1. 20:29, One Line
That night I typed a line: one false step brings everlasting regret.
Not a figure of speech — a sigh from the gut. Four numbers were on the screen:
| Maintenance margin / net liq | 96.3% |
| Excess liquidity | $210.24 |
| Cash | $140.42 |
| To NOK earnings | 2 days |
$210 is not a “buffer” — it is “a few breaths left.” For every 1% NOK falls, net liq and margin combined evaporate by about $192 (at a 4.15x delta-equivalent, net liq alone drops about $238; the $192 is the combined margin + net-liq change in the whatIf snapshot at the time, and the gap comes from the short put’s delta offset). Which means: one more 1% drop in the underlying and the broker’s liquidation routine fires. Not “might” — the program executes automatically. It does not ask about your strategy, it does not look at a LEAP with eight months left; it just hits sell.
Worse was the cash: $140. That number meant I could not even pay to buy back the five puts about to expire — roughly $400 (5 × $0.80 × 100).
You do not need to lose everything to die. You only need to have no cash at the wrong moment.
2. The Cause: Those 15 Contracts
Let me assign blame properly: this was not a black swan. It was a goose I raised myself.
Of those 35 NOK LEAPs, 15 were added in two tranches. The logic was plain — the underlying was falling, so I would trade around it and average down.
After the first add it bounced and the account recovered. I logged it mentally: I read that right.
After the second add it kept falling.
Identical actions, opposite outcomes. The first time I called it skill; the second time I learned it was luck. Diligence helps sometimes — but if the direction is wrong, it backfires.
Averaging down carries a hidden assumption: you have unlimited cash, and the underlying is certain to come back. The reality: my cash was down to $140, and the underlying had no idea what my cost basis was.
That day I understood: every averaging-down lot is a new bet placed to rescue an old bet. I thought I was managing risk; I was scaling up the position.
3. The First Counterintuitive Finding
The first thing I did was attribution — who exactly was eating the margin.
I assumed it was the stock. After all, 200 shares of RGTI sat on the books at a 36% unrealized loss; it looked like the obvious culprit.
Then I ran whatIf leg by leg, and the result ground my intuition into the floor:
| Leg | Margin used | Share of total |
|---|---|---|
| NOK 10C JUN27 ×35 (LEAP) | $3,037.52 | 54.9% |
| RGTI stock, 200 shares | $678.60 | 12.3% |
| NOK 9.5P ×5 | $670.88 | 12.1% |
| RGTI 14P ×2 | $557.36 | 10.1% |
Each long LEAP consumed $86.8 of margin.
It is not the ordinary kind of option you pay for in full and that uses no margin — under this account’s margin treatment it was simultaneously my largest asset (83% of net liq) and my largest margin sink, and it had no short call covering it.
That is not a PMCC. That is a 4x-leveraged directional bet.
Later I did the arithmetic: net delta exposure was 4.15x net liq. A 20% drop in the underlying should cost 80% of net liq. That month NOK fell from 11.05 to 8.85 — exactly 20%.
“The account got cut in half again” is not emotion. It is math.
4. First Cut: Close That 9.5P
One leg had to go first: five NOK 9.5 puts expiring October 23.
The problem was the timing, and it was extremely dangerous — earnings before the open on October 22, options expiring October 23.
A bad print and the puts go in the money; if it did not recover by expiry I would take 500 shares, requiring $4,750, against $140 of cash.
On paper, assignment is “buying a good company at a discount.” In an account, it is “buying goods you don’t want with money you don’t have.”
I bought back the five 9.5 puts at 0.80 — $400 notional, $2.71 commission.
Realized loss: $226.18.
It was the only operation that day that locked in a certain loss, and what it bought was margin falling from 86.3% to 76.3% (the $626 transfer came in before the cut; sequence in the next section) and excess liquidity rising from 210 to 1,528.
NOK later bounced from 8.85 to 9.03 and that put fell back to around 0.7. Sometimes the thought comes: see, you closed at a loss.
That is not luck, that is hindsight. At the time IV was 122%, two days remained, earnings sat ahead, and there was not enough cash to take assignment — any one of those alone argues for closing, and all four together leave no second option.
In a crisis the right decision is often not the optimal one, but the only one that keeps you alive to make the next decision.
5. SGD 800: A Transfusion Is Not a Cure
In between, I wired SGD 800 into the account, about $626.
This month probably means a few days of instant noodles.
The numbers slapped me back quickly:
| Before deposit | After deposit | |
|---|---|---|
| Excess liquidity | $210 | $917 |
| Margin ratio | 96.3% | 86.3% |
What $626 bought was pushing the liquidation line from “a 1% drop” to “a 4.8% drop.”
