1. 01:33, Two Numbers
At 01:33 there was one number on the screen: 59.7%.
That’s maintenance margin as a share of net liquidation value. Three hours earlier it was 75.8%.
Between the two sits a single trade: I sold 200 shares of RGTI at 9.665, for a loss of $1,639.46.
That’s not a figure of speech. Over the week the account went from 6,215 to 4,223 — −32%. NOK from 9.31 to 7.48 (−19.7%), RGTI from 12.44 to 9.62 (−22.7%).
I sat in the chair with one sentence looping in my head: I thought I was collecting rent. I was levering up and betting on direction.
2. This Week, Both Legs Broke at Once
| Date | NOK | RGTI | What happened |
|---|---|---|---|
| 10/29 | 9.31 | 12.06 | Everything looked like it was coming back |
| 10/30 | 8.56 | 11.25 | Semis dragged the AI complex down |
| 11/02 | 8.30 | 10.67 | Quantum sold off as a group |
| 11/03 | 8.48 | 11.09 | I bought 10 more LEAPs |
| 11/04 | 8.26 | 11.21 | Flat, and I called it a bottom |
| 11/05 | 7.48 | 9.62 | −9.4% / −14.2% on the day |
NOK’s earnings weren’t actually bad: revenue beat, AI & Cloud sales +123% year over year, comparable operating profit +35%. The market sold something else — free-cash-flow conversion guidance cut from 55–75% to 35–55%, and a CEO line on the call: “this isn’t just a shortage, it’s structural cost inflation.”
The income statement beat. The cash flow statement got downgraded. The market always sells the latter.
As for RGTI, there was no company-specific bad news. Quantum drifted lower for days after IonQ’s large secondary, and RGTI made a 52-week low. “Down a lot” has never meant “cheap.” I’ve written that sentence before, then proved it again on this one.
3. 11/03: I Added Again
On November 3, cash went from 1,992.91 to 924.13. The difference, 1,068.78, bought 10 NOK JUN’27 10C at an average of about 1.07. The 45-lot average came down from 2.235 to 1.975.
The reasoning at the time still looks entirely sound: the average dropped 12%, the underlying sat at 8.48, the LEAP had seven months left, delta was low, the cost was cheap, and “this is the last one.”
“This is the last one” is the most common line among people averaging down, and the least likely to come true.
I had written the rule: above 50% margin, reduce only, no adds; the only buy permitted is buying back my own short. I knew the rule. I had quoted it the night before. Then I opened the phone and placed the order.
It wasn’t impulse. Impulse has no reasons; I had a complete set. That is far more dangerous than impulse.
4. 12:23, I Sold 200 Shares of RGTI
Filled at 9.665. The day’s low was 9.19, the close 9.62. I sold above the day’s average price — not foresight, luck.
Cost 17.86, −44%, −1,639.46.
Why this one and not something else:
- It was the most liquid leg in the account. 200 shares clear in seconds, and the spread is rounding error.
- There was a January 15 14P sitting behind it. Strike 14, underlying 9.6 — deep in the money. If it got assigned I’d have to produce $2,800 inside two months, and the account held $921.
- It was the leg with no strategic role. I never built this position on purpose; the stock arrived through assignment.
One small thing happened after the sale, and it changed what the whole thing was: that 14P went from “an assignment accident waiting to happen” to a genuinely cash-secured put. $2,853 of cash against a $2,800 exercise obligation — just covered. (Forty-six minutes later I rolled it anyway — see the addendum.)
Same position, same unrealized loss, went from “will blow up” to “won’t blow up,” purely because there was cash in the account.
5. Counterintuitive: Closing a Position Up $1,035 Would Make the Account Riskier
Now the actual point.
What I hold is a NOK PMCC: 45 long LEAPs (10C JUN’27) + 35 short calls (12C MAR’27).
The 35 shorts: $1,631 of premium collected, now worth $596 — $1,035 unrealized gain, 63.4% captured.
Every options-selling textbook says the same thing here, in unison: take it off between 50% and 80%.
So I had it run the numbers on buying all 35 back right now:
| Metric | Now | After buying back 35 |
|---|---|---|
| Cash | $2,853 | $2,188 |
| Maintenance margin | 59.7% | 71.3% |
| Delta equivalent | 2.25x | 3.15x |
| Delta capacity | — | −9 contracts (over the cap) |
Three numbers, all pointing the wrong way.
The reason is simple and easy to skip over: those short calls are the cover for the 45 LEAPs. Buy them back and the 45 LEAPs become naked long.
