AInvest Option Flow Digest — 2026-08-24 · $313.3M

Thirteen institutional packages crossed the tape today across twelve names, and the striking thing is how few of them were anyone taking a view. One desk rolled a losing position and pulled $66.7M off the table. One bought insurance two days before the biggest earnings print of the quarter. One put on a two-strike put ladder in a utility and printed 680,000 shares against it in the same second. One struck a spread so close to at-the-money that it is a coin flip by construction. And one paid ≈$433,000 of time value to control ≈$275 million of stock.

That is not a bearish signal. It is a positioning signal — and positioning is usually more honest than conviction, because nobody pays $22 million for a hedge they think they will never need.

And two of today's twelve are not market views at all — they are the same trade, on the same date. JNJ and PBR both go ex-dividend tomorrow, August 25, and in both names someone bought deep-in-the-money calls at or below their intrinsic value — the signature of an early-exercise structure, not a bet on the stock. One was $4.08M in a Dividend King. The other was $45.5M across four Petrobras strikes, paying ≈$433,000 of time value to collect ≈$6.56 million of dividend. If you learn one mechanic from today's tape, make it that one.

$313.3M net structural premium across 12 names, 30 legs. (Gross across all legs was $951.9M — but gross double-counts both sides of every spread, so we publish the net: what was actually paid out or taken in.)

The day finished almost perfectly balanced: $161.1M of net debits paid against $152.2M of net credits collected. For most of the session credits led — then a single late four-strike program in Petrobras put $45.5M of buying on the board and evened it up.

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

Friday's board went through the next-day open-interest test and split three ways: nine names confirmed, two corrected, and three legs that can never be tested at all. Baseline is the 2026-08-21 snapshot; resolving is 2026-08-24.

🔄 Two corrections — read these first

CG — it was not a collar going on. The cap came OFF. The Sep-18 $52.50 call fell 141,303 → 70,908 (−70,395) while the $40 put rose 26,587 → 92,204 (+65,617). We published the fork in advance with no ambiguous middle: ≈241,300 if the collar completed, ≈41,300 if the short call was harvested. It harvested. Open interest cannot fall on an opening sale, so at least 70.4% of that 100,000-lot block was closed on both sides. ⭐ Concrete consequence: the $52.50 gamma wall, the largest structure on CG's board, is now roughly half its former size. The headline has been withdrawn and rewritten.

ABBV — the September leg closed. It was not a fresh gamma sale. Sep-18 $270 call 10,733 → 8,006 (−2,727); Jan-2027 $300 call 1,318 → 4,855 (+3,537, 101.1%). Of four published branches, the print landed on "the desk retired its own August block" — ≈7,500 forecast, 8,006 actual. The fresh-short branch is refuted by ≈10,200 contracts. The package is a roll out and up, not a new sale of near-dated volatility.

⚠️ Neither correction was a claim that turned out wrong — both legs carried explicit ⏳ PROVISIONAL labels and both articles published the full branch table in advance. What changed is the narrative built around them, and both articles are corrected in place.

✅ The nine that confirmed

• ⭐ EYPT — a falsifiable prediction, passed to within one contract. $2.50 call 30 → 10,058 (forecast ≈10,030); $12.50 call 321 → 10,320 (forecast ≈10,321).
• ⭐ ARM — Sep $300 call −5,704; Nov $320 call +9,042 (100.6%). The opening-sale branch that would have inverted the article is refuted by ≈14,700 contracts.
• ⭐ IBIT — Nov $57 call +29,892 (99.6%), and the $45 control line came back +708 instead of dropping ≈30,000, so the $57 block is additive risk, not a re-strike.
• GLD — Jan-2027 $450 call +19,794 (99.0%), inside the published band.
• IGV — Sep $100 call +19,583 (97.9%). That short-gamma exposure is new risk.
• GEO — both Sep legs opened (+20,732 and +19,216), so the ≈750,000 shares of long delta left standing after Friday's expiry are real.
• MU — Oct $980 put +5,514 and call +5,550. The jelly roll stays a financing trade.
• ORCL — the provisional 2027 $200 leg resolved OPEN (+5,172, 76.6%); the close branch is refuted by ≈11,900 contracts.
• IOVA — both legs rose together (+13,567 and +14,913).