At the time IV was 113%, and the market was pricing ±20% of movement for earnings week.
A 4.8% buffer against 20% of movement — that is not safety, that is commuting a death sentence to a reprieve, with two days on the clock.
Fresh capital buys time. It does not buy direction.
6. The Second Fork
With the puts closed, the real decision was on the table: what to do with 35 LEAPs.
Two roads:
A. Sell covered calls — collect premium, release margin, turn a naked long into a PMCC. B. Sell part of the LEAPs — take the loss, raise cash, cut leverage outright.
I ran the numbers on A first: the broker pairs the long LEAP with the short call into a spread automatically, and after selling the short call the margin actually came out net released — $11 to $37 per contract.
Selling the call both collects cash and returns margin. One action, two payoffs.
But on the strike choice, the AI got it wrong. More on that below.
7. The Math of Getting Back to Even
Start with the more basic question: can these 35 still get back to even?
Cost 2.235, last 1.36, down 39%.
| Break-even basis | NOK needed | Move |
|---|---|---|
| Sell partway to break even | 10.54 | +18.6% |
| Break even at expiry (K + cost) | 12.235 | +37.6% |
And then theta, the number nobody wants to look at:
| Point in time | DTE left | LEAP value (NOK unchanged) |
|---|---|---|
| Now | 240 | 1.33 |
| Month 3 | 149 | 0.96 |
| Month 5 | 88 | 0.65 |
| Month 8 | 0 | 0.00 |
If NOK goes sideways for eight months, all 35 of these go to zero.
That “zero” is not a scare tactic; it is what option pricing requires: a LEAP’s value = intrinsic value + time value. NOK at 9 against a 10 strike gives zero intrinsic; eight-plus months out, time value goes to zero too. What you hold is not “an asset” — it is a lottery ticket with an expiry date.
So there is exactly one path back to even: NOK has to rise. Collecting premium only reduces the bleeding while you wait.
8. The Council: Four AIs and a Chairman
At this point I did something I had never done before: I threw the decision to four AIs in a council.
The agenda was specific: expiry at 31, 59, 87, 150 or 178 days? Strike at 11, 12 or 13? Size 20, 25 or 35? Split into pre- and post-earnings tranches, or not?
All four members showed up. The debate ran 166 seconds.
Verdict: sell 25 of the April 16 11 calls, no split.
The reasoning was hard: among the combinations actually tested, only this one pushed margin down to 48.6% (March 19 was 50.4%, and 20 contracts was 53.7%).
The opposing side added one more cut, and it was brutal:
Selling a covered call only converts a “naked long” into a “long with a cap” — not enough protection on the way down, losses locked in on the way up.
The chairman finally labeled the conclusion in four words: lock the loss, buy survival.
One more entry for the record: the chairman said “reducing the LEAPs will not release maintenance margin.” That was wrong. I tested it — every 10 contracts sold released $943. The error came from textbook Reg T logic, while my account is charged margin on long LEAPs.
An AI can reason from a correct principle to a wrong conclusion, because it does not know how your broker classifies your account.
9. The AI Was Wrong Twice
The council gave an answer, but the AI’s recommendation was wrong twice, and that has to be written down.
Error one: the AI recommended 11.
The stated reason was “the 11 annualizes 18 to 20%, the 12 only 9 to 13%” — choosing a strike by how much premium it collects.
Yet in the same analysis it had plainly written: “the short call strike must be ≥ LEAP strike + net cost = 12.235.”
The AI knew where the break-even line was, then dropped it at the comparison step.
| 11C @0.74 | 12C @0.503 | |
|---|---|---|
| Sideways / down (to zero) | +0.74 | +0.503 |
| Above strike (assigned) | 11−10+0.74−2.235 = −0.495 | 12−10+0.503−2.235 = +0.268 |
To collect an extra $0.237 of premium, the AI took on $0.763 per share of locked-in loss risk.
In money: $592 more collected, $1,908 more lost. A schoolchild would not take that trade.
Error two: the AI’s roll model omitted theta.
The AI had first run a roll projection on “selling short calls 3–5 times,” concluding “rolling 8 times collects $4,550 cumulatively, and even assignment still leaves +227.”
That number was wildly wrong — the correct value is −3,988 — a gap of $8,538 from the +4,550 it claimed.
Where was the error? The AI added same-day quotes from different expiries directly. If you sell the 11/20 call on October 20, you cannot simultaneously sell the 11/6 — that stretch of option life no longer exists. And it assumed the LEAP’s value stays constant across those eight months.
Option value scales with the square root of time, not the sum of time.
