📊 Ainvest Option Flow Digest - 2026-08-20 | The Day $1.2 Billion of Headlines Turned Into $387M of Risk

17 tickers · $387.0M net premium · 100 legs · and one $339M "bearish bet" that risked about $185

⚡ The Quick Read

Start with the gap. Today's board printed $1,195.0M of gross premium and $386.9M of net — and on several names the headline was not slightly wrong but wildly wrong:

• TSLA's $339M is really $10.2M — and the actual capital at risk is about $185. Not $185 million. One hundred and eighty-five dollars.
• MU's $122.3M is really $25.7M — 4.3x — and the trade earns an estimated $9,100.
• ORCL's $117.2M is really ≈$32.3M — 78% of the headline is two sides of the same trade cancelling.
• QXO's $18.9M is really $1.85M — a 10.3x overstatement.
• AMAT's "$3.6M net debit" is not a debit at all. It is a credit.

Then look at what arrived after 14:30. Five separate names — AMAT, MU, ORCL, SPCX and TSLA — printed enormous ladders of deep-in-the-money puts expiring the next morning, negotiated on the exchange floor, priced at zero or negative time value. Together they carry a $841M gross headline.

None of them is a bearish bet, and the arithmetic is not close. A put struck far above the share price with no time value is not an opinion about anything — it is a share-delivery instrument. Tomorrow's implied move in AMAT is ±3.04%; its nearest strike needs +16.5% overnight to expire worthless. MU needs +16.1%. ORCL +20.0%. SPCX +32.5%. These are four-to-eight standard deviation events. They will be exercised, and everyone involved knows it.

The proofs are arithmetic, not opinion. TSLA's package balances to the single contract — 36,810 against 36,810 — with both halves priced off one stock reference to within $0.00005. MU's strike difference and price difference agree to eight decimal places. ORCL's decomposition is the only balanced split out of 1,024 possibilities and reconciles to the dollar.

One name today was a real directional bet: QQQ. A 37,000-lot September put spread was genuinely bought for a $14.2M debit — proven not by any buy/sell label but by watching the order get worked for 108 seconds while the price walked up as size grew. It expires two days after a Fed meeting that carries a fresh dot plot.

⚠️ And a warning that applies to every number you will read elsewhere today: the reported data was missing 15 legs. Not mispriced — absent. An entire expiration was missing from ORCL. A whole leg was missing from SPCX. Reported open interest was wrong on 9 contracts, once by 11x. Every figure in this letter comes from the exchange tape, not the feed.

One number to hold onto: our automated classifier labelled every leg of every trade a simple open. The tape overturned all seventeen names. That gap — between what a feed says and what a print means — is the entire job.

📈 Every Name on Today's Board

SPCX is shown since its June 12 listing, not as a one-year return — it has only 48 sessions of history, and the ticker previously belonged to an unrelated fund.

🔁 OI Review — Yesterday's Provisional Flags, Now Resolved

Wednesday's board went through the next-day open-interest test, and this time it held. Of 31 legs across 12 tickers, 17 opened, 13 closed, one produced no net open interest at all — and not a single published read inverted. After two consecutive sessions of corrections, that is worth stating plainly.

✅ The calls that landed

• ⭐ CRM — the $108.75M profit-take was right. We said a whale was closing a July bullish risk reversal, not opening a bearish bet. Open interest collapsed 45,981 → 8,662 on the calls (99.5% capture) and 45,309 → 5,202 on the puts (100.3%). A near-perfect close on both legs. The headline "$107.5M of calls sold" would have told readers the opposite of the truth.
• ⭐ AMZN — both roll-downs confirmed, all eight legs. Every low strike opened (+17,589, 100.5%; +17,657, 100.9%; +3,777; +4,000) and every high strike closed (−6,829; −7,455; −4,978; −4,430). That is the exact signature we predicted: new positions at the bottom, retirements at the top. It was never a fresh condor.
• ⭐ MDT — the cancel correction was right. We cut the published premium from $4.26M to $2,012,500 after finding a cancelled duplicate print. The test was clean: ≈19,000 if our correction held, ≈38,000 if it did not. Open interest came in at 20,940. The correction stands.
• QQQ — the partial-unwind read confirmed. We predicted ≈32,000 and ≈14,000 under a close; actual 33,827 and 14,915, both legs closing at 89-94% capture. And the headline was 9.6x the real capital.
• GDX — the protective-put ladder confirmed. The new $85 rung opened at 99.9% capture while the vacated $75 rung closed at 95.5% — precisely the roll-up-a-floor pattern we described.
• GPN — called to the contract. We predicted ≈4,581 if both sides opened. It printed 4,580.
• ROIV — the overwrite roll confirmed. All three September legs closed; the new December $39 line opened at exactly 100.0% capture from zero.
• PSKY — September closed (62.8%), December opened (103.2%). The calendar roll held.
• NBIS — the short-call roll confirmed: the bought leg closed, the sold leg opened at 96.7%.
• Clean opens elsewhere: GLD (both legs 101.0%), IBIT (both legs, and the November leg came in above our fully-fresh threshold, settling that it did not buy back an old short wing), and WMT's September $125 put at 100.2%.

