Does selling options before earnings pay? Across 416 US reports, the spread took the seller's small gain away

How this study happened

Earnings season is in full swing, and many of us are tempted to play the big moves with options. In my previous study, buying a call before the report mostly lost: 5 of 6 calls lost money. Selling options looks like the obvious answer, since the seller collects what buyers pay.

Before computing anything on the seller's side, I wrote down the rules in two parts. This page is part one: the reports from the buying study, seen from the other side. Part two is a new test on reports I haven't seen yet; I'll add it here in November.

Wojciech, 16 October 2026

The findings at a glance

−0.50%

of the stock price on average: a call and a put sold before the report and bought back at the next close, at real prices, 416 US earnings reports. Margin of error −1.10% to +0.10%: this could be chance

+0.54%

the same trades at the halfway price, between bid and ask. The gap, about 1% of the stock price, is what the spread cost

47%

of the trades kept money. The premium collected averaged 2.3% of the stock price

2.9×

the premium collected, or more: what the worst 1 in 10 trades lost. The worst: 61 times its premium

Claim and verdict

The claim and what the data says

Every claim from every study, in one list →

Claim tested	Finding	Verdict
Selling a call and a put before a US earnings report and buying them back at the next close pays	−0.50% of the stock price on average at real prices, 416 reports; margin of error −1.10% to +0.10%. 47% of the trades kept money; the worst 1 in 10 lost 2.9 times the premium or more. Counted trade by trade: −0.99%, −1.47% to −0.50%	Can't tell yet

1 · The main test

At the halfway price the seller kept +0.54% of the stock price; at real prices, −0.50%, a result luck alone could produce.

For each report I worked out the move the stock's own options priced in: what a call and a put at the share price cost together, as a share of the share price. The median was 9.4%. Then I sold the call with its strike that far above the share price and the put that far below, on the nearest expiry after the report, at the bid at the last close before the report, and bought both back at the ask at the first close after it. The bid is the price buyers offer; the ask is the price sellers want.

The seller's result is the premium collected minus the cost of buying back, as a share of the stock price, because a seller's loss has no upper limit and a share of the premium would distort it. Stocks that report on the same evening move together, so I average each evening first. 27 evenings had at least three reports, 385 of the 416 reports, and each evening counts the same. The average came to −0.50% of the stock price, with a margin of error from −1.10% to +0.10%. That range includes zero, so by the rules I wrote down first, these reports can't settle whether the seller lost. With 27 evenings, this test could only have picked out an average of about 0.85% of the stock price, either way.

At the halfway price, the same trades came to +0.54% of the stock price, roughly the mirror of the buyer's result at the halfway price in the previous study. So the spread cost the seller about 1% of the stock price per trade, about twice the edge at the halfway price.

Average result of the sale as a share of the stock price: at real prices −0.50% (margin of error −1.10% to +0.10%), at the halfway price +0.54% (+0.02% to +1.06%). They are the same trades: the gap, about 1% of the stock price, is what the spread cost. — WB Trading Labs · Fig. 1A call and a put with their strikes as far above and below the share price as the move the stock's options priced in. Real prices: sold at the bid at the last close before the report, bought back at the ask at the first close after it. Halfway price: between bid and ask, +0.54% with a margin of error from +0.02% to +1.06%. Average of the 27 evenings with at least three reports (385 of the 416 reports), each evening with the same weight; thin line: margin of error (95%). US earnings reports from 24 June to 9 October 2026. Data: my own record of US option prices, my calculations.

The average hides how lopsided a seller's result is. The most a seller can keep is the whole premium collected; the loss has no limit. The premium collected averaged 2.3% of the stock price. 47% of the trades kept money, and the worst 1 in 10 lost 2.9 times the premium or more. The worst single trade collected only $0.42 a share and lost about $25.50, 61 times its premium. The median trade came to −0.11%. Counted trade by trade, which gives the busiest evenings more weight, the average was a clear loss: −0.99%, with a margin of error from −1.47% to −0.50%. So the two ways of counting differ on whether the loss could be chance, and neither shows the seller keeping money. The verdict uses the evening average, the measure I chose before counting.

Histogram of the 416 trades by result as a multiple of the premium collected: gains stop at +1×, losses run far to the left. 47% kept money. The worst 1 in 10 lost 2.9× or more. 5 trades lost more than 8 times the premium, the worst 61 times. — WB Trading Labs · Fig. 2All 416 trades of the main test at real prices: each one's profit or loss divided by the premium it collected. Bars every 0.5×; trades worse than −8× in a separate, hatched bar. US earnings reports from 24 June to 9 October 2026. Data: my own record of US option prices, my calculations.