Both errors were the AI’s — the first a mismatch of standards, the second a methodological flaw. And both times I pulled it back with common sense:
“April 16 is too long.” “The 12 is better than the 11. Why are you pushing the 11?“
10. Why 12 Beats 11: One Formula
The real condition for not losing on assignment is not strike ≥ 12.235, but:
K_short ≥ (K_long + cost) − premium = 12.235 − premium
| Candidate | Premium | Strike needed to break even | Actual | Verdict |
|---|---|---|---|---|
| 11C | 0.74 | 11.495 | 11 | ❌ short by 0.495 |
| 12C | 0.503 | 11.732 | 12 | ✅ |
| 12.5C | 0.30 | 11.935 | 12.5 | ✅ |
| 13C | 0.17 | 12.065 | 13 | ✅ |
Premium is not income; it is a prepayment on an obligation. It can pull the break-even line down a little, but it cannot pull it $0.5.
11. The Final Blow
Forty minutes before the close, I finally placed the order.
SELL 25 × NOK 12C MAR19’27 @ 0.503
| Premium | $1,257.50 |
| Cash | $364 → $1,620 |
| Maintenance margin | $5,249 → $3,939 |
| Margin / net liq | 73.8% → 58.0% |
| Excess liquidity | $1,868 → $2,857 |
| Delta-equivalent | 2.4x → 1.64x (the basis after the put buyback had already brought it down from 4.15x) |
The 10 LEAPs left behind: if earnings send it up, those 10 capture the full upside; if it drops, the loss has already been cushioned in part by that $1,257.
Even assignment makes money: (12−10) + 0.503 − 2.235 = +0.268/share, 25 contracts = +$670.
Not a “small loss” — a small gain. That is the most essential difference between the 12 call and the 11 call.
12. Before and After Those 7 Hours
| Metric | 20:31 | 03:20 |
|---|---|---|
| Net liq | $5,747.82 | $6,796.01 |
| Cash | $140.42 | $1,620.08 |
| Maintenance margin | $5,537.58 | $3,939.43 |
| Margin / net liq | 96.3% | 58.0% |
| Excess liquidity | $210.24 | $2,856.58 |
| Delta-equivalent | 4.15x | 1.64x |
Excess liquidity rose 13.6 times. Margin utilization fell 38.3 percentage points.
Strictly speaking, 58% is still above the 50% safety line — another $1,000 deposit is due at month end, and only after it lands (about 50.5%) does this formally clear. I am not going to call “close” a “done.”
But at least, whatever earnings delivered, this account would not be liquidated. That was already the best outcome available to me that night.
13. What That Night Taught Me
First: discipline is not a moral question, it is a math question.
Everyone says “be disciplined,” as if it were quitting smoking. But discipline is essentially this: finish the decision before the emotion arrives. That is exactly how those 15 contracts went in — at the time it did not feel like gambling, it felt like a “tactical adjustment.” Real discipline asks one question before adding: if I am wrong, does my account die?
That night the answer was yes. Which means those 15 should not have been added.
Second: an AI will give you a structurally wrong answer, in a very confident voice.
All four members showed up, the analysis was high quality, and it produced an accurate framing in “lock the loss, buy survival.” But its recommendation of 11 was an inferior solution, and its claim that “reducing the position does not release margin” was wrong.
AI is good at computing; it is not good at knowing how your broker classifies your account. It can enumerate the options for you, but it cannot make that last judgment for you — especially when the judgment depends on your cost, your psychology and your cash, things it can only hear from your own mouth.
AI can raise efficiency, but personal experience still matters.
Third: survival comes before getting back to even.
For a while my question was “how do I rescue these 35.” That question was wrong from the start — it assumed the 35 were worth rescuing.
The right question: “If I had $4,900 of cash today, would I buy 35 NOK LEAPs?”
If the answer is “no,” then reduce. The market does not know your cost, and it does not know how much pride you spent averaging down.
One last thing.
That night I got three things right: closing the put, choosing the 12 over the 11, and keeping 10 contracts in reserve.
The account lived.
But I know: had I hesitated two more hours that night, or asked one fewer time “what if I am wrong,” this post would not exist — the account would have been liquidated at 8.85, and I would be writing a different piece titled “How I Turned 8,800 Into 0.”
The point of risk management is not to keep you from losing money. It is to keep you with a chance to stay in the game.
This is a personal trade record. All figures come from IBKR TWS snapshots (clientId 96/99, 2026-10-20 20:31 – 2026-10-21 03:20). Nothing here is investment advice. Option-selling strategies can produce losses far beyond expected principal in extreme conditions; assess your own risk tolerance.