IBKR’s margin algorithm doesn’t read your intent, your stop discipline, or the articles you’ve written. It reads structure. Change the structure, and the risk changes.
There’s a layer fewer people talk about: the short leg is the only thing in this account paying me every day (+$8.40/day), while the long leg bleeds (−$13.95/day). Closing a position that’s already up 63% means shutting off the only inflow with my own hands, just so the statement looks a little better.
In the end: I didn’t close it.
6. Will the Profit Shrink on a Bounce? Yes — but You’re Measuring the Wrong Leg
My original worry: “If I don’t close it, the first bounce eats that 1,035.”
True, but it describes one leg.
| NOK | Move | Long leg Δ | Short leg Δ | Net Δ |
|---|---|---|---|---|
| 8.00 | +6.7% | +$877 | −$286 | +$591 |
| 8.50 | +13.3% | +$1,875 | −$653 | +$1,222 |
| 9.50 | +26.7% | +$4,211 | −$1,650 | +$2,561 |
| 11.40 | +52% | +$9,728 | −$4,552 | +$5,176 |
The short leg’s profit does bleed as it rallies, but the long leg’s delta is 3.2x its size. The structure is net long 1,131 shares. A bounce is net profit, not net loss.
And “close it now, sell it back after the rally” carries a hidden assumption: that if the stock goes up, the option won’t get more expensive.
It gets more expensive.
So the right move isn’t “close now, re-sell later.” It’s wait for an actual move toward 11.4 — 5% below the short strike — and roll up in a single combo ticket: buy back the 12C and sell the 13C or 14C in the same order. That is selling higher. Sitting on a naked long exposure and waiting for the outcome you want is not.
7. I Was Wrong, and So Was the AI
This isn’t a success story, so this section has to be finished.
① I was wrong: the 11/03 add. A second add inside a freeze window. Complete reasons, defensible direction, disqualifying process.
② The AI was wrong: it logged a fat finger as a discipline breach. On October 16 I bought 2 RGTI 15.5C, and the system recorded it as “directional buy during a red-line window — violation.” It stayed on the books for three full weeks. Only today did I get around to correcting it: I had the order direction reversed — I meant to sell calls. One fat finger, logged as a discipline problem. It colored how I judged myself, and the wording of several reports after it.
③ The AI was wrong: a flipped sign turned the leverage into 3.83x. The first pass at portfolio Greeks put a plus sign on the short leg’s delta and produced “3.83x leverage.” The real figure is 2.25x. Had that number gone out without review, I’d have believed the account was near 4x levered and made a completely different decision, in a panic. A wrong number is more dangerous than no number.
④ The AI’s advice was wrong: buy back that 14P “to reduce risk.” I said no, and the reason was a number it had calculated itself: that put had $45 of time value left — 95% of it was intrinsic. Buying it back isn’t reducing risk. It’s closing a directional position at the low and pretending I’d done risk management.
Of the four errors, one was caught by me, one was caught by me, and two were caught by common sense. The AI isn’t for believing. It’s for checking.
8. The Result
| Metric | Before the sale (01:00) | After the sale (01:33) | Change |
|---|---|---|---|
| Net liq (NLV) | $4,361 | $4,223 | −$138 |
| Cash | $921 | $2,853 | +$1,932 |
| Maintenance margin | 75.8% | 59.7% | −16.1pt |
| Excess liquidity | $1,057 | $1,700 | +$643 |
| Delta equivalent | 2.51x | 2.25x | −0.26x |
| RGTI stock exposure | 200 shares | 0 | flat |
Net liq dropped only $138 — I sold at the market, and those losses were already marked. But the character of the account changed: from “a margin call could land any minute” to “there’s a buffer, there’s room, I can wait.”
Note that delta fell only 0.26x. The bulk of it was never those 200 shares; it was the 45 LEAPs. Which is the next thing on the list.
9. What Comes Next: I Stop Relying on Memory
I installed a trigger list so that every morning I don’t have to reconstruct what I decided the day before:
| Trigger | Action |
|---|---|
| Margin ≥ 65% / ≥ 75% | Sell 10 / 20 LEAPs |
| Delta ≥ 2.5x | Cut LEAPs; every other action is noise |
| NOK ≥ 9.00 | Bounce zone — re-check margin (appreciation raises the requirement) |
| NOK ≥ 10.50 / 11.00 | Evaluate roll-up / roll is mandatory |
| NOK ≥ 12.00 | Assignment risk: roll first, or exercise the LEAPs the same day to cover |
| NOK ≤ 7.00 / 6.50 | No panic trimming / re-underwrite the thesis |
| RGTI ≤ 10.00 | February 10P goes in the money: check cash ≥ $2,000 first, then take assignment or roll again |
| RGTI ≤ 9.00 | Re-underwrite: below a 52-week low there is no such thing as “cheap” |
| Short DTE ≤ 30 / ≤ 7 | Lock in the plan / final handling window |
And one hard date on the calendar: March 19, 2027, the short calls expire.