❗ Where we scored ourselves down

Three published numbers came in light and we are recording them rather than rounding them away: CG's $40 put floor of "≥100,000" printed +65,617 (our floor was stated too strongly — the true one-session bound is ≥46,826); IOVA's $10 call came 959 below our own band; ORCL's 2027 $200 captured 76.6%, not ≈100%. All three are still opens. What we got wrong was the size of the transfer component, not the direction.

📊 At a Glance

Ticker	Net Premium	Expiration Type	Catalyst (and its date)	The Option Play	What It Means
MCHP	$66.7M credit	Monthly (Oct-16) ← Quarterly (Dec-18)	Fiscal Q2 earnings   ≈Nov 5   —   after   the new October expiry	Delta-preserving call roll, up & in	Directional, but de-risked.   Same delta, less capital, no earnings
GLD	$57.9M credit	Quarterly (Sep-18)	FOMC Sep 16   — two trading days before expiry; Jackson Hole   Aug 28	At-the-money $420/$430 call vertical	Directional — but which way is unprovable.   $425 breakeven either side
IBIT	$74.1M debit	Monthly +   LEAP (Dec-2028)	Nov 4   Treasury refunding + Strategy Q3;   Sep 15   CLARITY vote	Three packages: double diagonal, 2028 tail spread, long straddle	Mixed — one desk bullish, one buying crash protection, one long volatility
NVDA	$22.0M debit	Jan-2027 / Jun-2027	Earnings Wednesday Aug 26 , after the close	Put ratio diagonal, 1.20:1	Hedging.   Buys the drawdown, sells the crash
XEL	$5.95M debit	Jan-2027	Dec–Jan Colorado wind-driven fire window   sits inside the expiry	Two-strike put ladder + 680,000-share hedge	Structure proven, side not.   Delta-flat either way; leans insurance
JNJ	$4.08M debit	Monthly (Oct-16) / Quarterly (Mar-2027)	Ex-dividend tomorrow, Aug 25, $1.34/share	Deep-ITM diagonal, bought below parity	Neither bullish nor bearish — a financing trade.   ≈6.5% annualised
PSKY	$2.95M debit	Quarterly (Dec-18)	≈Sep 24   bond-motion decision; trial   Mar 2, 2027   —   after   expiry	39,930 out-of-the-money calls, floor block	Unprovable.   Could be a new bet, or someone leaving at a loss
PBR	$45.5M debit	Quarterly (Sep-18)	Ex-dividend TOMORROW, Aug 25, ≈$0.441/ADR   — today is the last day to exercise for it	Four deep-ITM call strikes, floor-worked	Dividend capture, not a bet.   ≈$433K of time value to collect ≈$6.56M
VLO	$24.4M credit   (reported side)	Quarterly (Dec-18)	Q3 earnings   ≈Oct 22 ; Nov ex-dividend inside the window	Two Dec-18 call strikes, 2:3 ratio	Side unresolvable.   A ratio roll-up, or an outright sale
BEKE	$3.09M credit	Jan-2027 / Apr-2027	Q2 earnings   Aug 21   (last Friday); Q3   ≈Nov 10	Two opening $20 call blocks, 17 min apart	Side unproven.   Reported as sold — if so, ≈7.3% of spot collected
CCL	$4.26M debit	Quarterly (Dec-18)	Expires ≈3 days BEFORE Q4 earnings (≈Dec 21) ; fiscal Q3 ≈Sept 28 is inside	20,000 near-the-money puts + 800,000 shares	Fully delta-hedged.   102% hedge ratio; side unproven
PCT	$2.30M debit	LEAP (Jan-2028)	Ironton 24/7   Q4 2026 ; Thailand close   year-end 2026	10,000 LEAP calls + 320,000-share partial hedge	A long-dated bet — with its breakeven above the dilution trigger

🔍 The Twelve, In Detail

1. 🔄 MCHP — $66.7M off the table, and they gave up the earnings print

→ Read why a desk booked a ≈$115M loss to keep the same bet

One floor-negotiated package: sold 118,920 December $65 calls for $153.4M, bought 150,000 October $72.50 calls for $86.7M.

