Six names, sixteen legs, $347.60M of premium — and not one of those legs was a lit print.

Every single trade on today's board was negotiated away from the public order book: floor blocks, a cross, and two auctions. There was no sweep, no lifting of offers, no urgency anywhere. On trades like these the "buy" or "sell" label attached to each leg is a bookkeeping convention, not an observation — a broker matched two parties who had already agreed a price, and the exchange allocated that one package price across the legs afterwards. We say this most days. Today it applies to 16 of 16 legs, so it is worth saying first.

The second thing worth saying first: three of today's six trades are shaped to pay for moves the option market itself is not pricing. IWM at 7.5 : 1, BSX at 13.7 : 1, and NVDA at up to 41 : 1 — and all three break-evens sit outside the implied-move cone for their own window. That is not a coincidence, and it is not a prediction either. It is what cheap, bounded, low-probability structures look like when institutions want a specific outcome covered without paying much to cover it.

Two more are most likely positions coming OFF, not going on. MSFT's $35.97M "credit" and DRAM's $14.5M one both landed on strikes whose open interest has been frozen for a month — and then took prints equal to that entire open interest. That is the fingerprint of a liquidation, not a new view.

And one is not really an options trade at all. SPCX printed $100.87 million of headline premium containing $279,000 of time value — a 362× overstatement. Those are deep in-the-money puts trading at parity, which makes them synthetic stock, not optionality. We flag it rather than counting it.

⏳ Nothing here is settled until tomorrow morning. Ten of today's sixteen legs printed at or below their existing open interest, which means today's tape cannot tell us whether a position was opened or closed. Come back at ≈06:30 ET on the next trading day — the open-interest snapshot is the only thing that resolves it, and we score every published prediction in public, including the ones we get wrong. There were two of those this morning.

🔁 OI Review — Monday's Predictions, Scored

The 06:30 ET open-interest snapshot is in, and Monday's board resolved 17 legs open, 3 closed, 1 partially closed, 5 as pure transfers and 2 permanently unverifiable. We publish a capture rate on every leg before the answer exists, then come back and score it in public. Two of those calls were wrong, and one of them inverts a trade we described as insurance.

🔄 The inversion: SPY's 91,400-lot September $665 put was a CLOSE, not a new tail

We described Monday's SPY package as structural premium collection — 32,600 March-2027 $700 puts sold for $45.84M — financed against a small short-dated insurance tail: 91,400 September $665 puts bought for $2.65M. The March sale checks out: open interest went 5,837 → 38,435 (Δ +32,598, 100.0% capture), so $2.28 billion of assignment exposure below the market really was opened.

The tail was not. The $665 line had sat between 105,455 and 110,616 contracts every session for three weeks. On Monday it went 106,542 → 14,520 (Δ −92,022) — an 87% reduction, almost exactly the printed size. Open interest falls only when both counterparties are closing, so those 91,400 contracts extinguished a line that already existed instead of creating a new one. No tail was bought, and the September window is uncovered by anything that opened on August 31.

Neighbouring September $665-area strikes moved 0.1% to 5.5% over the same session while this one lost 87%, so this is strike-specific, not an exercise sweep. What open interest still cannot say — and never will on a negotiated print — is which counterparty initiated, or whether the desk closing the $665 line is the same one that sold the March $700 puts.

🔄 SNOW's "$12.81M credit" was neither a new bear spread nor a clean unwind

SNOW's January-2027 $340/$420 print, two days before earnings, was published as an open question: a new bear call spread, or an old bull spread cashing out? Open interest fell on both legs — $340C 6,200 → 5,530 (Δ −670) and $420C 6,247 → 5,313 (Δ −934) — but by only 11.8% and 16.4% of the printed size.

A new bear spread needed both lines to rise toward ≈11,890 and ≈11,937. They fell, so that reading is refuted: the $12.81M credit did not buy anyone a new bearish structure into earnings. A clean unwind needed a fall toward ≈510 and ≈557; roughly 85–89% of the standing open interest is still there, so that reading is not confirmed either. What is left is the third branch — a transfer with a small net close on top. Risk moved between books; very little left the market. The one thing we predicted correctly is that the two legs would move together, and they did.

❗ The call we got wrong: QQQ's November $680 put

We named QQQ's November $680 put "the leg to watch closely", led with the transfer branch, and put new opening at 0–25% because the arithmetic margin was only 18% — below the threshold where the size-minus-prior-OI floor is reliable. It opened at 102.7% (16,331 → 36,861, Δ +20,530).

The rule behind that call is sound and we are keeping it: at an 18% margin, the day's tape genuinely cannot prove an open. But "unprovable" is not "probably a transfer", and we attached the wrong lead branch to it. All four QQQ legs opened at 100% or better, so both collars — the full $4.15B of notional — are confirmed as genuinely new hedges.

