720 and -14,000 at 7,715, SPY -5,600 at 770, with the positive nodes stacked just above at 7,730 to 7,750. Dealers short gamma at spot amplify every swing inside the pocket and let neither edge print, and the front put book that would have been the squeeze fuel expired Thursday, a day before the print. Silva's morning anchors named the pocket in advance: zero-day put wall 7,700, call wall 7,755, gamma flip 7,651. The close landed on the put wall. That is a pin, and it graded exactly as the three-branch prior said a pin would. The convergence count for the index is three weak bullish inputs against one discounted bearish input and a breadth overlay, with the Fed on a hawkish-leaning hold and the hike arm live, so no directional stance is registered on the index and nothing is sized. The reader should take that as a statement, not a hedge: a pinned expiration day on a hot print tells you the tape's structure held, and nothing about where it goes once the structure expires.
The panel walk: what the eighteen pages said
Market Net Flow closed with calls at about +130 and puts near zero on the cumulative scale, the widest call-put gap of the week, and the Market DEX printed its fourth consecutive green bar at about +1.7B of net dealer deltas. The zero-day scatter for SPX shows the day's premium concentrated at 7,700 to 7,750 in both directions with almost nothing struck outside the pocket, which is the picture of a tape trading its own expiration rather than a view. The 0DTE gamma panels are the section above.
The Flow Map by expiration is the more interesting page. The Sep-04 line finished at about +30 calls and +15 puts, which is the expiring hedge closing out rather than new money. The 2027-04-16 expiry carried -35 on calls and -5 on puts, the one large negative on the map, and it is a long-dated call sale, which is the overwrite complex on SPX and the single names that the monetization ledger reads file-wide. The 2028-01-21 expiry took +13 calls and +7 puts. The Flow Timeline put the Sep-04 index line at +23 by the close, the Sep-18 line recovered from -85 to about -30 across two sessions, Sep-11 sat at +10, and Sep-25 held -35 to -45. Read with the rate-of-change rule, the Sep-18 line's two-session improvement is the largest slope on the page: the quad-witch hedge that was rebuilt into last Wednesday is being netted down, not added to.
The Dealers Diary is the page that builds the month. Sep-04 expired with puts -6 and calls +2.7. After it: Sep-08 about +2.4 long, and that book expires Tuesday during the session. Sep-10 about +0.6, Sep-15 about +0.4, Sep-18 puts -3.1 against calls +2.7 for a net of about -0.4, a mild short book into the quad-witch, Sep-29 about -0.7, Oct-21 about +0.7, Jan-15-2027 about +1.2. Flat-to-mildly-short through the 18th, long after it. That calendar is the spine of the September projection below.
Top Flow had MU, TSM, SNDK, SMH, INTC, DRAM, IREN, NVDA and EWY as its green bars and MSFT, DELL, PATH, META, BE, AMD and TSLA as its red ones. Three of the red bars are call writing on names that ran (AMD, BE, DELL), which is holders locking gains and is read that way. Sector Flow Premiums had technology at +4.6M as the only clearly green sector with financials and cyclicals red. The Darkpools cards named IVV the highest inflow at $6.48B, up 47% against its average, and the largest single trade at $1.32B, with technology at $35.06B of dark notional. Sector Net Darkpool showed technology at about +17.5B, the not-attributed bucket at about +14B and financials at about +6B on the raw, cross-included scale; the census stripped of the auction tells a different story, in the dark-tape section.
What the option tape said once the structure was stripped: financing, a tail bought at size, and four readable buys
The file was $12.62B gross across 34,619 deduplicated rows and its citable directional residue after the five-stage ladder was +$83M, under one percent. The file is structure-dominated, as every file has been since the ladder went in, and the index book is an artifact: SPX ran $5.59B gross with 638 structures stripped worth $2.69B of premium, leaving a +$124.8M residue at a 2.2% share that carries no direction. What the structures say is more useful than any residue would have been. The Nov-20 7000 call, 8000 put and 9000 put legs printed 3,000 by 3,000 by 900 as a conversion complex. The Dec-2032 7000 and 8000 calls and puts printed 700 by 700 with no side, a long-dated box. The Oct-30 6000, 7000 and 8000 calls printed as a deep call ladder financed by a 10,000-lot far-out-of-the-money write, and the Nov-30 8050 and 8100 calls printed 10,000 lots each below the bid at 15:57, which is opening call writing far above the market and is a structure label, not a bearish read. The Sep-18 7000 calls were bought to the ask in four clips, about $210M of synthetic long stock rolled into the quad-witch, and the same delta-one shape printed in the Oct and Nov 7000 calls. The one directional index print of size was the Mar-19-2027 6300 put, 7,200 lots bought to the ask for $44.1M at 24.3 implied vol, six months of protection 18% under spot, and the vol table below says what that cost.
SPY's residue was -$11.2M at a 4.3% share, inside the artifact zone, but its structure matters: the front hedge that vanished Thursday was rebuilt Friday morning as a Sep-18 760 / 745 put spread, 32,460 lots, the 760 bought above the ask and the 745 sold at the bid for about $6.5M net, a defined-risk hedge that sits exactly on the SPY zone low for Tuesday. QQQ's residue was -$32.1M at 10.6% and it is a collar: the Oct-16 700 calls sold to the bid and the Dec-31 700 calls bought to the ask, 10,000 by 10,000 in the same second, a deep-in-the-money calendar roll of synthetic long stock from October to December, with the Jan-2027 725 put bought to the ask, 6,000 lots, as the protection. A hedged long, which is a different thing from a bear. NVDA's $957M gross was 90 structures: the Oct-09 203 calls, 24,825 lots opened at 09:59 with the stock at 234 and almost no extrinsic, are synthetic long stock, the Sep-18 200 calls rolled to Oct and Nov, and the Nov-20 235 straddle, 11,500 by 11,500, is vol bought into the November print window. The $156M of “call buying” on the Top Flow panel is someone putting on long NVDA stock through the option market. A positioning fact, not a flow figure.