The craftsmanship is real. Old leg: 118,920 × 0.7348 delta = 87,375 delta-units. New leg: 150,000 × 0.5815 = 87,225. The two match to within 0.3% — they kept essentially identical directional exposure while pulling $66.7M back.

But here is the detail that should stop you copying it. There is no Microchip earnings report before the October 16 expiry — fiscal Q2 lands around November 5. The December expiry they just sold captured that print. The October expiry they bought does not. They kept the delta and gave up the event.

And they are selling at a large loss. Those December calls traded $20.43–$24.50 on August 4 and August 7; today they went at $12.90. Their first block went on before a beat-and-raise — a good call, the stock gapped +13.89% to $84.69. The second went on the day of the peak. Then eleven straight down sessions.

There was no bad news. The ≈13% slide is the round-trip give-back of that earnings gap, helped along by a peak-margin warning, the 30-year Treasury hitting 5.33% (highest since 2007), and the COO's departure.

2. 🥇 GLD — a $58M coin flip, struck into seven tier-1 macro events

→ See why we can prove this position is new but cannot prove which way it points

A block cross at 09:51:35, both legs 115,876 contracts: the September $420/$430 call vertical, $5.00 net on a $10-wide spread — $57.9M.

Both legs are proven opens (115,876 against prior open interest of 7,400 and 20,000). What we cannot prove is the direction. On a cross, a broker matched a buyer and a seller off the open book, and where each leg printed inside the bid/ask is an allocation artifact, not an aggressor signal. Both readings share the same $425 breakeven, with gold's ETF at ≈$428.50 — this is an at-the-money structure by construction.

What makes it interesting is the calendar. Seven tier-1 events before expiry with no quiet week — July PCE (Aug 26), Warsh's first Jackson Hole keynote (Aug 28), August payrolls (Sep 4), the Treasury buyback step-up (Sep 9), PPI (Sep 10), CPI (Sep 11), and the FOMC on Sep 16, two trading days before expiry. A wrong-way FOMC gap that late leaves almost no time value to recover from.

Worth knowing: the Fed's live debate in 2026 is hold-versus-hike, not hold-versus-cut.

3. ₿ IBIT — three packages, and they do not agree with each other

→ Unpack all three, including a tail hedge that expires in December 2028

Three separate blocks, almost certainly three different desks:

• $24.1M double diagonal — long November $40/$46 calls against short October $48/$54 calls. Whichever way open-versus-close resolves, this moves the holder's book ≈+2.03 million share-equivalents more bullish. That statement is true under both readings, which is why we lead with it.
• $5.3M tail hedge expiring December 2028 — a $20/$15 put spread, 2.3 years out. IBIT at $44.70 going to $20 is roughly a 55% bitcoin drawdown. Max payoff $25M on $5.3M risked, ≈4.7:1. Somebody has been assembling this for three weeks.
• $44.7M long straddle at the June-2027 $45 strike — and put-call parity checks out exactly, so it is a genuine straddle, not a disguised synthetic.

⏳ The straddle is the best open-versus-close question of the day. The identical straddle was opened on August 20 at ≈$13.80. It is worth $14.90 today. Today's print could be that holder closing for an 8% gain, or a second buyer arriving. Tomorrow's open interest is the only test — and there is no ambiguous middle.

4. 🛡️ NVDA — $22M of insurance, bought two days before the print

→ See the hedge, the financing, and the 12 million shares of tail risk it sells

At 13:57:11: bought 99,995 January-2027 $180 puts for $76.8M, sold 119,994 June-2027 $140 puts for $54.8M. Net $22.0M.

NVIDIA reports Wednesday, August 26, after the close. The four-day implied move is ±7.54% — an earnings-sized number.

Both legs are proven opens, and both open-interest series had been flat for a month, so there was no existing position to close. This is unambiguously new risk — a rare thing to be able to say with certainty.

Net delta ≈ −0.97 million shares, roughly $202M of downside exposure. The $180 strike sits just inside the January expected move — where a real hedge gets struck, not a lottery ticket.