✅ Confirmations

• NRG — the roll-down is real, on all four legs. $165C 26,489 → 5,996 (−20,493) and $200C 26,358 → 6,051 (−20,307) came off in near-matched size while $145C 240 → 20,704 (+20,464) and $170C 2,167 → 22,161 (+19,994) went on in near-matched size. Control strikes across the December chain were flat, so no exercise sweep. Note the closes are partial — ≈4,500 contracts per line are still outstanding.
• IBIT — the two legs split, exactly as we warned they might. $45C 42,634 → 19,653 (−22,981, a 77% close) confirms the profit-taking read on the leg where the classifier was confident. $57C 34,578 → 34,506 (Δ −72) is a pure transfer, so the close read on that leg is not confirmed. The BlackRock-overwrite-programme branch needed open interest to rise; it rose on neither line.
• IEF — both legs on their lead branches. $92.50C 1 → 35,254 (+35,253, 100.7%) opened in full; the $93C floor block, which we said would be a transfer against 38,346 of standing interest, moved the line by four contracts.
• XEL 222 → 10,070 (+9,848, 98.5%) — inside our published 98%–100% band.
• LPX $65P 9 → 5,010 and $55P 4 → 5,004 — both within one contract of target. RUM $8C and $11C both opened from zero at ≈100%. TEX 0 → 3,500, exact to the contract.
• The rest of SPY resolved as published: all three November 30 collar legs opened at 102%, confirming the $1.536B collar and its two-basis-point cost; the September 30 $775 put opened at 99.9%, confirming the protective-put roll; the October 30 $650 put opened cleanly. The October 16 $630 put did not — Δ +7 on a 30,000-lot print, the transfer branch we led with, so of the "two cheap far-downside tails" only one is new. Both August 31 puts expired that afternoon and have no next-day test, exactly as disclosed in advance.

What changed, and the standing lesson

Monday's board was published at $436.53M gross / $96.67M net. Nothing about those figures changes — they come off the trade tape, not open interest. What changes is what the money did: $2.65M of the SPY package removed an existing position rather than adding protection, and SNOW's $12.81M credit created almost no new January-2027 exposure at either strike. Roughly a fifth of Monday's legs moved risk between books without changing how much risk existed.

The lesson this session taught twice: a low arithmetic margin means the answer is unprovable today, not that a transfer is likely. We led with transfer on QQQ's $680 put and it opened at 102.7%; we led with transfer on SPY's $665 put and it turned out to be a full close of a 106,542-contract line. Both times the honest statement was the same — we cannot tell yet — and both times attaching a favourite to it was the error. The fix is cheap and we are applying it: on any large size ≤ OI leg, pull two to three weeks of open-interest history before writing, not just the prior session. A line that has been flat for three weeks and then collapses tells you more than any size-versus-OI ratio can.

📋 The Board at a Glance

Net is the debit paid or credit collected per package — the money actually at risk. Gross double-counts every spread and, on SPCX, is almost entirely intrinsic value rather than premium.

🔍 What's actually interesting today

The IWMIWM-- trade is bigger than it looks, and the timing is specific

166,080 contracts on each leg — $19.6M of cost controlling a spread worth up to $166 million. Prior open interest on the $275 leg was 588 contracts, so this is unambiguously new. Break-even is $283.82, −2.27%, and full payoff needs −5.34% inside ten days.

Set that against what the market charges for the same window. The implied move to September 18 — seventeen days — is ±2.75%. So the break-even sits essentially on the edge of the market's own 17-day range, with only ten days to get there, and the full payoff level is outside that range altogether.

Two details sharpen it. First, IWM has already fallen four straight sessions, from ≈$300 to $290.57 — this was bought into a slide, not ahead of one. Second, the September Fed meeting (Sep 15–16) falls after this option expires. So it is not a bet on the Fed's decision. It is a bet on the two data prints that will price that decision first: August payrolls (≈Sep 4) and August CPI (≈Sep 10–11) — neither date confirmed against the primary calendar, and we say so in the article.

SPCX: when a $100 million headline contains $279,000 of opinion

Four blocks, all deep in-the-money puts, all printed at or below parity. Bought at those prices, a put is not really an option — it is synthetic short stock. Struck against the $143.25 spot, the four legs work out to synthetic entries of $143.30, $143.05, $143.90 and $143.50 — every one within 65 cents of the market.

Why now is the interesting part. SpaceX listed on 2026-06-12 and used a tiered lock-up rather than a single 180-day one. Tranche 1 released 911.5M shares on Aug 6. Tranche 2 released 319M on Aug 17–20 — and the stock fell ≈6%, below its $135 IPO price. Tranche 3 unlocks September 9. September 18 is the first standard monthly expiry after it.

A holder facing tranche 3 has just watched the last unlock go badly. That is the leading hypothesis — and it is a hypothesis: the lock-up language itself could not be retrieved, and no lock-up date on or near September 18 could be sourced.