Four names carried a readable residue over the 25% floor, all buys. EWY at +$47.1M and a 72% share, the cleanest readable buy on the board: the Jun-2027 160 calls, 6,667 lots bought to the ask, with 10,000 of the 245 calls and 3,333 of the 120 puts bought in the same second, a long-dated bullish ratio call structure on Korea with a disaster put, and the 245 wing priced under the body at 42.8 implied against 46.9 for the near-the-money strike, the one genuinely cheap wing of the day. TSM at +$34.6M and 34%: the Nov-20 380 calls, 7,500 lots bought to the ask for $44.6M at 36.9 implied vol against an October at-the-money near 30 to 33, a leveraged long with about 20% of the price in extrinsic, so a real bet rather than delta-one, with the Oct 430 and Dec 450 calls written above it. PANW at +$6.5M and 35%, the one software name with a readable buy. EEM at +$8.1M and 32%, the emerging-markets wrapper bought alongside Korea. Nothing readable sold over the floor. TSLA at -$31.4M and 10% is near-dated puts bought against a price break with the deep Sep-18 500 and 520 puts sold to the bid, a synthetic long being unwound; AMD at -$36.7M runs against a +4.7% close and is long-dated call writing into a run, the harvest signature.
The VIX book rolled its wing. The Nov-18 31 calls, 125,998 lots, and the Nov-18 34 calls, 126,533 lots, were bought at the ask at 15:02 and 15:42 Central at 110.6 and 119.1 implied vol against a near-the-money November line at about 73, a ratio of 1.5x to 1.6x, while the Nov-18 25 calls, 35,000 lots, were sold at the bid. A quarter-million contracts of November tail rolled up and out, a standing hedge program re-subscribing at 1.5x the body: elevated rather than punitive, and the body at 19 to 25 is where the same view is priced fairly. The Mar-2027 SPX 6300 put paid 1.74x the December at-the-money for six months of protection; the professional bought a tail at size, and it is the size that makes it notable, not the price. MSTR's largest print, the Sep-11 45 calls, 5,737 lots bought to the ask at 15:58 for $56.3M, is 68% in the money with thirteen cents of extrinsic on a $98 premium: delta-one, synthetic long stock, 573,700 share-equivalents put on in the last two minutes of a long weekend, with the front-week implied at about 140 against 76 for December, an inverted front that makes a December long call spread the body and a front-week short call the wing.
The dark tape: memory in, software out, and every mega-cap “buy” was the closing cross
Strip the 16:00 auction and the census single names ran +$6.60B of net dark flow, but technology alone was +$8.23B, the largest intraday inflow in the census this week and nearly double Thursday's, which means everything outside technology was -$1.63B. Inside technology the inflow is the memory and semiconductor hardware cohort and it is price-confirmed in every name: MU +$1.98B on a +6.10% close, SNDK +$1.74B on +11.90%, INTC +$1.06B on +4.51%, AMD +$948M on +4.69%, MRVL +$556M on +7.05%, with WDC, STX, KLAC, LRCX, AMAT, TSM and EWY all up 2.6% to 7.3%. None of those prints carries a closing-cross line. By contrast every mega-cap headline on the darkpool cards is the auction: NVDA's +$3.76B is +$3.40B of cross, META's +$2.56B is 96% cross, MSFT's +$2.24B is 99% cross against a -2.04% close, and XOM's +$897M is 99% cross against -1.69%. The framework writes no mega-cap dark read this session because there is none to write.
The breadth is the other half. 254 names closed up by price against 326 down, from 375 to 193 on Thursday, and ten of the twelve sector chunks were majority down. Health care was the weakest breadth in the census at 38 of 44 decided names down with LLY's sixteen-session book at a new low of -$3.49B. Financials reversed Thursday's 36-of-40-up to 30-of-40-down, with V selling $738M intraday at a cumulative new low and MA and SCHW behind it; the sector's small ex-auction inflow was three names. Discretionary ran 14 of 23 down on LULU, TSLA and NKE. The index wrappers split: IVV took +$6.18B on volume up 45% against its average, the one index print on rising volume and the only bullish index dark input the count accepts, while SPY's +$5.02B came on falling volume and a down close with the census ladder flagging the contrast, and QQQ and IWM saw small ex-auction outflows on green closes.
The darkpool campaign test found no share-block campaign, five thick prints and twenty mid-bucket prints in 3,094 rows, and the monetization ledger read the sold-call side of the option file as a harvest: 85.2% of sold-call premium was holders monetizing against a +3.0% trailing run, with MU, NVDA, SKHY, MSTR, META, DELL, SPCX and SOXL flagged individually. Both verdicts are cited together on purpose, because they answer different questions: there is profit-taking through the option market, which caps upside into the September 18 expiry and is never a short, and there is price-confirmed selling in named non-technology names, PLTR, V, WMT, LLY and BKNG, which is a description of where money left. Neither is a bearish setup. The bears in this file are price bears, the names that broke on volume.
The four timeframes, as one argument: the week's edges are where Tuesday is decided
The daily band for Tuesday is 7,672.84 to 7,764.36 with the two-sigma at 7,627.08 and 7,810.12, anchored on Friday's close. A gap the size of the last US strike, when the low printed about 1.1% under the prior close, lands through the daily lower and on the daily two-sigma, and the daily band by itself only says a gap is a full-sigma event; it does not say where it stops. Asher's implied-vol daily bracket is 7,698.53 to 7,738.67, narrower than Silva's at 0.51 times, which means the option market priced Tuesday before the weekend news and that daily straddle is stale on a gap.