⏳ The one that stayed unresolved

WMT's August $119 put produced NO NET OPEN INTEREST — baseline 20,954, resolving 21,511, a change of just +557 against a 20,940-lot print. That is a pure transfer: the position changed hands rather than being created or destroyed. We had graded it an inferred close at medium-high confidence, and open interest cannot confirm or refute that — when one side closes and another opens the same size, the count does not move. This is the honest limit of the tool, and it is worth understanding: open interest counts contracts, not intentions.

📌 What it means

Two things. First, the corrections were the valuable part — CRM, AMZN, MDT and QQQ were all published against their headline reading, and all four were vindicated. Second, and less comfortably: the WMT desk's short September $125 put is now roughly $21 in the money after Walmart beat, raised and fell 9.13% this morning. Being right about the structure is not the same as the structure being right.

📋 At a Glance

The late-afternoon expiry-eve program — five names, all deep-ITM puts expiring the next morning, all negotiated on the floor at parity, all mechanical:

Ticker	Headline	Real net	Expiration	Catalyst in the window	The option play	What it means
SPCX	$188.16M	$188.16M	Aug 21 / 28 · Sep 18	None   — no earnings in any leg	9-leg deep-ITM put ladder	Negative time value   — bought at a 0.43%   discount   to parity
TSLA	$339.04M	$10.23M	Aug 21 ·   Monthly OPEX	Nothing at all   before Sep 18	18-leg balanced put verticals	≈$185 of actual risk   — headline overstates 33x
MU	$122.29M	$25.71M	Aug 21 ·   Monthly	Earnings ≈Sep 22   (est.)   —   after   expiry	7-leg parity strike roll	Zero economic profit   — earns ≈$9,100
ORCL	$117.15M	≈$32.30M	Aug 21 / 28 · Sep 18	Earnings ≈ Sep 8   (est.)   — inside Sep-18	Synthetic forward roll, 20 legs	Delta-flat   — +11,222 shares on $117M
AMAT	$74.46M	$3.92M credit	Aug 21 / 28 · Sep 18	Ex-div was today ; no earnings	16-leg parity put ladder	Not one package   — and the "debit" is a credit

The one genuinely directional trade of the session:

Ticker	Headline	Real net	Expiration	Catalyst in the window	The option play	What it means
QQQ	$44.59M	$14.25M   debit	Sep 18 ·   Quarterly	✅   FOMC Sep 15-16   — 2 days pre-expiry, with a fresh dot plot	$685/$660 put vertical, 37,000 lots	Bought, proven   — the order walked   up   as size grew

The morning board:

Ticker	Net Premium	Expiration	Catalyst in the window	The option play	What it means
TSM	$32.00M   debit	Sep 18 ·   Monthly	No TSMC earnings   — NVDA Aug 26 is the big one	10,000 deep-ITM $440 puts	Mostly intrinsic   — $25.9M of it, only $6.1M optionality
AMD	$24.54M   debit	Nov 20 + Oct 16 ·   Monthly	Q3 ≈Nov 3   (est.)   sits between the two	$430/$450 diagonal calendar roll	Delta-FLAT   — a time-and-vol trade, not direction
ARM	$15.66M   debit	Nov 20 + Sep 18 ·   Monthly	Earnings ≈Nov 4   (est.) , between the expiries	$310→$320 diagonal roll up and out	A roll   — our own classifier called it wrong
PTEN	$12.68M   debit	Nov 20 + Aug 21 ·   Monthly	Ex-div Sept 1; earnings ≈Oct 28   (est.)	Five-leg stock-tied roll out and up	Financing   — avoids a $43.6M cash exercise
IWM	$11.11M   debit	Sep 11 ·   Weekly	August CPI prints on expiry morning	$295/$285 put vertical, 55,273 lots	Tactical, not insurance   — the cap gives it away
ROST	$5.61M   debit	Sep 18 / Sep 25 / Oct 16 / Nov 20	Earnings TONIGHT at 4pm   (confirmed)	Four-leg delta-neutral vol roll	They SOLD the earnings vol, not bought it
TEAM	$3.17M   debit	Jan 15, 2027 ·   LEAP	≈One earnings in 148 days	1,500 outright $200 calls	100% real money — and 100% at risk
TLT	$2.75M   debit	Mar 19, 2027 ·   LEAP	Five FOMC meetings , last one 2 days pre-expiry	25,000 $77 puts	Rates   — a bet long-end yields keep rising
OWL	$2.00M   credit	Nov 20 + Aug 21 ·   Monthly	Ex-div ≈Nov 9-13   (est.) ; earnings ≈Oct 29	$12→$11 put roll down and out	Short vol   — sold at a trough, not delta
QXO	$1.85M   debit	Mar 19, 2027 ·   LEAP	No pending deal   — the thesis is the catalyst	1x2 ratio spread rolled $18/$25 → $15/$22	De-risking   — headline was 10.3x too big
XBI	$1.00M   debit	Oct 16 ·   Monthly	FOMC Sep 15-16; undated M&A flow	$185/$200 bull call vertical	Capped   — upside stops at $200