If NOK is still under 12 that day, the short leg goes to zero and I’m left with 45 naked LEAPs — a rent machine degraded into a pure directional bet. The next CC has to be sold before that date, or this PMCC is a PMCC in name only.
While I’m here, one counterintuitive fact: this account’s margin risk sits on the upside, not the downside. The underlying drops, the long leg’s market value shrinks, and the margin requirement eases on its own. The underlying rallies to 10, the net position’s value goes from 2,414 to 5,664, and the requirement goes up. So “cut when it falls” is wrong. “Re-check when it rises” is right.
10. Three Reflections
One: the leg losing the most is usually not the dangerous one. The dangerous one is the leg that looks like it’s making money — because it’s the only one that tempts you to touch it. People want to “do something” about an unrealized gain far more than about an unrealized loss.
Two: every add inside a freeze window can be given a perfectly reasonable justification after the fact. That’s exactly what makes it dangerous: it isn’t impulse, it’s narrative. And narrative talks itself into agreement.
Three: memory is worst where unrealized losses are concerned. So write rules as triggers, not as resolve. Resolve gets ground down by the tape. Triggers don’t.
I lost a third of this account in a week. I didn’t get any smarter. I just wrote down the mistake I’m guaranteed to make next time, and handed it to something that won’t go soft.
Risk management isn’t there to keep you from losing money. It’s there to make sure you still have choices when you do.
Addendum · 02:09: I Rolled That 14P After All
After finishing the above, at two in the morning, I opened the order ticket one more time.
13:09 ET, RGTI at 9.65. One combo order (BAG):
- Bought back RGTI JAN 15 ‘27 14P ×2 @4.72 (paid $944.00)
- Sold RGTI FEB 19 ‘27 10P ×2 @1.83 (collected $366.00)
- Net cash −$581.10 (including $3.10 commission; cash $2,853.08 → $2,271.98)
- Old leg settled at −$393.48 (position average 2.7526; TWS session basis −$394.90 including fees)
In section 7 I turned down the AI’s “buy back the 14P to reduce risk” advice. That judgment still stands: 95% of it was intrinsic value, and buying it back means closing a directional position at the low.
But a roll is not a buy-back. The position is still there. I just looked at this put as two halves: the intrinsic half got re-struck (strike 14 down to 10), the time half got sold (35 more days).
| Item | January version (14P) | February version (10P) | Change |
|---|---|---|---|
| Exercise obligation | $2,800 | $2,000 | −$800 |
| Expiry | 1/15 (71 days) | 2/19 (106 days) | +35 days |
| Δ | −0.95 | −0.44 | exposure halved |
| Cash ÷ obligation | 1.02x | 1.14x | +0.12 |
| Share cost if assigned | $11.25/share | $10.14/share | −$1.11 |
The time-value line explains it best: buying back cost about $74 of time value at the fill price (0.37 × 200; the 01:00 mid-price read was only $45, and the gap is the spread cost), while selling brought in $366 (1.83 × 200, all of it time value) — roughly $292 of time value taken in net, paid for with 4 points of strike and one realized loss.
The cost, written plainly: an unrealized loss became a realized one. Net liq $4,223 → $4,249, and that difference is NOK marking, nothing to do with this roll. Maintenance margin fell from 59.7% to 56.0%, not because anything was earned, but because a deep-ITM 14P is no longer carried as “assignment could happen any minute.”
What I actually bought is one thing: optionality. If it really breaks $10 before February, $2,000 in cash is sitting there — assignment goes from “accident” to “an executable outcome.” If it rallies back, the put goes to zero and the $366 stays.
The new leg is Δ −0.44; by delta approximation, the odds of trading below $10 before February are roughly 40%. So this isn’t a solution — it’s swapping into a bill I can afford.
One aside: that RGTI 15.5C expiring 11/27 is down to $0.095 and will most likely be zero in 22 days. That $99 is the final installment on this directional call.
This is a personal trading record. All figures come from Interactive Brokers TWS live sync and fill-by-fill executions, and none of it is investment advice. Short options carry unlimited risk; assess it yourself.