The weekly band, 7,623.25 to 7,813.95 for a four-session week that holds the CPI on Friday, is the level the whole map turns on. Its lower edge sits four points from the daily two-sigma and eight points from the Tuesday zone low at 7,619.94, with the gamma flip at 7,651 inside the same thirty-one-point stack, and the September playbook's measured base rate says an equity-index downside band break in this regime reverts the next session 89% of the time and re-enters the band within the week 74%. So a Tuesday tag of that stack is a buy-the-break setup graded on the Wednesday close, not a reason to sell, and a Wednesday close under 7,619.94 is the one event that converts the week's structure to the weekly two-sigma at 7,527.90 and Silva's net-gamma put wall at 7,500. Overhead, the weekly upper stacks with the daily two-sigma and the 7,800 call wall inside fourteen points; the bull trigger from Thursday's report, a close over the weekly upper, is unmet and now sits 95 points away with a war headline in between. Through the CPI week it is a ceiling, not a target. Asher's weekly bracket, 7,625.21 to 7,811.99, matches Silva's edges within two points on each side, the tightest weekly agreement in the September record, which makes those edges high-confidence.
The monthly band, 7,428.73 to 7,943.55 with the two-sigma at 7,171.32 and 8,200.96, has price dead-centre at +0.11 sigma, and it is where the projection's magnitude lives. A trough built at the weekly stack in the 7,620s is inside the monthly one-sigma, a normal September and not a regime event. The bull leg into the 7,890 to 7,944 stack, the JPM collar's short call and the monthly upper, is a one-sigma monthly event. The inverse path's monthly lower at 7,428.73 is 3.8% away, which is the honest size of the bear case in band terms: a hot CPI plus a hawkish Fed gets you a one-sigma monthly move, not a crash, unless the regime changes. The quarterly band, 6,929.27 to 8,069.45 through September 30, has price at +0.38 sigma with 351 points to the ceiling and no breach live on any index, so no regime flag; and because the quarterly upper sits above both the monthly upper and the collar, any September rip that reaches the 7,890 to 7,944 stack meets a three-level ceiling, which is why the base path fades into quarter-end rather than extending. The JPM collar rolls on the 30th, a dealer-flow event on the last two sessions of the month.
Across the indices: SPY at 770.19 has a Tuesday zone of 760.55 / 767.34 / 774.12, and the Sep-18 760 / 745 put spread bought Friday sits on the zone low, so the tape is hedged at the level a gap would reach. QQQ at 719.06 closed green on the S&P's red day because the memory bid carried it; its zone runs 701.88 to 725.44 with the 725 line stacked with Friday's Jan-2027 725 put buy, and a gap to the zone low lands on its own trend line at 703.30, the same buy-the-break shape. IWM at 295.95 is on its zone mid with the least room down of the three and the worst sensitivity to a hot CPI; no bid, no break, nothing to do until 302.61 or 289.24 prints. VIX cash at about 14.2 is at its zone low into a war weekend with its trend line at 18.23 above price; a Tuesday tag of the 16.26 zone high has been a fade within days in this vol-supply regime, and a zone-low VIX licenses no cheap-tail claim, because the tail is not priced at the cash level: the November wing sits at 1.5x the body on implied vol. TLT is the one asset outside a band, under its 82.50 quarterly floor with the ten-year at a 2023 high; that is continuation, the input that moved is the hike arm, and the reversion line is a TLT close back over 82.50 on the CPI.
The weekend: Iran, the gap, and why the playbook calls it a reversion setup rather than a reason
US strikes hit three Iranian tankers after a carrier and a destroyer were targeted in the region; Iran claims retaliation; the wires are still moving. Crude had already settled at 91.22 Friday, up 9.7% on the week and inside the new Silva weekly band of 86.74 to 95.70 after two weeks above the old one, and bitcoin traded about 80,000 on Saturday after the news. None of this is in Friday's tape and Monday is closed, so Tuesday's open carries two days of headline with no price discovery behind it.
The framework's routing is the playbook's, measured rather than felt. A headline is the most reversible driver class in the band-break study, and the base rates above say a downside band break in this regime is bought the next session about nine times in ten. So the Tuesday gap is a reversion setup at the 7,620 to 7,651 stack, graded on the Wednesday close back inside the weekly lower at 7,623.25, and it becomes continuation only if the input that moved keeps moving: crude settling over its weekly upper at 95.70 with the two-sigma at 100.18 behind it, and the ten-year holding 4.78% or higher, because a crude spike into a ten-year at a 2023 high is the rates input the software cohort cannot absorb. The precedent from the last US strike on September 2 is the third branch and not the minority one: the S&P held green that day while bitcoin fell 3% and crude ran. The one mechanical difference now is that the front hedge was rebuilt Friday as a small SPY put spread rather than left empty, so a gap has a floor at the zone low to lean on.
Two things the read does not do. It does not chase the gap: the event-morning card scored the payrolls week as a risk-on repricing and suspended tactical index shorts through Wednesday, and in any case the framework does not sell a gap it cannot enter, because in a vol-supply regime the spike is crushed within days, which is the second rule of the vol table. And it does not call the barrel from the equities: three sessions have now confirmed that the oil equities sold into a barrel that held, so the weekend re-opens the barrel leg and nothing else; XLE closed under its 64.39 continuation line and the equities flip back only on that line with crude over its weekly upper.