⚠️ Do not copy the short leg. Selling 119,994 far-out-of-the-money puts is 12 million shares of tail risk below $140. This structure protects a moderate decline and sells the crash. A professional running a big long book can carry that. Most people reading this cannot.

5. 🔌 XEL — two puts, 680,000 shares, and a fire-season calendar

→ What the 680,000-share block proves — and what it doesn't

10,000 January-2027 $75 puts plus 10,000 $70 puts for $5.95M, printed as one stock-and-options cross alongside a confirmed 680,000-share block at $76.6206.

Buying two different put strikes is an unusual shape — it could easily have been a spread. The hedge size settles it. Long both puts needs ≈615,000 shares of delta. A $75/$70 spread needs only ≈137,000. Observed: 680,000. The ladder reading matches; the spread reading is off by 5×.

⚠️ But the magnitude test is sign-blind. It settles ladder-versus-spread; it cannot settle who was long. Short both puts would carry delta +614,900 and hedge by selling ≈680,000 shares — an identical footprint. The reported side is BUY, and the wildfire calendar is the strongest reason to lean that way, but it is a lean, not a proof. The motive is on the calendar — the Marshall Fire ignited December 30, a winter downslope-wind event, and Xcel ran a 50,000-customer preemptive shutoff last December 17. That window sits directly inside a January expiry. Meanwhile only ≈$80M of a ≈$525M annual wildfire policy remained at June 30.

6. 💰 JNJ — a trade that is over before you read this

→ The call that traded below its own intrinsic value, and why

Bought 2,520 October $190 calls at $82.57, sold 2,520 March-2027 $210 calls at $66.37. Net $4.08M.

With the stock at $273, that October call has $83.00 of intrinsic value and traded at $82.57 — below parity. Options do not trade below intrinsic by accident.

Johnson & Johnson goes ex-dividend tomorrow, August 25, for $1.34 a share — which makes today the last day a call holder can exercise early and still be a holder of record. Immediate capture on 252,000 shares: ≈$108,000 of below-parity edge plus ≈$338,000 of dividend, ≈$446,000, with no time value forfeited.

Only one leg gets exercised, and that asymmetry is the design: for the October call the early-exercise break-even sits around $231–$308, so a $190 strike qualifies comfortably. For the March-2027 call the break-even is ≈$59–$79 — exercising it would never be rational for this dividend.

This is not a directional trade. Both legs are delta ≈1.0; the package's net delta is ≈12,600 share-equivalents. It is a financing structure yielding ≈3.8% over ≈7 months, ≈6.5% annualised, that only goes wrong below ≈$206.

7. 🎬 PSKY — the same size, the same strike, five days later

→ Three readings, ≈40,000 contracts apart, and tomorrow settles it

39,930 December $12 calls at $0.74 — $2.95M, a single-leg floor block with the stock at $10.35.

⏳ This one we genuinely cannot call, and that is the point. Size 39,930 sits below prior open interest of 42,753 — and that open interest is essentially one block created on August 19, when 42,167 contracts traded at $0.70–$0.94.

Three readings are live: a new buyer opening; the August 19 holder closing at a loss; or a pure transfer with no new conviction at all.

The timing makes the second reading very live. August 19 was the initial case management conference in the antitrust suit blocking the Warner Bros. Discovery merger, and the judge ordered the parties into mediation — the first court-sanctioned settlement path in the case. But the thesis degraded fast: the California AG demanded structural remedies on August 20 and cancelled the settlement meeting today.

Note the geometry: December 18 expires ≈2.5 months before the March 2027 trial. This was never a bet on winning at trial. It is a bet on a pre-trial settlement or on the ≈September 24 bond-motion decision.

8. ⛽ VLO — the one trade today we cannot read

→ Why no amount of open interest will tell us who bought this

One multi-leg auction package at 14:19:51, two Dec-18 call strikes: 2,998 at the $320 strike ($15.07M) and 4,497 at the $390 strike ($9.34M), a clean 2:3 ratio.

The $390 leg is a proven new open — 4,497 against prior open interest of just 45. The side, though, we cannot establish. On other names today the paired stock block settles it; here the largest VLO share print anywhere near the timestamp is 312 shares. There is no hedge to measure.