Gross premium across all 100 legs was $1,195.0M. Net — after the money one leg collects pays for another — is $386.9M. We publish net. Read the SPCX row carefully: its net equals its gross because a one-directional ladder has nothing to net, which is exactly why it is the largest real number on the board. And read the TSLA row carefully too: $339M of headline reduces to roughly the price of a nice dinner once you net the two halves against each other.

🔍 The Trades Worth Reading Twice

🧮 TSLA — a $339M headline, and about $185 of actual risk

This is the clearest example we have ever published of why a headline premium is not a risk number.

The feed showed $338M of deep-in-the-money puts expiring the next morning. It looks like an enormous bearish bet. It is not a bet at all.

The tape was missing a leg. An 18th print — 138 lots of the $410 put at 14:36:17 — never appeared in the reported data. Add it back and the package balances to the single contract: 36,810 against 36,810.

Then the pricing gives it away. Restate every leg as the stock price it implies, and the two halves agree at $343.294369 versus $343.294319. That is a difference of five hundred-thousandths of a dollar — $185 spread across 3,681,000 shares. Against 2,000 randomly drawn equal-size splits of the same legs, the observed split is the minimum of all 2,000. It did not happen by accident.

So the $10,228,185 that changes hands today comes back as $10,228,000 of strike-cash difference tomorrow. Net risk: about $185. The headline overstates it by 33 times.

Why would anyone do this? Because a deep-in-the-money put with no time value is not an opinion — it is a share-delivery instrument. Tomorrow's implied move is ±2.30%; the nearest strike, $375, needs a +9.3% overnight move to expire worthless. Six of the eleven strikes printed below intrinsic value. The reported implied volatilities of 0.81 to 2.93 are solver artifacts. There is zero volatility content and zero directional content here.

⚠️ And we cannot tell you which side is long. The buy/sell labels and the per-leg quote positions are the same measurement, not two agreeing signals — and they contradict the size decomposition. Anyone telling you this is bearish is reading a label, not a tape.

One more thing worth knowing: this is recurring. The same fingerprint appeared on the previous monthly-expiry eve at 14:37:46 with 17,850 contracts. Today's is roughly five times larger. And this block is only 64% of the day's campaign — floor prints of 50 lots or more on the expiring chain total 115,226 contracts.

🚀 SPCX — the biggest block of the day, bought at a discount to parity

$188.2M — the largest single print on the board, in SpaceX, which has existed as a public company for 48 trading sessions.

It carries negative time value. The ladder's buyer received $188,978,840 of immediate-exercise value and paid $188,160,000 — a 0.43% discount to parity. Zero optionality was purchased. In economic terms this is ≈2.40 million shares of synthetic stock changing hands, not $188M of "premium".

The most striking evidence has nothing to do with today's print. Look at what has been happening to the existing position:

• The August $175 put lost 1,972 contracts of open interest overnight — on 77 contracts of volume.
• The September $235 and $240 puts lost 76 and 179 contracts on zero volume, across ten sessions.

Open interest cannot fall without a trade unless holders are exercising. They are — putting stock to the writers at $175 to $240 on a $132 stock. On the September $240 put, exercising early is worth about $72 per contract against at most a dime of forfeited optionality. Behind all this sits a standing deep-in-the-money complex of roughly $1.6 billion.

⚠️ Three things we are not going to tell you, because they are not supportable. We cannot tell you the direction — the buy/sell column is provably a mechanical relabel of where each print sat in the quote. We are not publishing a financing rate from these prints; the deviations are so small relative to the quoted spreads that the implied rates span −124% to +96%. And SpaceX is not hard to borrow — measured against six expirations, the implied borrow is 0 to 70 basis points.

A lockup tranche was reported to release today. The timing is striking. But that report is single-sourced, its arithmetic does not reconcile, and this block is 0.75% of the tranche — so we are flagging it as context, not calling it the explanation.

⚠️ A caution on charts you will see elsewhere: the SPCX ticker previously belonged to an unrelated ETF. Any one-year SPCX chart splices two different companies together. There is no meaningful year-to-date figure for this name. Since listing on June 12 it is −17.3%, and it trades below its $135.00 IPO price.