The September projection: the framework's own map, Savino-style, with the magnitude taken from the boards
Savino has published nothing for September, so the framework built its own the way his charts are read: the calendar and the dealer book give the timing and the shape, the expected-move boards give every number, and the projection lines themselves are shape guides that are never graded against a price. The calendar, with each day-of-week computed: Tuesday the 8th is the Iran gap, the expiry of the Sep-08 long dealer book during the session, the Goldman technology conference and ORCL's print after the close; Wednesday the 9th the index-short suspension ends; Thursday the 10th is PPI and ADBE; Friday the 11th is the CPI, the binary Governor Waller named; Wednesday the 16th is the Fed, hold against hike with the hike arm live; Friday the 18th is the quad-witch and the reported Bank of Japan hike, and it is the day the net-short Sep-18 dealer book rolls off, which turns a mild floor into a move; the 22nd through the 25th is the post-expiration window of weakness the framework owns convexity into; Tuesday the 29th is the Sep-29 book; Wednesday the 30th is quarter-end and the JPM collar roll; Friday October 2 is September payrolls.
| Path | Timing | Direction | Structure | Magnitude, from the boards only | What kills it |
|---|---|---|---|---|---|
| BASE dip, base, rip, fade (about 45%) | trough forms Sep 8 to Sep 16 (the gap bought at the stack, then a CPI and Fed base); rip begins after Sep 18 as the net-short book expires; peak Sep 22 to 25; fade into Sep 30 | down, then flat, then up, then down or flat | trough, rally, fade: the shape the dealer calendar draws | trough at the weekly and zone stack 7,620 to 7,651, inside the monthly one-sigma; rip to the 7,890 to 7,944 stack (collar short call, monthly upper), under the quarterly ceiling 8,069; fade toward the monthly anchor near 7,686 | a Wednesday Sep 9 close under 7,619.94 (not a trough, goes to INVERSE); a close over 7,813.95 before Sep 18 (the rip came early, goes to BULL) |
| BULL SQUEEZE (about 30%) | no trough of consequence: the gap is bought Tuesday and the index closes over the weekly upper before the CPI; peak late, Sep 25 to 30; shallow fade | up, then flat | rally then plateau: Thursday's bull trigger firing on a cool CPI and a hold | 7,813.95, then 7,890 / 7,943.55 / 8,069.45; the monthly two-sigma 8,200.96 is the outer limit, not a target | a close back under 7,759.31 after the trigger; a hot CPI with the ten-year over 4.80% |
| INVERSE hot and hawkish (about 25%) | trough one on Sep 8 holds nothing; hot CPI Sep 11; a hike or a hawkish hold Sep 16; the post-expiration window Sep 22 to 25 is the low of the month; a quarter-end bounce only | down, then down, then up or flat | staircase down with a late bounce: the Sell-America shape, ten-year up, dollar down, metals up, equities de-rated at the multiple | first leg to the weekly two-sigma 7,527.90 and the 7,500 put wall; the month's low at the monthly lower 7,428.73, a one-sigma monthly event of 3.8%; the monthly two-sigma 7,171.32 and the collar's long put 7,090 are the outer band, not targets | a Wednesday Sep 9 close back over 7,623.25 (the reversion worked, goes to BASE); a cool CPI |
The structure of the three is the point. Base and bull differ only in when the rip begins, after the 18th or before the 11th. Base and inverse differ only in whether Tuesday's gap is a trough bought at the stack or the first stair down with crude over its weekly upper. The whole month's shape is therefore decided by one grade, the Wednesday close against 7,623.25, and one print, the CPI on the 11th. The framework has no opinion on the CPI's number. It has a map for each outcome, and every level on the map is Silva's, Asher's or the collar's. The projection is graded the way the timing protocol grades any projection: hit or miss on whether each inflection printed in its window and in its direction, never on a price level, with magnitude graded separately against the bands. The August chart graded its August 6 trough a hit and its August 21 peak a hit by one session and missed the fade; the same grid applies here.
Software: the reversion stalled on the real yield, and the sleeve's lines held by inches
Thursday's relief was nominal. The ten-year went back up to 4.78%, the hike arm re-opened on the print, and the real ten-year sat seven basis points over the breakeven, inside the deadband where the multiple cap is neither asserted nor released, so the cohort had no bid once the Waller relief was spent. MSFT gave back three-quarters of Thursday, closing at 499.70 and holding its 499.40 weekly line from last week by thirty cents, with the day's +$2.24B dark print 99% closing cross against a -2.04% close, a distribution-by-price read. NOW closed at 141.26 with a dead-cat leg verdict, holding its 140.87 death line by thirty-nine cents and under the 142.90 confirm, and the one readable print in its option book was a Jan-2028 120 put written to the bid, 2,000 lots, a willingness to own the stock 15% lower and no bid at spot. CRM, WDAY and ZS sold 2% to 5.4%, PATH lost 16.6% on its print, ADBE fell 6.7% on a CEO succession announcement with its own print Thursday, and NFLX dropped 5.4% with its sixteen-session book at -$1.00B and falling. PLTR sold $791M intraday, price-confirmed, the cohort's one clear institutional seller. ORCL was the exception, up 3.1% to its zone high with 0.04% of room into its own print Tuesday after the close.
The software cohort's description under the setup rules installed this weekend is abandonment: no call bid for weeks, holders overwriting, put writing far below spot. That describes and never times. The bearish reads inside it are the names that broke on volume, and they are price reads with lines, not flow calls. The software-reversion thesis from Thursday required both a second MSFT close over 499.40 and a NOW hold over 142.90; MSFT met its half, NOW did not, so the pair confirm failed and the thesis carries to its Tuesday review with the weaker line 0.39 away. A gap prints it. The sleeve re-read on MSFT is licensed only on a close under the monthly lower at 478.14, not on a gap open, and the one software name with a readable buy in the option book was PANW.