So we give you both, with numbers. Read as an outright sale of both strikes, it is $24.4M of credit and ≈−376,900 shares of delta. Read as a 2:3 ratio roll-up — sell the $320s you already own, buy 1.5× as many $390s — it is a $5.73M credit and ≈−31,700 shares, essentially delta-neutral. Those are wildly different trades. We lean to the roll-up, because a package routed as one complex order almost always has opposing legs, and because the $320 line has been worked down from 5,295 since August 12.

The lesson worth taking: open interest is side-blind. It rises on any opening trade regardless of who bought. Tomorrow tells us whether the $320 leg closed. It will never tell us who was on which side.

Context that matters for anyone writing upside here: VLO is +113.4% YTD at record territory, and every published analyst target sits below the current price — the Street high is $365, so the $390 strike is above the most bullish target on the tape. One more wrinkle: a November ex-dividend falls inside the option's life, which is an early-assignment trigger on a short deep-in-the-money call.

9. 🏠 BEKE — the cleanest print on the board

→ A zero-open-interest contract — and the one thing it still can't prove

Two block crosses 17 minutes apart, both at the $20 strike: 10,800 April-2027 calls at $1.60 and 13,500 January-2027 calls at $1.01 — $3.09M, with the stock at $17.78. The reported side is SELL.

Here is why this one is worth your attention even at a modest size: the April contract had zero prior open interest. Not low — zero. That makes 100% of that leg new risk by arithmetic, with no inference required at all. Almost all flow analysis lives in the gap between volume and open interest; here there is no gap.

The timing is deliberate. KE Holdings reported Q2 last Friday, August 21 — a 34% adjusted EPS beat that the market paid only +4.5% for, on revenue that still fell 5.7% year over year. Selling upside premium the session after a print is when implied volatility is richest.

⚠️ Two things we cannot tell you. First, the side: these were crosses, so the Buy/Sell label is a reporting convention, not proof — read as sold it is premium collection against a $20 cap; read as bought it is a 24,300-contract bullish call ladder. Second, if sold, whether the calls are covered by a long share position or naked. The tape cannot see stock held away from the print, and open interest will not settle it tomorrow either. Covered, this caps upside at $20 for ≈7.3% of spot. Naked, the risk above $20 is uncapped.

10. 🧪 PCT — a 2028 bet whose breakeven sits above its own dilution trigger

→ The detail that changes how this trade should be read

10,000 January-2028 $10 calls bought at $2.30 — $2.3M — with PureCycle at $6.67. That is a strike ≈50% out of the money, ≈2.4 years out, at an implied volatility of ≈97%.

The paired equity leg is confirmed: 320,000 shares at $6.65, 233 ms later, the largest PCT stock print of the day. It hedges ≈53% of the package's delta — a partial hedge, not a neutral one.

But here is the detail that reframes the whole thing. Breakeven is $10 + $2.30 = $12.30. PureCycle's $287.5M convertible converts at $11.08, into ≈25.9M shares — roughly 12.9% dilution. The trade only pays above the very level at which its largest dilution tranche converts, and convertible-arbitrage hedging tends to damp rallies into exactly that zone.

Be careful with the open/close here too: size 10,000 against prior open interest of 8,922 means only the 1,078-contract excess is provably new. The rest could be a transfer.

The milestone path is front-loaded (Ironton running 24/7 in Q4 2026, Thailand financial close at year-end) then hits a mid-2027 catalyst desert containing a warrant expiry and a revolver expiry — historically when this company raises capital. Short interest is ≈35% of float, which cuts both ways.

11. 🛢️ PBR — ≈$433,000 of time value for ≈$275 million of stock

→ The four-strike program the feed only half-showed

Between 15:20 and 15:25, 148,801 September-18 Petrobras calls printed across four strikes in synchronized clips — $13, $15, $16 and $17 — for $45.5M. Every strike is deep in the money, and every one traded at essentially zero time value. The $15 strike printed $0.075 below its own intrinsic value.

Add it up and the shape is unmistakable: they paid roughly $433,000 in total time premium to control ≈14.8 million shares, about $275 million of stock exposure, into September.

