the machine room
a $4.55 put became $10.75 because the engine was finally allowed to sit still — the read was ordinary, the hold was not.
oct 12 · why a put, why that one
The read was not clever. Four bearish flow sources on a name sitting under its own dealer flip, against a tape that was green everywhere else.
flow said down, the dealer map said the floor was close, and the stop was the morning’s high
Why the name. The candidate came in from a dark-flow divergence screen, and every flow source the engine reads agreed with it: the premarket options state, two confirmed-flow rows and the divergence screen itself, all four bearish, none bullish. The dealer map put the gamma flip at $79.35 and the biggest put wall at $78, so a stock at $79.81 was sitting just above the line where dealer hedging stops cushioning a fall. The broad tape argued the other way: SPY and QQQ both above their 21- and 50-day averages, VIX calm, regime tier 1. This was a put against a risk-on market, on one name’s own flow.
Why that contract. The selector looked at five contracts and took the $80 put expiring Oct 30, a hair in the money, delta −0.47, implied vol 63%, theta −$0.12 a day, open interest 6,567, spread $4.40 / $4.55. It paid the ask, $455 for one contract. Eighteen days gave a swing profile room; a minimum-30-DTE rail in shadow flagged it as too short and, being shadow, only wrote that down.
Where it was wrong. Invalidation was the opening-range high, $80.90. If the stock reclaimed that, the thesis was dead.
two nights, and the rule from oct 7 that made them possible
Until a week ago this contract would have been sold at Monday’s bell for about $4.10, a small loss, no matter what the read said.
a swing contract now exits on its own rails, not on the clock
Since September the options sleeve had flattened every position at the close, on the evidence of a Sep 16 post-mortem where carried options lost far more than same-day ones. On I lifted that for swing contracts, on a cap of two: they exit on take-profit, stop, invalidation or expiry, not because it is 3:50. CRWV is the first contract to ride that rule out to a win.
Day one went against it. The put marked as low as $4.05 at 15:38 ET, down 11%, and the stock printed $81.25 in the last half hour, through the $80.90 invalidation intrabar. It closed at $80.29, back under, and the rails only check on the :00 and :30 ticks, so nothing fired. It held at a loss into the night.
Day two paid, then gave half back. CoreWeave opened Tuesday at $76.27, down 5%, and the put opened at $6.30, +38%. By 12:30 ET the stock had bounced to $78.94 and the put fell to $5.25. A trailing stop running in shadow would have sold right there, at +15%. It is shadow, so it only logged WOULD_EXIT:trail_lock, and the take-profit at +50% had not been reached. The put closed around $5.90 with the stock at $77.80, under the $78 put wall. Second night.
| ET | What happened | Put mark |
|---|---|---|
| Mon 13:35 | ENTER 1 × CRWV 80P 10/30, stock $79.81, invalidation $80.90 | $4.55 paid |
| Mon 15:38 | Trough. Stock prints $81.25 late, closes $80.29, held overnight | $4.05 · −11% |
| Tue 09:30 | Stock opens $76.27, down 5%. Shadow velocity_take says sell | $6.30 · +38% |
| Tue 12:34 | Stock bounces to $78.94. Shadow trail says sell at the floor | $5.25 · +15% |
| Tue close | Stock $77.80, under the $78 put wall, held a second night | ~$5.90 |
| Wed 09:31 | Stock gaps about 8% lower. Live take_profit_50% fires | $9.95 · +119% |
| Wed 09:32 | First limit $9.45 cancelled unfilled as the bid ran to $10.70; re-priced to $10.70 | filled $10.75 |
oct 14 · the gap did the work, and the order chased it up
The +50% take-profit is the bluntest rule the sleeve has. On Wednesday the stock opened so far through it that the rule fired at +119% and sold at +136%.
first double in 24 trades, and the largest single options gain in the ledger
CoreWeave gapped roughly 8% lower before Wednesday’s open. The first manage pass at 09:31 ET marked the put at $9.95 and the live take-profit fired. The first sell limit went in at $9.45, the bid at that moment. Within seconds the bid had run to $10.70, so the order was cancelled with nothing filled and re-sent at $10.70. It filled at $10.75.
The record it joins. 23 closed options trades before this one. The best of them were Chevron at +60.6% and Exxon at +51.5%, both in August, and LRCX at +50.4% on Sep 22. Nothing had ever doubled. CRWV is the first trade over +100%, and at +$620 it edges LRCX’s +$610 as the biggest dollar winner too. The sleeve is now 8 winners in 24.
Why “letting it fly” is the right name for it. Every exit the engine sells itself on, the trail, the velocity take, the bell, would have closed this trade smaller. The bell rule would have taken a loss on Monday. Velocity would have taken +38% on Tuesday morning. The trail would have taken +15% at Tuesday’s low. What held it was the two simplest pieces of the system: a thesis stop that never got hit on a tick, and a profit target the market jumped straight over.
what one double does not establish
One trade is one trade, and two things about this one should make anyone uneasy rather than pleased.
the hold survived partly because nobody was looking at the wrong moment
The thesis was technically broken on day one. The stock traded at $81.25, above the $80.90 invalidation, in Monday’s last half hour. The rails only evaluate on ticks, and the tick saw $80.29. A continuous check would have sold this put at a loss. That is not a reason to check continuously; it is a reason not to call the hold a skill.
Overnight, a long option has no stop at all. The carry rule says so in its own notes: the first evaluation after the bell is the next session’s first tick. This time both gaps went the right way. The Sep 16 numbers that kept carry off, held options losing far more than same-day ones, have not been answered by one win.
And the shadow rails lost this one. Trail and velocity would both have sold early. That is one data point against them, not a verdict; a review is out to replay every rule on this trade and on the whole shadow ledger before anything gets promoted or loosened.
What would show this was more than luck. The carry rule has run for one week and produced one closed carried trade. The test is the next ten: whether carried swing contracts, as a group, beat the same contracts sold at the bell, net of the gaps that go the wrong way. The +136% is the headline. The 8-for-24 is the record. Ask again at thirty.
← the previous build note · Sep 22 · the machine room · build log & archive