Memory and semiconductor hardware: a leader candidate at one session, and SNDK is the name to watch for exhaustion
The cohort's inflow was price-confirmed everywhere and its option books are holders' books. MU at 1,016.59 printed a +$40.4M citable residue in agreement with its close, and the residue's content is the tell: the Jan-2027 1100 puts, 1,000 lots sold at the bid at 10:13 for $20.3M, are a put write 8% above spot, someone who wants MU stock at 1,100 in January, with the Sep-18 1000 calls overwritten against the run. That is a sponsored name being harvested at the same time, which is what a leader looks like mid-run, and the census's own ladder guard withholds the bullish input because the sixteen-session book is decaying at the margin, so the framework calls it a re-rate candidate at one session and not a re-rate. James's line, that MU must reclaim and hold 1,050 or the move is noise, is the external test; Silva's weekly upper at 1,081.15 is the band cap. SNDK at 1,740.00 was the strongest price in the census and the most chased, with the Sep-11 1525 to 1550 calls bought to the ask in the front week as leverage on the day. The setup scan tags it sponsorship rather than exhaustion because the call-chasing percentile has not reached the top decile; when it does, the exhaustion tag prints, and in the first walk-forward grade that tag carried 39% odds of a 10% drop within ten sessions against a 14% base. It is advisory and it is a trim-into-strength note for a holder, never a short. Arete's 1,600 call wall on SNDK is broken and behind.
TSM at 428.91 sits at its zone high with the widest zone range on the board, 107, a dominant uptrend by the board's own measure; the Nov-20 380 calls bought Friday are the same structure the framework's open TSM thesis holds, twenty-one points over its kill, so the read is healthy and the add is the 418.53 zone mid on a pullback, not the tag. EWY at 188.87 is at its zone high with 0.18% of room and is the same trade in a wrapper, Samsung and SK Hynix; the add is its 180.35 mid. AMD's -$36.7M residue against a +4.7% close is long-dated call writing into a run, the harvest shape on the Top Flow panel's red bar, and its 492.78 zone high is the cap into the Goldman conference. SMH's book is a bought risk reversal, the Oct-16 615 calls bought above the ask and the 535 puts written below the bid, 13,000 by 13,000, with Jan-2027 580 puts written above spot: bullish by structure, at fair vol. Across the cohort the tier rule holds: one price-confirmed cohort inflow after two opposing sessions is a stabilization watch, and the second confirming session is Tuesday's gap, graded on the Wednesday close.
The mega-cap board: NVDA's first close over its re-rate line, META at its decision point, and everything else was the auction
NVDA: three confirming closes, a book that has not turned
NVDA closed at 230.36, up 0.84%, its first close over the 229.65 re-rate line and its third confirming price session, at its zone high with 1.68% of room. The dark tape says the opposite of the price: the +$3.76B headline was +$3.40B of closing cross, the ladder is still a moderate distribution book at -$9.86B cumulative with a falling three-day slope, and the census's contrast flag vetoes the accumulation tag. The option book is delta-one: about $156M of Oct-09 203 and Oct-16 199 calls opened Friday morning with almost no extrinsic, which is synthetic long stock put on at 234, plus a Nov-20 235 straddle that is vol at par into the November print. So someone got long NVDA stock through the option market, the price crossed the line, and the dark book still has not turned. The read stays a stabilization watch: the re-rate is licensed when the three-day dark slope turns rising, and a name tagging its zone high with the book still falling is not added on the tag.
META: the add line is thirteen cents of room away
META closed at 616.77, a second close over the 600.28 weekly cap from last week, with a recovery-bid leg verdict, sponsorship tag and a harvest flag on a +6.7% trail. The monthly upper at 618.08 held by 1.31 and the zone high at 625.99 leaves 1.49% of room. The dark print was 96% cross, and the option book is a two-sided synthetic-stock structure being built around 616 into the September 18 expiry: Sep-18 730, 790 and 820 puts bought to the ask at 15:13 to 15:14 Central, deep in the money with under 3% extrinsic, which is synthetic short stock, against a Dec-18 450 call bought to the ask, synthetic long. Financing and conversion, not a bearish signal, and the same shape that printed on Wednesday and Thursday. The add-not-licensed stance from last week formally dies only on a close over 618.08; price has met the precondition and the flow leg cannot be verified because the census emits no ladder for META. Graded Tuesday.
The rest of the mega-caps, in one pass
AAPL +0.51% on a fast tape with its label net noise, inside its zone. GOOGL -1.17% to 338.46, under its zone mid, with GOOG -1.11% to 335.31 against its own zone of 330.56 / 337.12 / 343.68, the two share classes read separately; the dark-campaign thesis from late August reached its review date with price between its trigger and its death line and was expired as stale. AMZN -0.18%, inside its zone. AVGO +0.21% to 357.89 after Thursday's beat-sold day, with an Oct-09 360 straddle program of about 34,500 by 34,500 lots opened between 14:29 and 15:54, roughly $87M of at-the-money vol with the side unmeasurable; the dividend-data thesis's AVGO leg needed a close over 361.40 and did not get it. TSLA -1.99% is the one mega-cap where price, ladder and book agree on a bearish trend: a strong distribution ladder at -$3.35B cumulative with the tag agreeing, near-dated puts bought against the break, deep Sep-18 puts sold to the bid as a synthetic long unwinds, and a Mar-2027 500 call bought as a lottery ticket at 46.6 vol. Its level is the 337.08 zone low. MSFT is in the software section. Every mega-cap dark print Friday was the auction and no mega-cap dark read is written.