And the calendar confirms why. Petrobras goes ex-dividend tomorrow, August 25, for ≈$0.441 per ADR — which makes today the last day an exercise earns record-date entitlement. Run the arithmetic across all four strikes: ≈$6.56 million of dividend against ≈$433,000 of time value forfeited, a gross spread of ≈$6.13 million. The dividend is roughly 15× the premium given up.

⚠️ Two things worth knowing. First, it looks like a sweep and isn't one — twenty-plus clips lifting the offer over five minutes reads like aggression, but every contract printed as a floor block, broker-worked and negotiated, in uniform repeating clip sizes that are the fingerprint of an automated routine. Second, this is not riskless arbitrage: a naive exercise-and-hold leaves ≈14.9 million ADRs of naked overnight delta, and any offsetting short leg is invisible from the buy side.

All four strikes are proven opens — every leg dwarfs its prior open interest.

12. 🚢 CCL — a put block hedged to 102%

→ The 800,000 shares that show what this position really is

20,000 December-18 $25 puts at $2.13 — $4.26M, with Carnival at $25.80, so the strike is barely 3% out of the money. This is near-the-money protection, not a lottery ticket.

The position is a proven open: 20,000 contracts against an open interest of 7,034 that had not moved by more than ≈10 contracts in three weeks.

And the paired equity leg is confirmed — 800,000 shares at $25.91, the day's largest Carnival stock print by a factor of nine. Against a package delta of 781,200 shares, that is a 102.4% hedge: this position was delta-neutralised almost exactly at inception. Whoever put it on is not betting on direction; they are positioned in volatility and convexity into mid-December.

And the expiry date is chosen, not accidental. Carnival's fiscal year ends November 30 and its Q4 print has landed December 19–21 for four years running; in 2026 the only trading day in that band is Monday December 21 — three days after this expiry. So this is almost certainly not an earnings hedge. It covers the path into December: fuel above all, with Brent up 34% this year and an oil-driven selloff having knocked ≈5% off the stock on August 20.

One more thing worth knowing: not one of 30 analysts has a price target below the current price — the lowest is 11% above it — and there has been no sell-side action since July 23 while the stock fell 11%. Someone paying for near-the-money protection is positioned squarely against that.

⚠️ As with every cross today, the side is not proven — and the hedge cannot settle it either, because long puts hedge by buying stock and short puts by selling, and both produce the same 800,000-share footprint.

⏳ Tomorrow Morning Is the Scorecard

Ten of today's thirty legs printed at a size at or below existing open interest, which means today's tape alone cannot prove whether they opened or closed a position. (An eleventh, PCT, only clears its open interest by 1,078 contracts — so only that much of it is provably new.) The next-day open-interest snapshot (≈06:30 ET) is the definitive test, and we publish our predictions in advance so you can score us:

Test	If it opened	If it closed
MCHP   Dec $65	—	≈300   (a decline under ≈100,000 refutes our read)
IBIT   Nov $40 / $46	≈67,300 / ≈62,100	≈7,300 / ≈2,100
IBIT   Jun-27 $45 straddle	≈64,000 / ≈62,000	≈4,000 / ≈2,000
PSKY   Dec $12	≈82,700	≈2,800 — or   ≈42,700 unchanged = pure transfer
VLO   Dec $320	≈7,614	≈1,618 ⚠️   settles open vs close — but never the side
PCT   Jan-2028 $10	≈18,900 if fully new	≈8,900–10,000 if largely a transfer
JNJ   Oct $190	≈8,600	≈3,600 ⚠️   not a clean test — early exercise around the ex-dividend retires open interest on its own
PBR   Sep $13 / $17	≈50,200 / ≈92,900	⚠️   same trap as JNJ — these are deep-ITM calls; if they are exercised around an ex-dividend, open interest falls for that reason alone and would NOT refute the opening read
CCL   Dec $25	≈27,000	—   open already proven; OI had been flat within ≈10 contracts for three weeks

And two that need no open/close test at all. BEKE's April-2027 $20 call had zero prior open interest, and NVDA's and XEL's legs all printed larger than the open interest that existed — those opens are arithmetic, not inference. ⚠️ But note what that does not buy you: open interest is side-blind. It rises on any opening trade regardless of who bought. On today's crosses and auctions — BEKE, GLD, IBIT, NVDA, PCT, VLO and XEL — tomorrow confirms the position opened and will never reveal the direction.