🧾 MU — seven legs, and a spread that appeared to cost more than its maximum value

The reported data showed two legs and $109.7M. The tape shows seven legs of a single package — five were missing entirely.

Start with the thing that looked impossible. The $1,190 and $1,100 puts are $90 apart, and the pair appeared to trade at $90.32 — more than the spread can ever be worth. That impossibility was the clue. The +$0.32 apparent overpay on one leg (+$87,254.74) is offset to the dollar by four small legs priced below reference (−$87,259.75).

This is worth internalising: on a negotiated package, a broker prices the whole thing and then distributes the cents across the legs. Per-leg prices are bookkeeping, not decisions. Reading them as separate trades produces nonsense — in this case, a spread that costs more than it can pay.

Once assembled, the package is exact. Both sides are 3,040 contracts. The weighted strike difference is 84.555921 and the weighted price difference is 84.555921 — a parity error of zero to eight decimal places. Five of the day's six Micron floor episodes price to exactly zero parity error.

Real net: $25,705,000, not $109.7M — a 4.3x overstatement. And that figure is exactly the difference in tomorrow's exercise cash, with an identical 304,000 shares delivered either way. The economic profit is roughly $9,100 of interest.

The history explains it. That 11,307-lot $1,190 put was built in two lumps — +7,783 on July 21 and +4,000 on July 23 — and then sat untouched for a month: dormant, priced at exact intrinsic, with 31 days to run. That is a financing position, not a directional one.

Say the key line plainly: a put struck 25% above the share price with zero time value is a synthetic stock sale at the market. It is not a bet on a decline.

Context worth holding: Micron is +237% year-to-date — and roughly 21% below its June 25 close of $1,213.56, about 23% below its $1,255 52-week high. The $1,190 strike sits essentially at that old high, which is precisely why it is so deep in the money today.

🤝 ORCL — two executions, balanced to the dollar, deliberately built to earn nothing

Twenty legs, $117.2M on the tape against $112.9M reported — six legs were missing, including an entire expiration.

It is two separate executions, not one package — different venue, different mechanism, different pricing conventions.

The second cluster decomposes uniquely. Two sides of 9,180 contracts each. Net premium $30,415,000 equals net strike notional $30,415,000 — to the dollar. Both sides imply the same stock price to the cent. Of the 1,024 possible ways to split those ten legs, only one reconciles at zero error; the runner-up misses by $25,200.

What it builds is a synthetic forward roll: long Oracle into August expiry, short into September, about 193,000 shares. Net delta across the entire $117M package is +11,222 shares — 1.2%. Flat.

And here is the elegant part: it earns nothing, and it cannot. The package struck at exactly parity — 0.00% implied financing. The reason is a piece of options mechanics worth learning: the September $250 put's "hold to expiry" value ($107.79) is below what you get by exercising it today ($108.49). The early-exercise boundary binds, so a deep-in-the-money put calendar physically cannot express an interest rate.

It is a fresh position, not a roll of the same contracts. The $170 put traded 15,295 contracts against 4,109 of open interest — you cannot close 9,120 contracts of a 4,109 book. The old book was dismantled over the prior two sessions.

The human story is in the dates. The $175 strike was built on April 22 with Oracle at $187.50. The $180 on May 6 at $194.03. The $200 on May 28 at $203.70 — all roughly at the money then. Oracle now trades at $141.51. Those puts are $33 to $58 in the money, and whoever holds them is sitting on an enormous gain.

The backdrop is a funding story, not an operations story: free cash flow of −$23.7B, debt up 54% to $167.4B, an S&P downgrade to BBB− on July 9 — one notch above junk — credit protection costs at a record, and a $20B stock issuance still to come. The drawdown began on June 11 on an earnings beat.

⚠️ Which side of this is the customer is genuinely unknowable, and the two possibilities point in opposite directions. A large deep-in-the-money put block on a stock down 27% is not evidence of a fresh bearish view.

🧩 AMAT — a ladder where half the legs printed below intrinsic value

Sixteen legs, $74.5M, and a pricing pattern that looks broken: several puts printed for less than their intrinsic value, with the discrepancies alternating sign leg by leg.

It is not a data error and it is not financing. It is how the exchange allocated the price.

Fourteen of the sixteen prints landed exactly to the penny on the displayed bid or the displayed ask. Restate each leg as the stock price it implies — strike minus premium — and every one snaps onto one of two rails: $483.80 or $492.00, against a fair next-day forward of $487.71. The $630, $670, $680 and $730 puts are all $483.80 to the cent. The $580, $610, $620 and $660 are all $492.00. One leg, the $590, is the "plug" that absorbs the rounding for the whole ladder.