The earnings sort, graded: a clean raise is paid for one session and then harvested, and a blemish is sold hard
LULU missed and cut its full-year guide and fell 17.4% to 100.61, one and a half times its measured 11.9% median reaction; the deep-in-the-money December puts that printed Wednesday afternoon, about $181M of synthetic short opened three hours before the print, were paid. Post-print the surface collapsed from about 80 implied to 39.5 and any new put there is bought at par vol, so there is no cheap short and none is proposed; Click Capital's crash-buy at 100 against his fair value of 194 is logged as a commentary position, not adopted. PATH lost 16.6% and printed 13,000 lots of deep Sep-18 18.5 puts to the ask after the drop, synthetic shorts, structure not signal. ZS fell 4.5% on a beat with a soft outlook. SNOW, which was paid 16.6% on Thursday for its clean raise, gave back 5.4% on Friday on profit-taking with its dark book at a new low. HPE reversed again, -4.5% after +5.0%. AVGO's beat-sold day was followed by a flat day. DELL, the cleanest raise of the week, extended a third session to 524.14 with an accumulation ladder at a new high and a harvest flag on a +14.9% trail. The discriminator stands, a clean raise is paid and any blemish is sold, with a second rule added this week: the beat is paid for one session and then harvested, which is the covered-call regime the monetization ledger reads file-wide.
Ahead: ORCL reports Tuesday after the close at its zone high with 0.04% of room, and ADBE reports Thursday after a CEO succession day that was bought with synthetic stock and hedged with a January put and November vol. Neither name has a measured reaction history on file, so any “typical move” for them is unverified and is written that way; the implied moves are what the option market says, and the asymmetry at a zone high on a blemish is the only thing the sort can say in advance.
Gold, the dollar and the bond: the Sell-America line held on a red day
GLD fell 0.84% to 406.77, NEM lost 1.8%, SLV 1.2% and GDX 2.2%, because the hot print lifted the hike arm and the real yield, and the line held anyway because the dollar did not rally: DXY closed 99.16 with the zone board's trend line at 100.18 above price, a downtrend by its own measure. The hard-asset thesis from last Tuesday required GLD over 402.78 with the dollar under 100.03, both conditions met on the close, and it resolved confirmed on the level. The option tape on the red day was bullish structure: a one-year 406.77 / 445.41 call spread opened at the close, an Oct 425 / 450 call spread, and Sep-18 410 calls, 5,400 lots bought at the ask at 24.6 implied against a 23.8 at-the-money, par vol across the curve with the 450 wing only 12% richer than the body. The buyers used the dip. The next tests are a DXY close over 100.03, which is the death line, or a GLD close under its 395.62 weekly floor; the Tuesday zone is 391.73 / 411.55 / 431.36 with 6% of room up and 3.7% down, and the add, if the framework adds, is the weekly floor and not the mid-band. Kramer's and Arete's reads, that the ten-year at a 2023 high is the metals' real headwind for a day and rates matter more than the dollar for that day, are the same causal claim from two desks and are logged as one.
Energy: the barrel held, the equities did not, and the weekend re-opens only the barrel
For the third session the oil equities sold into a barrel that held. XLE closed 64.06, under the 64.39 line that the continuation branch of last week's oil thesis required, so the equity leg of that thesis is negated and it resolved partial; 24 of 43 decided energy names were down, XOM lost 1.7% with its +$897M dark print 99% cross and its ladder decaying, SLB's ladder decaying too, COP and XOM both carrying the census's contrast flag. Crude settled 91.22, up 9.7% on the week and inside the new weekly band of 86.74 to 95.70. Three sessions make it a board fact rather than a thesis: the equities do not follow the barrel in this tape. What the weekend re-opens is the barrel leg alone. A Tuesday settle over the weekly upper at 95.70 is continuation, with 100.18 the two-sigma behind it, and it is the rates input the software cohort cannot absorb; a settle back under Friday's 91.22 is the headline faded. The equities flip back to a continuation read only on XLE over 64.39 with the barrel over its weekly upper at the same time, and the book's oil exposure is discussed in the operator section below rather than here.
Crypto-beta: price gave back, the books made new highs, and the long weekend is a coin flip
IBIT fell 2.4% to 45.23 with its accumulation ladder at a new high of +$629M and the contrast flag on, MSTR fell 1.4% to 142.80 on a fast tape with its sixteen-session book at a new high and rising, HOOD lost 2.1% with 84% of its period volume in the closing cross, and COIN lost 4.2%. The ETF bid persisted while price fell: James's count has about 10,000 bitcoin bought through the ETFs on Thursday. Bitcoin traded about 80,000 on Saturday after the tanker news. The Tuesday question is which precedent governs, James's stated base rate of 70% up across a US long weekend, which is commentary-sourced and not independently measured, or the September 2 shape where bitcoin fell 3% on the last US strike while the index held. The framework holds the convexity sleeve on its lines, IBIT's 40.40 zone high from August and its 38.05 weekly two-sigma floor, both more than 10% below, and does not size the coin flip. MSTR's close print, the delta-one Sep-11 45 calls, is a positioning fact worth knowing for anyone short the front: 573,700 share-equivalents got long in the last two minutes into a front-week implied vol near 140 against 76 for December, and an inverted front like that is the shape that produced the prior squeezes.
