How Much Does IV Drop After Earnings? Real Crush Numbers by Name

If you are asking how much does IV drop after earnings, the honest answer is: it depends on the name, and the difference is measurable. This post puts real numbers on earnings IV crush using event-history data, then explains the mechanics that make META's crush three times NVDA's.

IV crush, defined in one paragraph: implied volatility before earnings contains a one-time jump premium for the announcement. The moment results are out, that uncertainty is resolved and the jump premium evaporates - ATM implied volatility drops discontinuously, typically at the next open. The size of the drop is the share of total option-implied variance that the event itself represented.

The measured distributions

From GET /v1/earnings/iv-crush/{symbol} on 2026-08-03, which returns the live expected-crush estimate plus the distribution over up to 20 past events:

Symbol	Median crush	P25	P75	Best event	Events	Next earnings
META	42.5%	41.0%	43.4%	46.5%	6	2026-10-27
AAPL	23.2%	20.4%	27.5%	28.6%	6	2026-10-28
NVDA	14.8%	14.1%	15.5%	16.8%	4	2026-08-26

Crush here is the percentage drop in ATM IV from the last pre-event reading to the first post-event reading, front expiry.

Honest footnotes: the samples are the platform's covered event history (four to six events per name at the time of writing, growing each quarter), and a zero in a distribution's worst column - both AAPL and META carry one - marks an event where the measured crush did not materialise in the data. The interquartile range is the robust read, not the extremes.

Why META crushes 3x harder than NVDA

Pre-event IV is a blend of two components: baseline diffusion (the vol the stock runs on ordinary days) and the event jump. The crush percentage is essentially the event's share of total implied variance. Two things drive it:

• How large the expected jump is relative to baseline vol. META's post-earnings moves have repeatedly been double-digit percent against a baseline vol in the 30s - the event dominates the front expiry, so resolving it removes most of the IV. NVDA runs a high baseline vol (around 37% in early August 2026, three weeks before its report) with an implied move that is large in dollars but smaller relative to that baseline, so the event share - and the crush - is smaller.
• Days to expiry at the event. The shorter the expiry, the larger the event's share of remaining variance, the more violent the crush. Weeklies crush hardest; a 60-day option barely notices.

The expected-move endpoint performs exactly this decomposition - splitting the front-expiry straddle into jump and diffusion using the pre/post-event term structure - which is what makes a live crush estimate possible before the event.

The ramp: expected crush grows into the event

Three weeks before NVDA's 2026-08-26 report, the live estimate read a modest expected crush (about 4% at the front expiry, pre-IV 42.5 against post-IV 40.8) - far below the 15% the distribution says the event delivers.

That is not a contradiction; it is the ramp. Event variance concentrates into the front expiry as the calendar rolls toward the report: with three weeks of ordinary trading days still in the expiry, the jump is a small share of total variance. By the week of the event, the front expiry is mostly jump, and the expected crush converges toward the historical distribution.

Watching the live estimate ramp against the historical median is the cleanest way to see whether this quarter's event premium is building rich or cheap relative to the name's own history.

Trading implications, honestly stated

Crush is not free money. The stock moves at the same moment the IV collapses, and whether short-premium structures win depends on implied vs realised move, not on the existence of crush. Measured separately across 70 events: the median event harvested a third of the implied move with a 67% win rate and a fat left tail.

The per-name crush distribution adds the structure-selection layer. Names with large, reliable crush and modest realised moves favour short-vega structures; names where the crush is small relative to gap risk favour defined-risk or long-gamma-into-ramp structures.

Pulling it yourself

import requests

for sym in ["AAPL", "META", "NVDA"]:
    r = requests.get(
        f"https://lab.flashalpha.com/v1/earnings/iv-crush/{sym}",
        headers={"X-Api-Key": KEY},
    ).json()
    est, dist = r["current_estimate"], r["distribution"]
    print(sym, r["earnings_date"],
          f"expected {est['expected_crush_pct']}%",
          f"median {dist['median']}%  p25 {dist['p25']}  p75 {dist['p75']}")

Per-event detail - EPS and revenue surprises, implied vs actual moves, realised crush per event - comes from the companion /v1/earnings/history/{symbol} endpoint, and the upcoming calendar from /v1/earnings/calendar.

FAQ

How much does IV drop after earnings on average?
Across the names measured here, median crush ranges from about 15% (NVDA) through 23% (AAPL) to 43% (META) of pre-event ATM IV at the front expiry. There is no useful single average: the number is a per-name property driven by the event's share of total implied variance, and it is stable enough per name to be worth looking up rather than guessing.

When exactly does IV crush happen?
At the resolution of the uncertainty: effectively instantaneous at the first quotes after the announcement (the next open for after-close reporters). The decay into the event is a separate, slower effect - the jump premium itself holds until the news is out.

Can you profit from IV crush by selling options before earnings?
Only when the implied move overprices the realised move - the crush and the gap arrive together. The measured base rate: about two-thirds of events pay the seller something, the median event pays a third of the implied move, and the tail events cost multiples of it.

Does IV crush affect all expirations equally?
No. Crush concentrates in the front expiry, where the event is the dominant share of remaining variance. Back-month IV barely moves, which is why calendar structures are one of the standard earnings expressions and why measuring crush requires expiry-matched pre/post readings.

Wrapping up

IV crush is real, large, and - the part almost nobody quantifies - radically different across names: the same mega-cap quarter produced a 15% median crush in NVDA and a 43% median crush in META. The distribution for any covered name is one API call, and the live estimate ramps against it into each event.

Originally published at flashalpha.com.