The term structure kills the financing explanation. That ±$4 toll is identical at one day to expiry and at 29 days. Real carry scales with time — one day of carry on a $630 strike is seven cents, not four dollars.

And it is provably not one package. As a single book it would forward-buy 348,500 shares at $491.76 while forward-selling 236,000 at $484.78 — a guaranteed $1.65M loss. In a related batch minutes later, the same contract printed at both the bid and the ask.

⚠️ Correcting the headline economics: the reported "$3.6M net debit" is not a debit. It is a credit. Neither $73.7M nor $3.6M is capital at risk. What is real: $357,090,000 of strike cash changes hands tomorrow, and 584,500 shares move for a net acquisition of 112,500 shares.

Two useful facts. Applied Materials went ex-dividend today, not tomorrow — so ≈$0.53 of today's decline is mechanical, and there is no dividend left before September. And that matters mechanically: put early exercise is driven by interest carry, not dividends, so the absence of a dividend makes early exercise of a deep-in-the-money put more likely, not less. That is a large part of why these things trade at parity.

⚠️ We cannot tell you who started it. Open interest rises when a buyer-to-open meets a seller-to-open, so "opening" attaches to both sides equally. The mechanism is proven; the initiator is not.

🛡️ QQQ — the one genuinely directional trade of the session

Everything above is mechanical. This one is a real position, and we can prove which way it goes.

A 37,000-lot September $685/$660 put spread. The feed reported both legs as sells at identical size — an allocation artifact. It was bought, for a $14,245,000 net debit. The $44.6M headline overstates real capital at risk by 3.13x.

How we know the direction, and this is the most useful thing in this letter:

The package price cannot be faked. A broker can shuffle cents between legs, but not the agreed net of $3.85. Against a package market of $3.76 by $3.87, the print sat $0.02 inside the offer.

The order was worked for 108 seconds, and the price walked up as size grew — 4,195 lots at $3.78, then 2,981 at $3.83, then 37,000 at $3.85. A buyer gets charged for size. A seller gets walked down. The full worked order was 44,176 spreads.

The opposite reading would require a seller hitting the offer three times, across four venues, at progressively worse prices.

The timing is not accidental. ✅ The September FOMC meeting is September 15–16 — confirmed on the Federal Reserve's own calendar — and it carries a fresh set of economic projections. It is the only Fed meeting between now and expiry, and it ends two days before these options expire. July's meeting held rates on a 9–3 vote, with three members dissenting in favour of an increase.

Concentration does the rest of the work: the top ten holdings are 48.2% of the fund and semiconductors alone 21.9%. NVIDIA reports August 26 and Broadcom September 2 — 11.6% of the index inside this option's life.

⚠️ But do not confuse a big payout with good odds. Breakeven is $681.15 — 4.11% below today's price — and the payout ratio is 5.49 to 1. The actual risk-neutral probabilities are 70.8% total loss, 26.8% some profit, and 16.4% for the maximum. Expected payoff equals the price by construction. A payout ratio is not a probability, and there is no free lunch in the geometry.

⏳ One thing is genuinely unresolved: whether the $660 leg opens a new position or closes an old one. The two readings predict ≈92,800 versus ≈18,800 contracts of open interest tomorrow — 74,000 apart, and settled by a single number at 06:30 ET. Both readings are on the same side of the market.

🔄 ROST — they sold the earnings volatility, hours before their own print

Four put legs, $18.33M of gross premium, bought into an earnings report due at 4pm. The obvious reading is a big bearish bet. It is the opposite of what happened.

The 2:1 leg ratios were sized to cancel delta. Back-solved from the traded prices, the ratios come out at 0.521 and 0.512 — both within 2% of perfect neutrality, chosen twice, 49 minutes apart, in different strikes and different expiries. Read as four outright buys the package carries −710,543 shares of short exposure. Read correctly — selling the near legs, buying the far ones — it is +12,886 shares. Flat.

The open-interest trail is what proves it. The September-25 $220 strike held literally zero open interest until August 12, jumped to 3,119 in a single session, then sat untouched for five sessions. Today's 3,100 traded that entire single-owner block. That is not a coincidence of size — it is one position, closed in full.

Three more things point the same way: both near legs printed at or above the national offer (a seller's price) while both far legs printed around 72% across; the sold leg sits at a 30-day volatility peak (z = +1.79) and the bought leg at a trough (z = −1.92); and under the roll reading both packages clear at or better than the mid, while under four-buys they would pay 4-8% through it.

So here is the actual story. TJX beat and raised on Tuesday and fell. Walmart beat, raised, and fell 9.13% this morning. Front-month volatility in every off-price retailer got rich. And this desk sold that richness and rolled into intrinsic-heavy October and November puts — keeping delta unchanged — six hours before Ross reported. Real money: $5.61M net, not $18.33M.

