The Way It Ended: Nvidia Went Down, Our Option Trade Went Up

NVDAThe scorecard on our August calendar spread, and why a week against us still paid

POSITIONS CLOSED AUGUST 21, 2026. Figures compare August 12 entry midpoints against August 21 closing marks. Frontwater Capital and accounts it manages hold positions in Nvidia.

When we published Using Options Around Nvidia’s August 26th Earnings Announcement on August 13, we ended it with a promise: we would come back with the scorecard rather than the thesis. We closed the entire structure last Friday, exactly as that piece said we would, and Nvidia reports on Wednesday. So here is the scorecard now, before the report. Publishing it afterward would prove nothing.

The short version, so you know where this is going. Nvidia fell in every single session of that final week and finished below the structure we built. The trade made money anyway. Both halves of that sentence need explaining, and we will take them in order.

+44.9%

Returned on the
basic structure

+94.8%

Returned on the
ratioed variant

$214.72

Where Nvidia closed
below our $215 edge

Those are good numbers. Now here is the part that makes this worth writing, and the reason we aren’t simply taking a victory lap.

The trade finished outside the profit zone we drew for you. We told you the structure wanted Nvidia to land between $215 and $230 on August 21. It closed at $214.72. Twenty-eight cents below the lower edge, on the one day of the nine that actually mattered.

And it made money anyway. Not despite finishing outside the zone, but in a sense because of where it finished. That deserves an explanation rather than a shrug, because if we let you believe the thesis was simply correct, you would draw the wrong lesson and so would we.

Why it still paid

The week behaved almost exactly as we hoped on volatility, and almost exactly as we feared on direction. The structure survived because a calendar spread doesn’t really pay you for a stock standing still. It pays you for the gap between two expiries, and that gap widened more than we modelled.

What the week actually did

Nvidia went down every single session. Not dramatically, and never on a headline anyone will remember, but relentlessly. Five sessions, five lower closes, and the worst of them was a 2.34% day.

Session	Close	Change	Percent
Friday, August 14	$225.16		starting point
Monday, August 17	$225.01	$0.15 lower	0.07%
Tuesday, August 18	$219.74	$5.27 lower	2.34%
Wednesday, August 19	$217.56	$2.18 lower	0.99%
Thursday, August 20	$216.85	$0.71 lower	0.33%
Friday, August 21	$214.72	$2.13 lower	0.98%

Closing prices, Nasdaq consolidated tape. Cumulative move over the week: $10.44 lower, or 4.64%.

Nine days, one direction

Nvidia daily closes from entry through expiry, against the zone the structure needed.

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

Look at the shape of that line. There is no crash in it, no gap, no bad headline. It is a slow walk downhill, and it stepped over the edge of our zone in the last few hours of the last day. Nvidia held the $215 handle until 3:54 in the afternoon and never got it back before the bell.

Quiet isn’t the same as still

This is the finding we didn’t expect, and it is the most useful thing in this letter.

Our whole argument was that the week before earnings is uneventful, so the options expiring inside it are overpriced. On that specific question, we were right. The August 21 options we sold carried an implied volatility of 33.8%. What Nvidia actually delivered that week, measured on its daily moves, was 19.7%. We sold volatility at 33.8 and the market realized 19.7. As a volatility seller, that is precisely the outcome you underwrite.

And we still walked out of the profit zone. Both of those things are true at once, and reconciling them is the education.

The week was quiet. It simply wasn’t still.

Volatility measures how much a stock jumps around day to day. It says nothing about whether those days point the same way. Nvidia’s week had almost no jumpiness in it and yet moved $10.44, because every one of those small moves pointed downhill. Run the arithmetic and the displacement was 1.67 times what a genuinely random week at 19.7% volatility would have produced. The path didn’t wander. It trended.

A calendar spread is short two things at once, and we only wrote about one of them. It is short variance, meaning it wants small daily moves, and we got those. It is also short drift, meaning it wants the stock to finish roughly where it started, and we didn’t get that. In our view this is the risk that deserves top billing in any future version of this trade, because it is the one that is invisible in an implied volatility number.

The open question from part one, now answered

We closed the original piece by admitting we had never measured whether the pre-earnings week genuinely realizes less volatility than it is priced at. For this one instance, it did, by a wide margin: 19.7% realized against 33.8% sold. One observation isn’t a rule, and we will keep counting. But the lesson we take from it is that being right about volatility wasn’t sufficient on its own.

Which leaves the obvious question, and it is the one we set out to answer. If Nvidia moved against us all week and finished outside the range we said the structure needed, where did the return come from? It turns out to have very little to do with the half of the trade we spent the most time explaining.

So why did it pay?

Here is every leg, at entry and at exit. This is the basic one-lot version, the teaching example from the original piece.

Leg	Entry	Exit	Result
Sold · Aug 21 · 215 Put	$1.58	$0.28	+$130
Owned · Aug 28 · 215 Put	$3.88	$6.60	+$272
Sold · Aug 21 · 230 Call	$2.34	$0.00	+$234
Owned · Aug 28 · 230 Call	$5.65	$1.81	−$384
Net on a $561 debit			+$252

Entry at Interactive Brokers midpoints, August 12. Exit at closing bid/ask midpoints, August 21. Excludes commissions.

Read that table again and you will see something the original piece never told you, because we didn’t know it yet. This wasn’t one trade. It was two, and they disagreed.

+175%

The put calendar
on a $230 debit

−45%

The call calendar
on a $331 debit

The put side carried the entire result and then some. The call side lost nearly half its money, because Nvidia walked away from $230 and the call we owned went from nearly at-the-money to plainly out of it. When you put a double calendar on, you are running two independent positions that happen to share a ticket, and a directional move will reward one and punish the other.