MSFT and DRAM: the tell is a line that hasn't moved in a month

Both trades landed on strikes whose open interest has been frozen for 21 sessions. MSFT's four lines held bands of 0.0% to 0.5% — the $705 call sat at exactly 8,093 contracts for eight consecutive sessions — and then each took a print equal to its entire size, twenty seconds apart. DRAM's $45 call held a 0.2% band and took a 20,000-lot print.

A line that does not move for a month is a dormant position with one owner. When it suddenly trades its whole size, the simplest explanation is that the owner left. We are saying those lines will FALL tomorrow. If they rise instead, we are wrong and both credits are new bearish positions. That is the whole test.

BSX: a 27.7% break-even that isn't as wild as it sounds

The $60/$75 call spread needs +27.7% to break even by December — which sounds heroic until you see where the stock has been. Boston Scientific is −49.6% year to date and −54.43% over 52 weeks, from $109.50 to ≈$48.

Its 200-day moving average is $67.59. So the $60 strike sits below the stock's own average price over the last ten months, and the consensus analyst target of $62.69 already clears it. The CEO bought 186,240 shares in the open market on August 5.

There is no takeover here — and the M&A that exists runs the other way: BSXBSX-- is the acquirer, having announced a $14.5B purchase of Penumbra in January. This is mean reversion on a broken compounder. The honest counterweight: the 45-day implied move tops out at $54.49, so even the $60 break-even is outside the market's own cone. Cheap and plausible is not the same as likely.

👥 What this means for you

🎰 If you trade for the big score. Today's board is full of exactly what you like — 7.5:1, 13.7:1, up to 41:1 — and it is worth noticing how the professionals built them. Every one of those payoffs is capped. They sold a further-out strike against the one they bought, which cut the cost and gave up the fantasy tail. NVDA is the clearest lesson on the page: that structure pays up to ≈$500M if the stock lands near $100, and exactly nothing if it goes to zero. Someone deliberately gave up the apocalypse scenario to make the realistic one cheap. A capped trade you can afford to be wrong on five times beats an uncapped one you can afford once.

📈 If you swing trade. The levels worth writing down are where the gamma actually sits, not the strikes in these trades. IWM: spot is sitting on a large negative-gamma shelf at $290 while the heavy positive walls are at $304–$305 — worth understanding, because negative gamma below the price tends to make declines move faster, not slower. NVDA: $215–$220 is a very strong cluster right at spot. MSFT: $500 is the heaviest shelf on its board and the stock is right on it. Those matter far more to a two-week horizon than any 2027 strike.

💰 If you sell premium. Two cautionary tales today. SPCX shows a $100.87M "premium" number containing $279,000 of actual time value — if you are screening for fat premium, that is the kind of number that will fool you. And MSFT's $35.97M "credit" is most likely not income at all; it is a two-year-old bullish position being cashed in. Before you treat any credit as an edge, strip out the intrinsic value and ask whether the seller was opening or closing.

🌱 If you are new. The single most useful idea on this page is intrinsic value versus time value. A $205 put on a $143 stock is worth at least $62 no matter what — that is intrinsic, it is just arithmetic. Anything above that is time value, and time value is the only part that carries an opinion. Today someone traded $100.87 million of those puts and the opinion inside was worth $279,000. Learn to subtract the intrinsic before you get impressed by a headline. It will save you more money than any strategy.

🧯 Risk control and patience

Everything above describes what other people did with their money. None of it is a recommendation, and most of it is not yet provable.

• Sixteen of sixteen legs were negotiated blocks. No aggressor exists anywhere on this board. Do not read urgency into a number that does not contain any.
• Ten legs printed at or below existing open interest. For those, we cannot tell whether a position was opened or closed — and the two point in opposite directions.
• This morning we published two calls that were wrong, both for the same reason: we attached a favourite branch to a leg the tape genuinely could not resolve. The fix — pulling weeks of open-interest history rather than one session — is what produced today's strongest read on MSFTMSFT--. Being wrong in public is how the method improves.
• Three of today's structures need a move the option market says is unlikely. They are cheap because they are unlikely. That is the trade, not a flaw in it — but do not mistake a 41:1 payoff for a 41:1 opportunity.

Position sizes here run to nine figures. Yours does not have to. The most valuable habit in this business is being willing to do nothing on a day you do not fully understand what you are looking at.

⏳ Tomorrow morning, ≈06:30 ET

We publish a capture rate, never a single number. Three outcomes are always live on any leg: it opened in full, it closed, or it landed in the middle as a partial close or a pure transfer between two existing holders — and the middle is common, not exotic.

MSFT is the most falsifiable call on the board and we are stating it plainly: four lines each holding ≈8,000 contracts should FALL. If they rise, we are wrong.

This newsletter describes trades that appeared on the public options tape. It is not investment advice and nothing here is a recommendation to buy or sell. Options can lose their entire value. The tape cannot tell us who traded, why, or what else they hold — and we say so every time it matters.