Why Didn't Options Get More Expensive After Two Down Sessions? All Nine Tickers Sit in the Bottom Two IV Deciles, and the Hardest-Hit Name Has the Least Protection

Data basis (read this first)

• Every core conclusion rests on the settled 2026-09-23 session. The U.S. options snapshot was collected in-house on 2026-09-24 and maps to the 09-23 trading day (date_offset_applied=1), covering 289 tickers; nine are used here. Hong Kong figures are the 09-23 close.
• At publication the 2026-09-24 U.S. session was still open. Figures labelled "09-24 intraday" were taken at 22:30 Beijing time / 10:30 U.S. Eastern on 2026-09-24. They are not settled closing prices and are used only in the "first test" section — they do not support any core conclusion.
• IV percentiles come from a self-collected sample, n=99–128 trading days, series start 2026-03-20. This is not a 252-day window and must not be read as a "one-year low."
• The Korean market had no trading session on 2026-09-24 (000660.KS and 005930.KS both end at 09-23, while Hong Kong data from the identical retrieval path does contain 09-24). We have not verified the reason for the closure and offer no attribution.
• This is public-layer market-structure analysis. It contains no positions, sizing, or trade actions.

1. The question: two sessions of declines, so why did options not get more expensive?

In the 2026-09-23 session, all nine core U.S. tickers fell, and Hong Kong fell alongside them. The intuitive reading is: a broad decline means the market is hedging, protection demand rises, and options get more expensive.

The options data rejects that explanation. The implied-volatility percentile for all nine tickers sat in the bottom two deciles of our self-collected sample. Not one became more expensive because of the decline.

This article answers the follow-up question: if it was not hedging, what was it?

2. Two markets, one shape: declines ordered strictly by risk exposure

U.S. equities, 2026-09-23 close

Hong Kong equities, 2026-09-23 close

The key observation: two markets with entirely different constituents produced the same ordering — broad indices fell least, high-volatility assets fell most, and the middle was strictly monotonic.

The U.S. runs from SPY at −0.72% down to SanDisk at −3.73%. Hong Kong runs from Jiangxi Copper at −0.41% and the Tracker Fund at −1.17% down to Tencent at −2.35% and Alibaba at −4.36%.

Two direct implications:

• This was not a sector event. Storage was simply the steepest end of the gradient, not a standalone story. Hong Kong internet names (Alibaba −4.36%) fell further than U.S. storage (SanDisk −3.73%). The same ordering explains both; storage does not need its own narrative.
• The driving variable was risk exposure itself, not fundamental news. Had this been industry news, declines would cluster by industry rather than line up monotonically by volatility.

3. If this were hedging, options should have become more expensive — they did not

This is the decisive evidence separating de-risking from hedging.

The highest IV percentile across the nine is 19.7 and the lowest is 0.8 — every one sits in the bottom two deciles of the self-collected sample. NVIDIA's implied volatility was 10.70 points below its own trailing 20-day realized volatility (VRP −10.70).

Genuine hedging bids up the cost of protection. That did not happen here. The conclusion: this was de-risking (cutting exposure), not hedging (buying protection). The two look identical on a price chart and completely different on an options screen.

4. The most counterintuitive cell: the names being sold are not the names being protected

Comparing changes against put/call ratios in the same cross-section (one session, so contract-roll effects do not distort it) reveals a structure that runs against intuition:

• TSMC 1.344, Nasdaq 100 1.276, AMD 1.223, S&P 500 1.104 — put/call ratios above 1, meaning more puts bought than calls. These four are precisely the tier that fell least.
• Micron at 0.369 is the lowest in the table, yet it fell 2.22%, the third-largest decline.
• SanDisk fell the most (−3.73%) with a put/call ratio of 0.628, also below 1.

Quantifying the structure with rank correlation: the Spearman coefficient between change and put/call ratio is ρ = +0.65 (n=9; Pearson r = +0.59). The positive sign means exactly this: the more a name fell, the less protection was bought on it.

⚠️ Methodological note: n=9 is a descriptive cross-sectional statistic, not a significance test. It must not be extrapolated to larger samples or other time windows. It describes the arrangement of nine tickers in the 09-23 session only.

What this implies: the buffer was bought at the layer that barely fell, while exposure was left on at the layer that fell hardest. That is the signature of a de-risking path — sell the most liquid first, hedge at the index layer — rather than a hedging path, which would concentrate protection where losses are steepest.

