The Nvidia Earnings Trade That Ends Before Earnings

NVDAOpened August 12, closed August 21, returned 44.9%. Nvidia doesn’t report for another five days.

POSITION CLOSED AUGUST 21, 2026. Entry at Interactive Brokers midpoints, August 12. Exit at closing midpoints, August 21. Frontwater Capital and accounts it manages hold positions in Nvidia.

Nvidia reports earnings on Wednesday. The calendar spread we wrote on our web page two weeks ago closed last Friday. It made a return of 44.9%. That number is now settled and nothing the company announces this week can change it.

This was an earnings trade in every meaningful sense: it existed because of the announcement, it was priced off the announcement, and it made its money from the announcement. It simply never waited around to hear what the announcement said. Sounds confusing, right?

Stated in advance, not after the fact

This trade was published on the Frontwater website on August 13, with the outcome unknown and Nvidia’s earnings still two weeks away. The strikes, the cost, the exit date and the risk were all disclosed then. Anyone can find a winning trade after it has finished winning. This piece follows one that committed to the position beforehand, and it reports what actually happened, including the part that went against us.

The reason it paid isn’t the one most people would guess, and it is the part worth reading. It was not that the stock behaved. The stock didn’t behave. It fell in every session of the final week and finished outside the range the structure was built around.

What the trade was

Two expiries, opened August 12 with Nvidia near $224.

Leg	Price	Net
Sell · August 21 · 215 Put	$1.58
Buy · August 28 · 215 Put	$3.88	$2.30
Sell · August 21 · 230 Call	$2.34
Buy · August 28 · 230 Call	$5.65	$3.31
Net cost, and the most it could lose		$5.61

Bid/ask midpoints, Interactive Brokers, August 12, 2026. Excludes commissions. $5.61 per share is $561 for the structure.

This is a double calendar spread. The legs we sold expire August 21, the quiet week before earnings. The legs we owned expire August 28, the week after. Each pair shares a strike across the two dates, so the position starts close to neutral on direction. It is a position on the relationship between two dates rather than on the price of the stock.

Only the August 28 expiry contains the earnings announcement. The August 21 expiry covers a week in which, barring an accident, very little is scheduled to happen.

You trade the calendar, not the earnings.

The maximum loss was $5.61 per share, fixed and known from the moment the position went on. That is the single most useful thing to know about the basic structure, and it is true for the version priced above. It is not true of the variant we actually ran in client accounts, which is covered on page 6.

What the quiet week actually did

Nvidia drifted down. No shock, no gap, no headline. A steady slide of about 4%, from $224.09 on the day we opened to $214.72 at the close on August 21, across five consecutive lower sessions. The worst single day was 2.34%.

Nine days, one direction

Nvidia daily closes from entry through expiry, against the range the structure was built around.

Blue band is inside the range, sand is outside it. The inset magnifies the final gap.

Look closely at where that line finishes. The structure was built around a range of $215 to $230, and Nvidia closed at $214.72. Twenty-eight cents below the lower strike. The range didn’t hold.

We are stating that plainly because the alternative would be to round $214.72 up to $215 and claim the band worked. It didn’t, and the more interesting fact is what happened anyway: the trade returned 44.9% with the stock outside the structure, moving against it, for five sessions running. If the position had depended on the range holding, that would have been impossible.

Why exiting before earnings is what paid

Two things happened to volatility in those nine days, and between them they are the whole return.

1

The front week got quieter

We sold the August 21 options at an implied volatility of 33.8%. What the stock actually delivered that week was 19.7%. The premium we collected was richer than the movement that arrived. The market had overpriced the quiet week, which is exactly the proposition we underwrote.

2

The back week got louder

Implied volatility on the August 28 expiry rose from 42.2% to 55.1% as the report approached and the market grew more anxious. The legs we owned were worth more at the exit than at the entry, purely because of rising apprehension about an event that hadn’t occurred.

Short the leg that got quieter, long the leg that got louder.

That spread is what paid. The direction of the stock was, in the end, a secondary concern, which is why finishing outside the range cost so little. Note what the second force means: we were paid by the market’s fear of the announcement, not by the announcement. Fear is available in advance. The result isn’t.

One asymmetry inside the result

A double calendar is really two positions sharing a ticket, and they disagreed. The put side returned about 175% because the stock came down to meet it and left our long put perfectly at the money. The call side lost roughly 45%, because Nvidia walked away from $230 and the call we owned drifted out of the money. The blended result was positive. It wasn’t uniformly positive.

Twenty-eight cents, and the ninety minutes after the bell

Finishing below $215 had a consequence. The August 21 put we sold closed in the money, and the clearing house automatically exercises any option that finishes as little as one cent in the money. On that rule alone, we were being assigned stock over the weekend.

