The Options Read · 2026-07-14

In plain English: this page reads last session's stock-options market — the insurance market attached to each stock — and reports what changed. It forecasts one thing only: which names are likelier than usual to move big tomorrow, never which direction.

In plain English: the night's headline facts in one paragraph — where the options market got busiest, whose insurance repriced hardest, and how big the priced-in moves are.

A hot tape: 18 turbulence admits against a 6.8-a-day run rate, above the Poisson threshold. The T3 census posted 18 versus a 20-day average of 13.0. Across 290 quotable names, median ATM IV posted 71.7%, median spread 14.2%, and front-expiry movement priced ±5.0%.

AEHR repriced 46.6 points to 296.8%, its historical maximum, with ±21.3% priced through Friday; IBM repriced 26.6 points to 73.4%. SERV, KORU and HIMX abandoned 26.5, 26.0 and 20.3 points. Stock-move fields: unavailable. NFLX priced ±8.0% around Thursday’s event; TSM ±5.2%. IBM alone posted loud crowd and high premium, while 22 names—including ALAB, ASML, JNJ and NBIS—posted high premium with a quiet crowd.

lede: codex (facts-fed, lint-passed) · passes the same banned-lexicon lint as everything else

In plain English: yesterday this page named names it said should move more than usual. Here is how that went — hits and misses both, scored against ordinary luck and against a dumb alternative (just picking yesterday's biggest movers).

Grades this page's only forecast — turbulence — at the tier level, unsigned, next to the dumb benchmark it must beat.

Every board is graded here, hits and misses, unsigned. Expected hit rate is printed because a lift without its base is marketing.

In plain English: a weather report for the whole options tape: how many names tripped our alarms today versus how many would trip them on a random day. Only when today's count runs far above that chance rate does the page call the tape hot.

Regime context and null accounting. Five numbers, no more.

• median iv vs index (Tier A)— (Δ —; 20d band 0.95..1.49)
• Tier-A census86 names (Δ -23 d/d) · panel median 79
• admits vs run rateTurbulence 18 · λ ≈ 6.8/day (47-session calibration pack); permutation nulls for the anomaly families: repriced ~0.9/day, quality_zb ~0.8/day, borrow ~0.9/day
• Poisson surpriseFLAG ON — admits exceed λ + 2·√λ; the only thing this page may ever call a hot tape
• T3 census vs 20d18 today · 20d avg 13.0

Context, not signals. The 'expected by chance' figure is the null run rate of our own thresholds, estimated by permutation and refreshed monthly.

In plain English: a census of today's options market — the typical price of insurance, the typical toll to trade it, and the day's extremes with names attached. Descriptions, not picks.

Universe medians and the day's extremes, named. Arithmetic, not signals.

• quotable universe290 names (spread bands A+B)
• median ATM IV71.7%
• median IV − RV20—
• median ATM spread14.2%
• median priced move (front expiry)±5.0%

own-history column: 2-year baselines (456-session replay panel), DEEP/MEDIUM names only, n printed per claim; '10:30 hist' marks the grid-snapshot history (regime-comparable to tonight's EOD row, not identical)

Cross-sectional facts about today's universe — medians and extremes, no ranking beyond arithmetic. Context, not signals.

In plain English: we watch how violently market makers shuffled their price quotes on each stock's options today. T3 means the top tier of that churn — names where the options market got busiest rearranging itself. Historically those names move more than usual the NEXT day (about 1.5x the normal odds of a >5% move) — but in an unknown direction.

Names in this tier have historically moved more (either direction) the next day: P(|move|>5%) ≈ 15% vs 10.5% baseline, monotone, held out-of-sample (dayT +11.5 design / +7.5 holdout, incremental to vol clustering).

book churn is |Δbook| of bid-size quote-book tilt, NOT open interest (C3); tiers from the frozen cuts 0.17/0.42, all rows T3 by admission

strike-adds column: awaiting panel v2 (listings diff exists in raw chains per the Phase-0 audit; Tue–Thu only). Struck, not approximated.