⚖️ AMD — $150M of gross premium, and delta-flat

The two legs look mismatched: 11,500 November calls bought against 12,500 October calls sold. They are not mismatched — that ratio is a hedge. 11,500/12,500 = 0.9200, against a delta ratio of 0.9162. Agreement to four-tenths of one percent.

The consequence is that package delta is +3,195 share-equivalents. On $150M of gross premium, this expresses essentially no view on AMD's direction. What it does express is time and volatility: it collects $115,185 a day in decay and is long vega, entered when implied volatility sat at a two-sigma low.

Two details a headline would bury. 52% of the $75.90 paid for the November call is time value, so it is not the deep-in-the-money stock substitute it appears to be. And the effective breakeven, net of the sale, is $451.33 — 3.25% below today's price. A standalone reading gives $505.90 and would tell readers this needs a 9% rally. It does not.

The calendar explains the shape: Q3 earnings land around November 3 (estimated), so the October leg dies 18 days before the print and the November leg survives 17 days past it. This is a decision to hold through earnings.

🔄 QXO — the biggest overstatement we have published

Six call legs, $18.9M gross. The real money is $1,845,650 — a 10.3x gap, because the gross figure counts both the closing and opening sides of a roll.

A desk rolled a 1x2 short-call-ratio spread down the ladder, from $18/$25 to $15/$22. Two things make that airtight. Cluster A's net delta is +6,490 shares against 4.26 million shares of gross leg delta — 0.15%. Nobody builds a new $18.9M directional package at zero delta. And the ratio is preserved to four decimal places: closed at 1:2.0004, reopened at 1:2.0006.

The $22 strike is the tell — prior open interest of 25 against 75,503 sold, roughly 3,020 times over, so that leg is unambiguously new while the $18 and $25 legs sit inside existing positions.

And the direction of travel matters: QXO has fallen to within 1.1% of its 52-week low, and this roll lowered the breakeven from ≈$18.60 to $15.72 and the profit peak from $25 to $22. That is de-risking a losing position, not a bullish whale.

🧊 TSM — the board's biggest print, and most of it is not a bet

$32M of Taiwan Semiconductor puts is the largest premium on the board. But $25.9M of that is intrinsic value — the strike sits 6.3% above the stock, so the option behaves much like short stock. Only $6.1M is actual optionality.

What settles it is history. A 10,000-lot block printed in this exact contract yesterday at $32.30 — and it added only +5,618 to open interest, meaning ≈44% of that was already transfer. Two identical-size floor blocks on consecutive sessions, 26 cents apart, reads as one position being worked across sessions, not fresh conviction.

Worth knowing: there is no TSMC earnings inside this option's life. The biggest scheduled event in the window belongs to NVIDIA on August 26. And the $440 strike sits exactly at the lowest analyst target on one panel — though a second panel shows a $330 low, and we publish both rather than pick.

🐻 IWM — expiring on the exact morning August CPI prints

A $11.1M put spread, 55,273 contracts a side, expiring September 11 — the morning August CPI is released at 8:30. The FOMC meets four days later. So the trade owns the data that reprices the Fed, not the decision itself.

The macro backdrop is genuinely unusual: markets are pricing ≈77% odds of a rate HIKE in September, three FOMC members dissented for one in July, and the minutes had officials judging conditions "may not have been restrictive enough." The transmission into small caps is mechanical — more than half of Russell 2000 debt is floating-rate against 24% for the S&P 500, and ≈40% of its constituents have no earnings. A hike lands as an earnings shock, not a valuation one.

The counter-argument is strong and we will not hide it: small caps are leading, not lagging — IWM is +21.7% year to date against the S&P's +12.9%, and this trade fades that leadership within 2.4% of the 52-week high.

One detail worth a raised eyebrow: the $295 strike they bought is the single largest gamma wall on the entire chain. And a sharp read on intent — the $285 cap argues this is tactical, not insurance. A real hedge does not stop protecting exactly where the pain starts.

🛢️ PTEN and 🦉 OWL — two rolls, two very different motives

PTEN looks like $24.7M of call buying. It is a stock-tied financing package — the combination condition covers 155,568 of 156,151 contracts, 99.6% of size. The motive is plain once you see it: the August legs are pure synthetic stock with five cents of time value, and rolling them out avoids posting $43.6M of cash to exercise 40,668 expiring in-the-money calls tomorrow. It is not a covered-call roll, though — the far legs are bought, and exposure increases 1.73x.

OWL is a put roll down and out for a $2.0M credit — and the real exposure is vega, not delta: net −$95,250 per volatility point, sold at a trough (z = −1.58), which is the unflattering version. Two corrections worth carrying: the $11 strike is not near the lows — the stock is +47.7% above its 52-week low — and a round trip back there would cost ≈$11.4M, or 5.7x the credit collected.