Financials and health care: rotation in, one session out; distribution by price, described not shorted
Financials reversed. Thursday's 36 of 40 up on a +$3.15B intraday dark inflow became 30 of 40 down by price on Friday, with the payments names the sellers: V sold $738M intraday at a cumulative new low, MA $245M, SCHW $201M, none of it cross, while the sector's small ex-auction inflow was JPM, GS and BAC and JPM's print was entirely the auction. XLF closed 58.10, above its zone mid and under the 58.74 zone high, and the sector premium bar on the dashboard stayed red. The financials-in thesis from Thursday is neither confirmed, which needs XLF over 58.74 with the premium bar green, nor dead, which needs XLF under 57.93 with a negative ex-auction net; it reviews Wednesday. Health care was the weakest breadth in the census for a second session, 38 of 44 decided names down with an ex-auction outflow of $566M, LLY's book at a new low and UNH's +$1.04B print entirely cross. That is a description by price and the framework writes it as one: XLV at 171.45 is inside its zone and above its trend line at 162.69, so by the board's own measure the sector trend is intact, and nothing about two sessions of selling licenses a short in a bull regime.
Credit, sentiment, breadth, and what changed in the framework this week
HYG sat at its zone mid with no credit gate live. FOM sentiment printed 51.8 neutral, down 2.0 on the day and 8.2 on the week, with both contrarian arms inactive and no velocity input. The Fed balance sheet grew $6.3B on the week, one week and not a regime; the yen kept firming with USD/JPY at 156.22 into the reported Bank of Japan hike on the 18th, which is the shadow side of the liquidity picture and is analyst-fed rather than computed. The speculative basket alarm is standing at 15 of 27 names distributing on its first night and the AI basket alarm cleared at 6 of 33. The stance-mix check at the close found the census falling and the framework calling bears by price on TSLA, NFLX, LULU, PATH, ZS and LLY, no drift either way.
The reader should know what changed in the rulebook this week, because it changes how this report reads sellers. A replay of every 2026 decline found that no seller-based read, not put buying, not far-out-of-the-money call writing, not a day of dark pool at the bid, not distribution into strength, led a single one of them; the grinding bears of the year were ignored, with no call bid for weeks, and the only bearish setup the flow can see in advance is exhaustion, a vertical run with call chasing in the top decile and holders selling calls into it in the top quartile, which carried about 35% odds of a 10% drop in twenty sessions against a 20% base. So heavy call selling on a running name is now read as a holder locking gains, never a short; the tripwire that used to flag far-out-of-the-money opening call writing as bearish was re-keyed after it fired once in a 9% selloff on a META call butterfly; the setup scan tags sponsorship, exhaustion and abandonment nightly, and a stance-mix check warns at the close if the census is falling while the framework calls no bears or the reverse. What is bearish is unchanged: price breaking on volume, a trend below the bands and the 200-day, an exhaustion tag, a regime shift, sentiment over 80 with fragility, a bearish earnings regime, a failed bottom test. Every bearish line in this edition is one of those.
The desks against the tape
James at InvestAnswers read the hike odds as barely moved by the print, counted about 10,000 bitcoin of ETF buying on Thursday, put the long-weekend base rate at 70% up, set MU's line at a reclaim-and-hold of 1,050 and said he has traps set for the next AVGO dip. Arete carried the memory data, HBM demand up 88% next year on UBS's count, RAM prices up 10% on the month, DRAM sold out to 2030, read the Goldman conference as the Tuesday front-run, said software funds the rotation, and had SNDK's 1,600 call wall as a structure level, which broke. Michael Kramer had the ten-year's weekly close as the highest since October 2023 and ISM services prices paid at 72.3, a series record, and made the multiple-cap argument. Click Capital bought LULU's crash at 100 and read the September hike odds at just over half. The Maverick, a standing perma-bear whose thesis is context and never a signal, has a September top; his shape is the inverse path above and it is graded on the same inflections, and his mood carries no exit authority over any operator line. Silva's morning anchors, put wall 7,700 and call wall 7,755 and flip 7,651, were the map that graded the day a pin, and his zone document is the source of every zone level in this report. Where these desks agree with the tape they are cited beside it; where they lead it, the position is logged and not adopted.
Grading Thursday's report: the pin branch printed, the vanished hedge was the right call, and the bull trigger never fired
Thursday's edition, “The Hedge That Vanished Into the Print,” carried a three-branch prior of Bull 40 / Pin 35 / Bear 25 for a payrolls-plus-expiration Friday and named +162k as the bear trigger. The print landed hot and the pin branch printed: hit on the structure read, and a lesson on the prior, because the event card had tilted the pin toward bull and the tape did neither. The structural claim, that the front-expiry hedge had expired the day before the print and the Sep-18 line was being netted down, graded a hit on both counts; Friday's tape rebuilt a smaller front hedge as the SPY 760 / 745 spread, which is the correction the report said would have to come from somewhere. The software-reversion thesis graded held-not-confirmed, the weakest possible hold, with MSFT clearing its line and NOW missing. The financials-in, energy-out rotation graded half right: energy out held for a third session, financials in reversed in one. The hard-asset thesis graded confirmed on the level. The bull trigger, a close over the weekly upper, never came within 60 points. The GOOGL dark-campaign thesis reached its review date unresolved and was expired.
Thursday's Unusual Prints, graded on the Friday close: the MSFT November 420 calls are behind with the stock down 2%. The SPX Sep-30 7525 puts sold are ahead, the index 190 points above the strike. The EWY June 2027 call ladder is ahead, the fund up 4.6% and its option book printing the cleanest readable buy of Friday on the same structure. The QQQ June 2027 740 puts sold are ahead. The LULU deep-in-the-money puts resolved, the synthetic short paid in full. The TLT September 87 puts are ahead with the long bond still under its floor. The MCD deep-in-the-money puts sold are flat to ahead. The VIX October 70 calls were graded a lesson rather than a trade and stay one: the wing was 2.4x the body then and the November wing was rolled at 1.5x on Friday, priced in vol either way. Grade for the edition: a B. The map held, the structural read was right, the prior over-weighted the bull leg of a print that did not run.