5. Micron reports in four trading days, and options have not begun to price it

• Micron FQ4 FY26 earnings: after the close on 2026-09-30 (16:30 U.S. Eastern / 04:30 Beijing time on 10-01). Primary source: Micron Investor Relations press release dated 2026-08-26.
• As of the 09-23 close, that is four trading days away.
• Micron's IV percentile of 8.7 and put/call ratio of 0.369 are both the lowest in the table.

A stock reporting earnings in four trading days carries an implied-volatility percentile in the bottom decile of our sample and the lowest put/call ratio of the nine. Stated purely as a pricing fact: the premium currently available for selling time value sits at the thinnest level in our self-collected sample.

Other publicly verifiable facts

• Wells Fargo cut its Micron price target to $1,400 on 2026-09-23 — still about 30.6% above the 09-23 close of $1,071.88.
• Sell-side consensus (collected 2026-09-24): Micron price target median $1,500, mean $1,542.50, with 58 buy / 10 hold / 2 sell.

  ⚠️ Note: some data sources report upside_pct 43.9 computed from the mean while labelling it "median." Because that is internally inconsistent, this article cites only the raw target prices, not that percentage.

• SK Hynix replied to a Korea Exchange rumour inquiry on 2026-09-18 with the conclusion "미확정 / undetermined" (filing number 20260918800583). Under exchange convention, "undetermined" is not a denial — a denial must state "사실무근 / factually groundless." The reply carries a hard re-disclosure deadline of 2026-12-17, and can be resolved on any day before then.

6. Methodological warning: the "net change" field in Hong Kong custody data cannot be cited directly

This is a public-data trap that readily produces wrong conclusions, and it deserves its own section.

In Hong Kong's Central Clearing and Settlement System (CCASS), the participant-level "net change" in holdings is systematically overstated because newly-entering participants are recorded entirely as positive. Evidence from the full 09-23 table:

Of the four tickers showing a positive headline net change, two flip negative once new entrants are removed — including a mega-cap such as Tencent. The field is unusable directly across the entire table, not only for leveraged ETFs. The correct treatment is to read participant direction and breadth together while excluding newly-entered participants separately.

7. First test: 09-24 intraday (not settled, not a conclusion)

As of 22:30 Beijing time / 10:30 U.S. Eastern on 2026-09-24, with the 09-24 U.S. session still trading:

The second session broadly continues the ordering: broad indices fell least (SPY −0.25%, QQQ −0.50%), semiconductors sat in the middle (SMH −1.20%), and storage was deepest (MU −1.61%, WDC −3.01%).

Two departures must be named rather than smoothed over:

1. AMD turned positive at +0.39%, breaking the monotonic ordering outright.
2. Broadcom at −1.93% fell further than Micron at −1.61%, reversing their relative positions from 09-23.

The accurate statement is therefore: the 09-24 intraday session continued the broad-index → semiconductor → storage gradient in direction, but strict monotonicity at the single-stock level has already broken. This session had not settled; the figures above are not final and do not enter the core criteria.

8. What would invalidate this reading

Falsification conditions, fixed in advance rather than moved afterwards:

• If Micron's IV percentile rises above 30%, or its put/call ratio breaks above 1.0 → the "options have not priced the earnings" reading is void immediately; the market has begun buying protection.
• If the nine tickers' IV percentiles collectively leave the bottom two deciles while declines continue → the character shifts from de-risking to hedging, and the conclusion in section 3 is void.
• If the rank correlation between change and put/call ratio turns negative → the "inverted protection" structure in section 4 has disappeared.
• If subsequent sessions stop ordering declines by risk exposure and begin clustering by industry → the cross-market de-risking explanation in section 2 is void and should be replaced by a sector-event explanation.
• If AMD departs from the gradient for two consecutive sessions → the breadth claim of "broad-based exposure reduction" must be narrowed and restated.

Disclaimer

This article is a structural analysis of public market data. It contains no positions, sizing, trade actions, or investment advice. All data sources and timestamps are stated in the "Data basis" note above. IV percentiles are derived from a self-collected sample (series start 2026-03-20, n=99–128 trading days) and are not comparable to the conventional 252-day convention. Markets carry risk; any action taken on this basis is at the reader's own risk.