nyasekai

Graveyard / Options and volatility

Selling straddles into earnings (IV crush)

Can you profit by selling options before earnings, when implied volatility collapses?

Family
Options and volatility
Verdict
Dead
Cause
No edge
Buried
22 November 2026

The idea

Option prices rise into an earnings report and fall right after (implied volatility crush). Sell a straddle (a call and a put at the same strike) before the report, buy it back after. In theory you collect the premium that evaporates.

How we tested it

DataFree historical option daily bars for about 155 large caps. Short the at-the-money straddle at the close before the print, close at the close after. Test 1: all events. Test 2: only when the implied move divided by the historical move is at least 1.0. A long-all control was also run. A second test bought straddles only when implied divided by historical move was 0.8 or less.
PeriodFebruary 2024 to September 2026 (1,252 usable events).
CostsBid-ask spreads modelled at 2.5% to 7%. Results are also shown at half the spread.
HoldoutDevelopment and holdout halves.
ControlA random non-earnings twin.

Result

Events1,252
Net per trade, all events−94 bp of the stock price (t −5.6)
Development / holdout−85 bp / −103 bp
Filtered by implied-to-historical move−96 bp (holdout −104 bp)
Net at half the spread−60 to −62 bp
Average implied vs realised move5.66% vs 5.46%
Worst trade−32% of the stock price
Long cheap straddles (96 events)+37 bp, holdout +47 bp, p 0.42

Why it failed

Options were priced almost exactly right: the implied move averaged 5.66% and the realised move 5.46%, a gap of about 20 bp. Spreads are several times larger than that, and one bad trade lost 32% of the stock price. The reverse idea, buying cheap straddles, looked slightly positive on 96 events but dies at twice the spread and is below the 150-event minimum.

What would make it worth retesting

Real bid-ask history with an average half-spread under 1% and an implied-minus-realised gap of at least +60 bp on a sample that includes a bear market. For cheap straddles: at least 150 events with real spreads.

Past results of a backtest say nothing certain about the future.