Except that isn’t where it ends, and this is a piece of plumbing that rarely gets written about. The holder of an option has until 5:30 in the evening, ninety minutes after the closing bell, to instruct their broker not to exercise. In those ninety minutes Nvidia went back up. It reclaimed $215 by 4:17, sat at $215.05 at the 5:30 deadline, and drifted to $215.38 by evening. A rational holder of a 215 put, watching the stock trade above the strike, files to abandon rather than sell shares for less than the market will pay.

Why we booked the charge anyway

Our figures assume the put cost us its full 28 cents, as though we had bought stock back at $215 that we had sold at $214.72. In all likelihood most of those puts were simply abandoned and cost nothing. We would rather publish the conservative number and revise it up later than publish the flattering one and correct it down.

The lesson isn’t that after-hours drift is a risk management plan. It is that we got a favourable roll of the dice on something the original piece had already told readers how to control: buy the short leg back if it finishes near the strike. That instruction was right, it would have cost about 25 cents to follow, and the alternative meant handing the decision to somebody else and finding out on Monday.

The variant we actually ran, and what it risked

Everything to this point describes the straightforward one-lot structure, where the most that can be lost is the $5.61 paid. Because we already hold Nvidia in client portfolios, the version actually executed was a ratioed one, and its risk isn’t capped.

Ratioed variant	Price	Cash
Sell 4 · Aug 21 · 215 Puts	$1.58	collect $632
Buy 2 · Aug 28 · 215 Puts	$3.88	pay $776
Sell 2 · Aug 21 · 230 Calls	$2.34	collect $468
Buy 2 · Aug 28 · 230 Calls	$5.65	pay $1,130
Net debit		$806

Bid/ask midpoints, Interactive Brokers, August 12, 2026. Excludes commissions.

Only the puts are ratioed: four short August 21 puts against two long August 28 puts. Two of those four short puts aren’t offset by anything. They are cash secured short puts, and had Nvidia finished meaningfully below $215 they would have committed us to buying 200 shares at $215, or $43,000. The premium collected on that uncovered portion was $316.

That $316 isn’t a discount

The straight two-lot version of this structure would have cost $1,122 and could not have lost a dollar more than that. The ratioed version cost less, $806, precisely because those two extra short puts brought in $316 the capped version never collects. It is payment for taking on the uncapped side. Anyone running the ratio should size it against the $43,000 commitment, not against the $806 debit.

It returned 94.8%, and it was closer to trouble than that number suggests. Its breakeven sat at $212.50 and Nvidia closed at $214.72, so the cushion was about $2.22. A further 2.2% decline would have erased the entire debit, and below that the naked puts keep bleeding.

The worst case, which traders skip

Any option strategy deserves a clear view of what could have gone wrong. For the basic structure the worst case was simple and fixed: lose the $5.61 per share paid. That would have required Nvidia to make a large move during the quiet week, big enough to inflate the August 21 options rather than let them decay. A sharp headline ahead of earnings, a regulatory surprise, or a sudden selloff across the semiconductor sector could each have done it. None occurred.

For the ratioed variant the worst case isn’t fixed and doesn’t have a floor short of the stock going to zero. That is a different trade, and it belongs only in an account that genuinely wants to own more Nvidia lower.

The idea worth carrying away

The earnings announcement was the reason this trade existed. It set the price of the options, it opened the gap between the two expiries, and it supplied every dollar the position earned.

And when we closed the position on Friday afternoon, it still hadn’t happened.

In a calendar spread the event is the source of the opportunity. It isn’t something you have to sit through to collect it. Holding the August 28 legs through Wednesday would have been a different trade altogether, a bet on the size of the earnings move, purchased at 55% implied volatility with the collapse arriving the following morning. That isn’t what we underwrote, and by the time Nvidia steps to the microphone we will own none of it.

Jeff Kaminker, CFA, CFP, FRM
President, Frontwater Capital Inc.

Educational content only. This is not investment advice, an offer or solicitation to buy or sell any security, or a recommendation of any strategy for any particular investor.

This is a single trade shown for illustrative purposes, discussed here because it was written up publicly before the outcome was known. It is not representative of results generally, and past performance does not indicate future results. Options involve substantial risk, including total loss of the premium paid. Multi-leg strategies add complexity and execution risk, and ratioed structures with uncovered short options can lose considerably more than the premium paid. Investors should read the Characteristics and Risks of Standardized Options before trading.

Entry prices are bid/ask midpoints sourced from Interactive Brokers on August 12, 2026. Exit prices are closing bid/ask midpoints on August 21, 2026. Figures exclude commissions and are illustrative of the structure described rather than a statement of any client account’s returns. The short August 21 215 put finished $0.28 in the money and has been charged at that amount; final assignment status was pending at publication and figures will be restated if it differs. Frontwater Capital and accounts it manages hold positions in Nvidia. Positions may change at any time without notice.

Please speak with your portfolio manager before acting on anything discussed here.