The tape on 18 September 2026 produced a headline number that looks like a verdict:
- 2,380 prints clearing the unusual-activity floor
- 425 distinct symbols
- $2.7B call premium
- $6.3B put premium
- Largest single print: an SPCX put at the $205 strike, $561.3M
Put premium at 2.33 times call premium. Read at face value, that is a market buying protection with both hands, and it would be easy to write a confident paragraph about institutional caution.
That paragraph would be wrong, and the reason is on the calendar rather than in the data.
The date is doing most of the work
18 September 2026 was the third Friday of the month, which makes it quarterly options expiration. Monthly and quarterly contracts settle, index options settle, and positions that have been accumulating for months get closed, rolled or exercised in a single session.
The fingerprint is all over the day's prints. Scroll the tape and the expiry column repeats the same value again and again: 09/18/26, one day. Tesla puts at $362.50, $365, $367.50, $370 and $372.50, all expiring that afternoon. Advanced Micro Devices calls at ten adjacent strikes from $535 to $575, same expiry. Micron on both sides of $980. Amazon, Apple, Meta, Nvidia, all with clusters of same-day contracts.
On an ordinary session, a large put print is a decision. On quarterly expiry, a large put print is frequently the end of a decision made months ago, or a hedge being maintained, or an index position being rolled forward. The premium is real money and the print is real, but the direction of the trade tells you very little about what anybody expects next week.
The lesson generalises past this one session: aggregate put-call premium ratios are only comparable across days with similar structure. Comparing a quarterly expiry to a random Wednesday is comparing two different activities that happen to use the same instruments.
What to do with the number instead
The ratio is not useless. It is useless in isolation. Three adjustments make it informative.
Adjust for the calendar. Compare like with like. A quarterly expiry belongs next to the previous quarterly expiry, not next to yesterday. The daily archive keeps previous sessions addressable by date, which is what makes that comparison a lookup rather than a research project.
Strip the short tenors. Filter out anything expiring within a few days and recompute. What remains is positioning that survives the session, and its put-call balance is a far better sentiment reading than the headline. On a quad-witching day this typically removes most of the volume and nearly all of the noise.
Separate index from single name. A put on a broad ETF and a put on one company carry entirely different information. The first is usually portfolio insurance; the second is a view on a business. OptionsBell's treatment of single-name versus ETF flow asymmetry is the argument in full, and it matters most exactly on days like this one, when index-linked activity dominates the totals.
The eight names that carried the day
Concentration is the other thing the headline number hides. The most active tickers by premium:
| Ticker | Premium |
|---|---|
| SPCX | $1.3B |
| TSLA | $662.7M |
| FRO | $413.4M |
| AMGN | $363.3M |
| MU | $339.7M |
| AMD | $329.6M |
| LULU | $319.5M |
| BABA | $299.9M |
Eight symbols out of 425 account for roughly $4.0B of the $9.0B total. Any statement about "the market" on that session is, arithmetically, mostly a statement about these eight names.
Which is good news, because eight companies is a workload you can actually handle. Instead of theorising about aggregate sentiment, you can read what each of these businesses said on its most recent call and ask whether the positioning fits.
A worked example: Micron
Micron printed $339.7M on 18 September, with heavy activity on both sides of the $980 strike:
| Side | Strike | Expiry | Premium | IV |
|---|---|---|---|---|
| Call | $980 | 09/18/26 (1d) | $31.3M | 55% |
| Put | $980 | 09/18/26 (1d) | $21.6M | 52% |
| Put | $975 | 09/18/26 (1d) | $16.6M | 56% |
| Call | $985 | 09/18/26 (1d) | $10.6M | 55% |
| Put | $970 | 09/18/26 (1d) | $10.1M | 56% |
Balanced, at-the-money, expiring that day. On the flow alone this is unreadable: someone is long, someone is short, and on expiry day much of it is neither.
Now open the 10 August 2026 transcript and the picture acquires content that no amount of tape-reading would have produced.
The quarter itself beat. Revenue of $6.5B against a $6.2B estimate, up 15% year over year. EPS of $1.10, fifteen cents ahead. Operating margin at 8%, up from 7% the prior quarter.
The forward-looking commentary is where it gets interesting. Management described AI as a "transformative capability" and a "once in a lifetime opportunity," which is the sort of language that is easy to discount. What is much harder to discount is the operational statement underneath it:
"We still don't have line of sight as to when the supply is going to be able to meet demand."
That is a company saying it cannot yet see the end of a shortage, with the implication running into 2027. Alongside it, a commitment to raise US manufacturing investment from $200B to $250B, and $22B in cash commitments from supply agreements, $18B of it in cash.
Put those together and you have a genuine two-sided setup, which explains balanced positioning far better than any sentiment reading. Persistent undersupply is excellent for pricing and margins in the near term. A $250B capital commitment, made while demand cannot be met, is precisely how memory cycles have historically ended. Both the bull case and the bear case come from the same three sentences.
A balanced option chain over a genuinely two-sided business is not indecision. It is the market correctly pricing a situation where reasonable people reach opposite conclusions from identical facts.
The routine
This is the version worth running on any high-volume session, and it takes about twenty minutes:
- Read the concentration, not the total. Identify the handful of names carrying most of the premium.
- Filter to tenors past the current week. Discard expiry-day noise unless you are specifically studying it.
- Pull each surviving name's most recent transcript. Read the Q&A first.
- Ask what is two-sided. Where management's own commentary supports opposite conclusions, expect and accept balanced positioning.
- Flag the mismatches. A name where the transcript is unambiguous but the positioning is heavily one-directional against it is the case worth a second look.
Step three is the one people skip, and it is the one that converts a table of numbers into an argument about businesses. Our transcripts are free to start reading, and the live flow is free to watch without an account, so the only real cost of this routine is the twenty minutes.
The discipline underneath
Every aggregate market statistic has a structural explanation competing with the interpretive one, and the structural explanation is usually right.
Put premium exceeded call premium by 2.33 times on 18 September 2026. The interpretive reading is that institutions were hedging into an uncertain autumn. The structural reading is that it was quarterly expiry and a large share of that premium belonged to positions being closed rather than opened.
You cannot distinguish the two from the ratio. You distinguish them by looking at expiry dates, by comparing against structurally similar sessions, and by reading what the underlying companies actually said about their own businesses.
That last step is the one that keeps the exercise honest. The tape tells you money moved. Only the transcript tells you what it moved around.