The premium did more than hold

The core claim was that the August 28 options would keep their value because the earnings report still sat inside them. They did better than that. Implied volatility on that expiry went from 42.2% at entry to 55.1% at exit, a rise of nearly 13 volatility points in nine days. We weren’t merely renting premium that held its price. We were holding an asset the market re-rated into the event.

A correction to our own framing

We described a “profit zone” of $215 to $230. That was imprecise, and the outcome exposed it. Those two numbers mark where the options we sold expire without being assigned. They are an assignment boundary. A calendar spread’s profit actually peaks at the strikes, not in the middle between them, because that is where the option you still own carries the most time value. Landing at $214.72 put our long put almost perfectly at the money with its volatility freshly marked up, which is close to the best place it could have finished. We were outside the zone and near the sweet spot simultaneously.

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

The August 21 215 put finished in the money. Not by much, but the size isn’t the point. Nvidia’s official close was $214.72, and the clearing house automatically exercises any option that finishes as little as one cent in the money. On that rule alone, we were getting assigned stock over the weekend.

Except that isn’t where the story ends, and this is a piece of plumbing that almost never 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. And 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 the evening. Any rational holder of a 215 put, watching the stock trade above the strike, files to abandon rather than sell their shares for less than the market will pay.

Why we booked the charge anyway

We have assumed the put cost us its full 28 cents, as though we had bought back stock at $215 that we had shorted at $214.72. In all likelihood most of those puts were simply abandoned and cost us nothing at all. We would rather publish the conservative number and revise it upward later than publish the flattering one and correct it downward. On the ratioed version the difference is $112, which would lift the return from 94.8% to 108.7%. It doesn’t change a single conclusion here.

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

Twenty-five cents is a cheap way to stop being a spectator in your own position.

The trade we deliberately didn’t take

We closed everything on Friday. Nvidia reports on Wednesday evening and we will be watching it the way everyone else does, with no position at stake. That was the plan in print before the fact, and it is worth showing you what it was worth, because the alternative was genuinely tempting.

Suppose we had kept the August 28 options and held them to their own expiry, two days after the report. Those two legs are a 215 put and a 230 call. Held to expiry with nothing sold against them, that is a strangle, and a strangle only pays if the stock finishes well outside both strikes.

If Nvidia finishes August 28 at	Basic structure	Ratioed variant
$200.00	+$911	+$2,082
$210.41	−$130	breakeven
Anywhere from $215 to $230	−$589	−$918
$234.59	−$130	breakeven
$240.00	+$411	+$1,082

Payoff at expiry is intrinsic value only. Total outlay includes the 28 cents charged on the short puts. Illustrative, not a forecast.

That middle row is the whole argument. A fifteen dollar wide band in which the position is a total loss, and Nvidia closed Friday at $214.72, sitting right on its edge. Meanwhile the options market is pricing a move of roughly $13.12 in either direction through August 28, about 6.1%. So holding on would have meant swapping a realized 94.8% for something close to a coin flip, and the dead zone covers the price the stock is nearest to today.

We want to be fair about this. If Nvidia gaps hard on Wednesday night, holding would have been the better trade and we will say so. That isn’t the same as it having been the better decision. We underwrote a rental of earnings premium, not a bet on the earnings result, and changing the trade at the finish line because the ending looked exciting is how a disciplined position quietly turns into a different one.

The bottom line

The structure returned 44.9% in its basic form and 94.8% in the ratioed version, over nine days: $252 of profit on a $561 outlay, and $764 on $806, on a stock that moved against it all week and finished outside the boundaries we published. The straight two-lot version, the fully capped one, returned $504 on a $1,122 outlay, the same 44.9%.

Three things we would carry forward, and one we would change.

1

Selling the pre-earnings week worked

We sold volatility at 33.8% and the week delivered 19.7%. The premise of the trade was sound and the evidence supports running it again.

2

Owning the event expiry worked better than we claimed

We argued that premium would hold. It rose, from 42.2% to 55.1%. The gap between the two expiries is the asset, and it widened.

3

Closing on schedule worked

Exiting Friday converted a paper position into a realized result and took the earnings report entirely off our books. We hold nothing into Wednesday.

4

What we would change: respect the drift

We wrote at length about volatility and almost nothing about direction. A quiet, persistent slide is the single path that satisfies every word of our thesis while still walking the stock out of the structure. Next time that risk goes on page one, not in the footnotes.

Where that leaves us

None of this makes calendar spreads a good idea for every investor, or this one repeatable on demand. It is a narrow, mechanical trade that lives on a specific and temporary distortion in the options market, it requires you to be at the screen on the day it ends, and it can lose its entire cost. The ratioed version can lose considerably more than that, which is why we sized it against the $43,000 of stock it committed us to rather than against its $806 price tag.

We said we would come back with the scorecard rather than the thesis. Nvidia reports on Wednesday, we own no options into it, and we will keep counting these weeks until we know whether the pattern is a rule or an anecdote.

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

This publication is produced by Frontwater Capital for educational and informational purposes only. It is not investment advice, an offer or solicitation to buy or sell any security, or a recommendation of any strategy for any particular investor.

Options involve significant risk and are not suitable for all investors. Multi-leg strategies carry additional risks including early assignment on short legs and execution risk across legs. Ratioed structures with uncovered short options can lose 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 the figures will be restated if it differs. Implied-move estimates are derived from option prices and are not forecasts. Frontwater Capital and accounts it manages hold positions in Nvidia. Positions may change at any time without notice.

Past performance is not indicative of future results. Please speak with your portfolio manager before acting on anything discussed here.