2 · Other versions

No version kept money for the seller at real prices; the one with the loss capped lost money.

These are side results: I wrote them down to see the whole picture, and none of them settles anything on its own or is compared with the others. I tried the same sale with the loss capped, by also buying a call and a put twice as far from the share price, the call alone, and a call and a put with the strike at the share price.

In the version with the loss capped, the worst 1 in 10 trades lost 6.9 times the premium or more; on average it lost 1.36% of the stock price, with its margin of error below zero. In 65 of its trades the options bought cost as much as the ones sold, or more, so there was no premium at all; those trades are left out of the premium multiples. The cap holds only at expiry: bought back at the next close at real prices, 14 of the 376 trades lost more than it.

Options sold before US earnings reports, four versions, at real prices

What was sold	Average (margin of error)	Halfway price	Kept money	Worst 1 in 10
Call and put at the priced move (main, 416)	−0.50% −1.10% to +0.10%	+0.54%	47%	−2.9×
The same, loss capped (376)	−1.36% −1.98% to −0.74%	+0.32%	31%	−6.9×
The call alone (445)	−0.22% −0.77% to +0.33%	+0.40%	58%	−4.2×
Call and put at the share price (344)	−0.92% −2.13% to +0.29%	+0.83%	39%	−1.1×

WB Trading Labs · Table 1Result as a share of the stock price, averaged over the evenings with at least three reports, each evening with the same weight; margin of error 95%. Worst 1 in 10: that trade's result as a multiple of the premium collected. Loss capped: a call and a put also bought twice as far from the share price. In brackets, the number of reports: rows leave out reports with no listed strike close enough to the target or no bid on an option sold, so they cover different reports. In bold: the test I chose as the main one before counting. Data: my own record of US option prices, my calculations.

What this can't answer

This is the other side of the same trades, not a new test.

At the halfway price, the seller's result is roughly the mirror of the buyer's, which I've already published, so this part could largely be predicted before I counted anything. What's new is what the spread costs the seller and how large the losses get. The new test, on reports from 13 October 2026, runs under the same written rules on the first Friday from 6 November with at least 20 evenings of three or more reports, and no later than 20 November.

This is one stretch of one earnings season, 24 June to 9 October 2026, mostly volatile stocks. I count closing prices only: no early exercise by the buyer, no margin calls and no commissions. Selling options can lose many times the premium collected.

Method

How I counted

Rules written down first: on 16 October 2026, before computing any of the seller's results, I wrote down what to sell, at which prices, how to treat a missing price at the buy-back, how to average, and how I'd read each possible answer, for both parts of the study.
Data: my own record of US option prices (the bid and ask of every listed contract, saved after each US close); the stocks' closing prices; the report dates and whether each company reported before the open or after the close. I publish aggregates only.
The priced move and the strikes: as in the buying study: the midpoint prices of the call and the put with the strike nearest the share price, added together and divided by the share price; then the listed strike nearest the share price plus (call) or minus (put) that move, kept if it was within a quarter of the move of that target (and never less than 1% of the share price).
When: sold at the bid at the last close before the report, bought back at the ask at the close of the first session that contained the news. The result is the premium collected minus the cost of buying back, divided by the share price at the sale. Commissions aren't included. No split or large dividend fell inside any trade.
No price at the buy-back: a sold option with no ask at the buy-back would be bought back at what it would be worth if exercised, plus its spread at the sale, never dropped; in this part it never happened. An option with no bid at the sale drops out of that row: there was no one to sell it to.
Evenings, not trades: stocks that report on the same evening move together, so their results aren't independent; a report released the next morning before the open counts with the evening before. I average each evening first, then the evenings, each with the same weight, using only evenings with at least three reports. The median, the percentages and the worst 1 in 10 use every trade.
Reports left out: of 562 reports, 510 had full prices: 2 were recorded twice, 6 had no option prices, 1 had its after-report price more than 3 days late, 9 had no matching expiry, and 34 had no readable priced move. 416 made it into the main test: 72 had no listed strike close enough to the target, and in 22 an option to be sold had no bid.

I post new studies on X first.

Follow @WojciechBernPL

All studies

Does selling options before earnings pay? Across 416 US earnings reports, the spread took the seller's small gain away