Tuesday's map: the levels own the gap, the Wednesday close owns the week
The covered session is Tuesday the 8th, not an expiration, so no expiration prior is armed; the index-short suspension from the payrolls card runs through Wednesday; the Sep-08 long dealer book expires during the session; and the weekend added a gap driver the tape has not priced. Three branches, qualitative and unmodelled. Gap-and-hold: the S&P opens 0.7% to 1.2% lower, tags the stack at 7,619.94 / 7,623.25 / 7,651, and closes back inside the weekly band; the trigger is any print at or under the 7,651 flip, and the grade is the Wednesday close back over 7,623.25, the base path's trough beginning and the memory cohort's second confirming session on the same close. Gap-and-go: the stack fails with a Tuesday close under 7,619.94 while crude settles over 95.70 and the ten-year holds 4.78% or higher; the next structure is the weekly two-sigma at 7,527.90 and the 7,500 put wall, then the monthly lower at 7,428.73, and the software sleeve's monthly-lower lines print first; the projection re-scores toward inverse and index shorts stay suspended regardless, because the framework does not chase a gap. No gap: the headline is faded overnight as it was on the 2nd, the index opens inside its daily band and pins under the weekly upper, bitcoin and crude carry the headline instead, the rotation continues into the Goldman conference, and the bull trigger stays unmet.
Levels, by instrument, source named. SPX up: 7,759.31 zone high, 7,764.36 daily upper, then the three-level ceiling of the 7,800 call wall, the 7,810.12 daily two-sigma and the 7,813.95 weekly upper, then the 7,890 collar and the 7,943.55 monthly upper, then 8,069.45 quarterly. SPX down: 7,689.63 zone mid, 7,672.84 daily lower, 7,651 flip, then the stack at 7,627.08 / 7,623.25 / 7,619.94, then the second stack at 7,527.90 / 7,527.28 / 7,500, then 7,428.73 monthly. SPY 774.12 / 767.34 / 760.55 with the new front hedge on the low. QQQ 725.44 / 713.66 / 701.88 with the trend at 703.30. IWM 302.61 / 295.93 / 289.24. VIX 16.26 the zone high, a fade in this regime. Crude 95.70 continuation and 100.18 the two-sigma, 91.22 the fade line; USO 146.26 the zone high. GLD 395.62 the weekly floor and the add, 402.78 the confirm, DXY 100.03 the death line. TLT 82.50 the reversion line on the CPI. The single names: MSFT 499.40 and 478.14, NOW 140.87 and 142.90, META 618.08, NVDA 229.65 and 234.23, TSM 418.53 the add and 408.13 the kill, IBIT 40.40, NFLX 76.54, XLE 64.39.
Bottom line
A hot print that the tape pinned on its own expiration rather than sold, a rotation underneath it into memory and semiconductor hardware on real intraday dark prints and out of software, health care and payments by price, and a hard-asset line that held on a red day because the dollar would not rally. The index count is three weak bullish inputs against one discounted bearish input with the Fed on a hawkish-leaning hold, so nothing is licensed on the index and nothing is sized. The weekend's escalation lands Tuesday as a gap the tape has not priced, and the playbook's measured base rates make it a reversion setup at the 7,620 to 7,651 stack, bought at the break with defined risk and graded on Wednesday's close, with crude over its weekly upper at 95.70 the one thing that turns it into continuation. The month is drawn: dip, base, rip after the quad-witch, fade into quarter-end, with the CPI on the 11th and the Wednesday close the two things that decide which of three shapes it takes, and every number on that map comes from the expected-move boards.
Unusual Prints — Top Trades to Follow
1. EWY June 2027 ratio call structure — the cleanest readable buy on the board, with the wing cheaper than the body
2. TSM November 380 calls — a leveraged long bought at fair vol, with the overwrite above it
3. MU January 2027 1100 puts sold — someone wants the stock 8% higher than it closed
4. SPX March 2027 6300 puts — the one directional index print of size, a tail bought at 1.7x the body
5. VIX November 31 and 34 calls — a quarter-million contracts of tail rolled up and out
6. SPY September 760 / 745 put spread — the front hedge that vanished Thursday was rebuilt Friday, on the zone low
7. MSTR September 11 45 calls — 573,700 share-equivalents got long in the last two minutes
8. META September 18 deep-in-the-money puts — the synthetic short that keeps printing at the decision point
Built from the 09/04 close over the Labor Day weekend: the full options tape (35,858 rows, deduplicated and structure-adjusted through the five-stage ladder, with the index book reconstructed print by print), the dark-pool census across 651 names and all twelve sector chunks with the closing cross stripped, the darkpool campaign test and the monetization ledger, the nightly setup scan, the four Silva expected-move timeframes for the 09/08 session with the Asher implied-vol brackets beside them, the 09/08 zone sheets (the ZONE DOCUMENT numeric table and the Zone Visual trend layer), the 26-page options dashboard read panel by panel, the FOM sentiment gauge, the four intraday cuts and the close recap, the FOM and Maverick morning briefs, the InvestAnswers, Arete, Kramer and Click Capital commentary, the JPM Q3 collar, the September playbook and the band-break base-rate study, the timing-model protocol and the August projection precedent, the thesis ledger, the correction log and the session logs through 09/05, and the operator's reconciled books. The August payrolls figures, the ADBE succession, the LULU guide, the S&P inclusion slate and the weekend Iran developments are wire- and search-sourced; the full-text article fetches failed and are cited from summaries. Savino has published no September projection; the map above is the framework's own and is graded on its inflections, never on a price.