🎯 TEAM and 🧬 XBI — the exception, and the one that got smaller

TEAM is the only headline on today's board that is 100% real money — $3.17M, entirely time value, genuinely standalone, nothing to strip out. It is also 100% at risk: the $200 strike sits 8.7% above the 52-week high, the breakeven 20.2% above it, and not one of fourteen post-earnings analyst targets reaches that breakeven — the highest is $200, the strike itself. Atlassian is +125% in seven weeks and 67.6% above its 50-day average.

XBI is the reverse. Reported as a $1.6M call buy, it is a $999,900 bull call vertical — and the proof is elegant: both strikes traded identical volume, 5,567 each, all session. They only ever traded as a pair. That moves the breakeven to $186.80 and caps the upside at $200.

📅 Upcoming Catalysts — and Which Expiration Actually Covers Them

A catalyst only matters to an option still alive when it happens. Event on the left; the option that survives to see it on the right.

Date	Event	Which of today's trades is alive for it?
Thu Aug 20, 16:00 ET	Ross Stores fiscal Q2 earnings   (issuer-confirmed)	✅ All four ROST legs — every expiry covers it
Fri Aug 21	Monthly expiration — the big one	⚠️   PTEN's two August legs and OWL's August leg EXPIRE   · ⚠️   and so do the entire TSLA, MU, AMAT and SPCX August ladders plus ORCL's August cluster   — ≈$800M of gross headline settles into share delivery
Sat Aug 22	(after Aug-21 expiry)	📦   Settlement:   TSLA 7,362,000 shares / $2.87B of strike cash · ORCL 2,297,000 shares / $417M · SPCX 1,289,000 shares / $251M · AMAT 584,500 shares / $357M · MU 304,000 shares.   Net share flow on the balanced packages is roughly zero
Wed Aug 26	NVIDIA earnings	✅ TSM Sep-18 · ✅ AMD both legs · ✅ ARM both legs · ✅ IWM Sep-11 · ✅   QQQ Sep-18 — NVDA is 8.50% of the fund   · ✅ ORCL Sep-18
Wed Aug 26	July PCE inflation	✅ IWM Sep-11 · ✅ TLT Mar-2027
Thu Aug 27	Burlington earnings   (off-price peer)	✅ ROST Sep-18 and later
Thu-Sat Aug 27-29	Jackson Hole   (dates semi-confirmed)	✅ IWM Sep-11 · ✅ TLT
Tue Sep 1	PTEN ex-dividend $0.10   (confirmed)	✅ PTEN Nov-20 — and note the August short calls expire   before   it, so no early-exercise risk
Fri Sep 4	August payrolls	✅ IWM Sep-11 · ✅ TLT
Tue Sep 8	ROST ex-dividend $0.445	✅ All four ROST legs
Tue Sep 8	ORCL fiscal Q1 earnings   (estimated, NOT issuer-confirmed)	✅   ORCL Sep-18 straddles the print — the August legs do not
Tue Sep 8 / Wed Sep 9	AMAT fireside chats (Citi TMT, Goldman Communacopia)   (confirmed)	✅ AMAT Sep-18 — the only confirmed AMAT events in the window
Wed Sep 2	Broadcom earnings	✅   QQQ Sep-18 — AVGO is 3.09% of the fund
Thu Sep 10	AMAT dividend pay date	—   (ex-date was Aug 20)
Thu Sep 10 / Fri Sep 11	August PPI, then August CPI	✅ QQQ Sep-18 — the last inflation prints before the Fed decides
≈Thu Sep 10	TSMC August monthly revenue   (est. from cadence)	✅ TSM Sep-18 — the only TSM-specific event in its window
Fri Sep 11, 08:30 ET	August CPI — on IWM's expiration morning	⚠️   IWM expires that day   · ✅ TLT
Mon-Tue Sep 15-16	FOMC with a fresh dot plot   (confirmed on the Fed's own calendar)	⭐   QQQ Sep-18 — the meeting ends TWO DAYS before it expires   · ❌ IWM expired 4 days earlier · ✅ XBI Oct-16 · ✅ AMD · ✅ ARM · ✅ TLT
Fri Sep 18	Quarterly triple-witching expiration	⚠️   QQQ, TSM, ORCL, AMAT and SPCX September legs all EXPIRE
≈Tue Sep 22	MU fiscal Q4 earnings   (estimated, not confirmed)	❌   After every MU leg has expired   — the September options die four days early
Fri Sep 18	Quarterly expiration	⚠️ TSM and ARM's September leg   expire
≈Thu Oct 15	TSMC Q3 earnings   (est., not issuer-confirmed)	❌   TSM's option expired Sep 18
Fri Oct 16	Monthly expiration	⚠️   AMD's October leg and XBI expire
≈Wed Oct 28	PTEN Q3 earnings   (est.)	✅ PTEN Nov-20 — inside the November window, outside the August one
≈Thu Oct 29	Atlassian Q1 FY27   (est.)   · OWL Q3   (est.)	✅ TEAM Jan-2027 · ✅ OWL Nov-20
Tue-Wed Oct 27-28	FOMC   (no projections)	❌   XBI expired 11 days earlier   · ✅ AMD Nov-20 · ✅ TLT
≈Tue Nov 3	AMD Q3 earnings   (est.)	✅ AMD Nov-20 — this is why October was rolled into November
≈Wed Nov 4	ARM earnings   (est.)	✅ ARM Nov-20
≈Nov 9-13	OWL Q4 ex-dividend   (est., not declared)	✅ OWL Nov-20
Tue-Wed Dec 8-9	FOMC with projections	✅ TEAM Jan-2027 · ✅ QXO · ✅ TLT
Mon-Tue Mar 16-17, 2027	FOMC with projections — two days before TLT expires	✅ TLT Mar-2027