👥 How to Read Today, By Who You Are

🎲 The YOLO trader. Two names are actually built for you today, and both come with a catch worth reading first. PCT is a January-2028 call at ≈97% implied volatility on a stock with ≈35% of its float sold short — genuine convexity. But its breakeven of $12.30 sits above the $11.08 price at which the company's convertible converts and dilutes, which is a structural headwind sitting exactly where you need the stock to go. PSKY is the other, and we cannot even prove someone was buying it. Meanwhile: NVDA's short $140 put leg is exactly the kind of trade that looks like free money and ends careers. Twelve million shares of tail risk to finance a hedge is a professional's position, not a retail one.

📈 The swing trader. Three calendars matter more than any strike here. NVIDIA reports Wednesday after the close with a ±7.54% four-day implied move. Gold's expiry lands two trading days after the September 16 FOMC. And Valero's Q3 print lands ≈October 22, inside a December expiry, on a stock up 113% YTD where every published analyst target sits below the current price. All three are event risk you can plan around, in either direction. The MCHP roll is your cautionary tale: someone with far better information than either of us bought before earnings, was right, added at the top, and gave it all back in eleven sessions.

💵 The premium collector. Today looks like your day — but read the fine print first. Several names are reported as premium sales, and on a cross that label is a reporting convention rather than a proven side. BEKE is the cleanest teaching example if the reported side is right: two call blocks at a strike 12.5% out of the money, struck the session after earnings when implied volatility is richest, collecting ≈7.3% of spot over five to eight months. JNJ is the professional yield structure — ≈6.5% annualised on a Dividend King with 64 straight years of increases, and honestly one you probably cannot replicate profitably after commissions and exercise costs. But hold on to the two caveats that run through all of them: we cannot prove the side on a cross, and we cannot tell whether any short call is covered. Covered, BEKE caps upside for a fee. Naked, the risk above $20 is uncapped. That distinction is the entire difference between an income strategy and a blow-up, and no amount of flow data will settle it for you. More broadly, watch how the day actually finished: $161.1M of debits against $152.2M of credits — almost exactly balanced. Credits led for most of the session, and then one late Petrobras program put $45.5M of deep-in-the-money buying on the board. A tape that looks one-directional at 2pm can even out by the close, which is a decent argument against reading too much into any single hour of flow.

🌱 The beginner. Start with PSKY, the smallest single-leg trade on the board and the best lesson available: when a trade's size is smaller than the open interest already sitting in that contract, nobody can tell from the tape whether someone opened a new position or closed an old one. A "$3M bullish call buy" headline may be somebody getting out at a loss. Then read BEKE for the opposite case — a contract with zero prior open interest, where the proof is pure arithmetic. Then VLO, which shows the limit of the whole method: sometimes even tomorrow's open interest cannot tell you who was on which side. Then PBR, for the trap that catches almost everyone: twenty-plus blocks lifting the offer in five minutes looks like frantic buying, but every one printed on the exchange floor — negotiated, not swept. Those four articles together will teach you more about reading flow than any indicator.

⚠️ Risk, Plainly

A large trade is not a recommendation, and it is not a prediction. Every institution on this list has information, hedges, mandates, and time horizons you cannot see. The OPRA tape does not carry the broker, the customer, the order ID, or any stock and futures position held away from the print. Everything we label intent is an inference from structure, size versus open interest, strike geometry and pricing — never a disclosure.

Today's board makes the point better than usual: the largest premium on the tape belongs to a desk realising a nine-figure loss. Size is not skill.

Three habits worth more than any flow alert:

Wait for the open-interest confirmation before building a thesis on a trade whose size is below existing open interest. That is eleven of today's nineteen legs.

Never copy the short leg of someone else's financed hedge. Their long book is the reason the structure is survivable.

Position so that being wrong is survivable, because on any given trade you will be — including when you are copying someone who was right last month.

Nothing here is investment advice. Options carry the risk of total loss of premium, and short option positions carry risk far beyond the premium collected.