The pattern worth noticing: several of today's structures expire just before the events everyone watches — IWM dies four days before the Fed, XBI eleven days before it, TSM a month before TSMC's own earnings. Others deliberately step past a print: AMD and ARM both rolled from a September or October expiry into November precisely to hold through earnings. Neither choice is accidental.

👥 Four Ways to Read Today's Board

🎲 The YOLO trader

Ross reports at 4pm today — that is the only same-day fuse here. But understand what you would be buying: the options market has priced a ±8.9% move for tomorrow's expiry against a full-quarter expectation of ±11.33%. Nearly the entire quarter's expected movement is packed into 24 hours, which means you can be right on direction and still lose once volatility collapses after the print. And note what the professionals actually did: they sold that volatility rather than buying it. ROST's gamma map is nearly empty — 16 strikes, zero walls, no resistance at all — so there is little dealer positioning to cushion a move in either direction.

📈 The swing trader

TEAM is the cleanest expression on the board — one leg, no structural tricks, 100% time value. It is also the hardest sell: zero of fourteen analyst targets reach its $221.10 breakeven, and the highest target is the strike. XBI is the disciplined alternative — defined risk, 7.33:1, capped at $200 — but it needs a fresh record, and the market's own October band tops out at $181.60, below the breakeven. That gap between what a trade needs and what the market expects is the number to check before you copy anything.

💰 The premium collector

Today's lesson is ROST, and it is a good one. A desk sold volatility at a 30-day peak (z = +1.79) and bought it back at a trough (z = −1.92) — in the same package, at the same instant, with delta cancelled to within 2%. That is what selling premium properly looks like: not a directional bet dressed up, but a paid transfer of volatility risk. Contrast OWL, which sold vega at a trough (z = −1.58) — the same mechanic, the wrong price. The mechanic is not the edge; the level is.

🌱 The beginner

Three things from today, no maths required.

One — the headline is not the money. $309M of gross premium became $112M of net. On QXO it was 10x off, because the number counted both sides of a roll: the position being closed and the one being opened.

Two — a big trade can express no opinion at all. AMD's $150M package has a net delta of about 3,000 shares. It is a bet on time and volatility, not on whether AMD goes up. Four of today's eleven names are like that.

Three — check where a strike sits before you judge the trade. TEAM's $200 calls need a price the stock has never reached in a year. TSM's $440 puts are already 6% in the money, so most of that "$32M" is not a wager at all — it is intrinsic value that gets handed back at settlement.

⚠️ Before You Trade Any of This

Eleven names today, and the tape corrected every one. Our own automated classifier labelled all 28 legs simple opens. Eleven of them are closes. Three headline premiums were overstated by more than 60%, one by 10.3x. If you traded from the headlines alone, you would have had the story backwards on at least four names.

These are negotiated blocks, not panicked buying. Every structure here was arranged with a counterparty who had already agreed on price. Nobody lifted an offer in anger. A $150M cross is $150M changing hands between two parties who shook on it — not a stampede, and not a signal to chase.

And be honest about what open interest cannot tell you. Yesterday's board settled almost perfectly — CRM, AMZN, MDT and QQQ were all published against their headline reading and all four were vindicated. But WMT's August put moved open interest by just 557 contracts on a 20,940-lot print — a pure transfer, where one side closed as another opened. The count cannot settle that, and we say so rather than pretending otherwise.

One more thing, and it is the uncomfortable one: that WMT desk's short September put is now roughly $21 in the money after this morning's 9% drop. Being right about a structure is not the same as the structure being right. Position sizes here are institutional, hedged in ways an alert never shows you, and patience costs nothing but opportunity.

This letter is research and education, not investment advice. Options carry risk including total loss of premium. Do your own work and size positions you can afford to be wrong about.