S3 shadow check (declared per the changelog rule): the fixed-strike-set variant retains its tier spread out-of-window (t +3.8 holdout / +6.6 out-of-time) but daily ranks agree at corr 0.65 — below the 0.8 bar, so the board switches to the fixed-set column (cuts re-frozen once, pre-launch: 0.146/0.357) the day panel v2 lands it. Until then admission runs on the floating validated artifact.

convergence counts run over ~3.8 effective families (S5 family-correlation pack); turbulence + strike-adds + Δbook firing together counts once.

Which way? Unknown — not 'we're not sure': no direction signal from option state has ever survived our holdouts. This is a watchlist, not a side. No positive-EV expression exists: even the best-gated straddle tier loses after taker costs.

Own-history stats are computed on at most N=53 sessions per name today. The finest honest claim at this depth is 'today is the max of its own history' (p ≈ 1/(N+1) ≈ 2%). Percentile claims sharpen as the panel ages; N is printed with every stat.

In plain English: your actual positions and candidates, checked against tonight's churn tiers. A T3 name in your book is likelier than usual to swing hard tomorrow — either way — which is a reason to revisit position size, not a reason to trade its options.

Application of the turbulence tiers and the mechanics veto to held and candidate names. A sizing conversation, not a trade list.

LOADED: 5 of 17 book names (T3 ≈ 1.5× base odds of a >5% move)

Today's live book

Seated-pool candidates (tomorrow's forming book)

P(|move|>5%) tier base rates (47-session panel): T3 17.2% · below-T3 11.7% — the base rate is a property of the tier, so it is printed once here instead of on every row.

Your positions against tonight's universe read. A T3 book name is a sizing conversation, not a signal to trade options on it.

In plain English: the price of insurance on a stock jumped or collapsed today in a way the stock's own move does not explain. When that happens, someone repriced risk on information you cannot see in the chart.

Observation-class, accruing a grade from day one. Largest residual ΔIV — a description of today, not a forecast.

This is repricing net of what the stock move and the index vol move already explain. Raw IV changes on down days are mostly beta; the residual is where information lives — and it is a description of today, not a forecast.

In plain English: the biggest one-day changes in what insurance costs on each stock, before subtracting the mechanical part. Insurance normally reprices when the stock itself moves, so read the stock-move column beside each ΔIV: where the stock barely moved, the repricing is the interesting part.

Largest raw day-over-day ATM-IV changes, with yesterday's stock move alongside. Includes spot-beta until the residual board lands.

raw ΔATM-IV — includes spot-beta until the residual board lands; read the yday-move column next to the ΔIV column and much of the 'signal' dissolves into beta. That is the point of printing both.

Raw day-over-day ATM-IV changes include spot-beta until the residual board lands; yesterday's stock move is printed beside each so the beta is visible. A description of today, not a forecast.

In plain English: what a one-week 'it moves a lot, either way' position costs — which tells you how big a move the options market itself has priced in. That is the market's estimate, not ours, and history says paying it is a losing trade.

Largest front-expiry ATM straddle-implied moves (mid straddle / spot). What the market priced, not what will happen.

mid straddle at the panel's ATM strike, front expiry, both legs live-quoted (bid > 0); the spread column is the toll a taker would pay to express any view on it

The straddle-implied move is the market's own pricing of movement by expiry — descriptive of what is priced, never an endorsement of paying it (the premium-selling backtest died in holdout).

In plain English: two dials on every stock: how loudly the crowd is posting about it, and how much its options insurance costs. The square to watch is quiet + high premium — options repriced without a crowd, so whoever repriced was not reacting to chatter.

Attention z-score × IV percentile of own history, four quadrants. A divergence map, not a signal.

cuts: attention z ≥ +1 / ≤ −1 (own 60-session history) × ATM-IV percentile ≥ 0.8 / ≤ 0.2 of 2-year baselines for the high tail on DEEP/MEDIUM names (10:30-snapshot history; fixed quantiles resolve ≥ p80 only), panel own history (≤47 sessions) for the low tail and the rest; 261 names carried both coordinates tonight

the S4 panel priors for these cells are printed under Quiet loading; the 'reading' column is a label for the axes, not an inference

Attention and premium cross-read, four quadrants, alphabetical within each. Quadrant membership predicted nothing out-of-sample; the quiet-loading hypothesis accrues separately above.

In plain English: names from tonight's turbulence board that nobody on social media is talking about. Churn without chatter is the version most likely to mean something is happening away from public view — but that idea is still on trial, and its scorecard prints below.

Frozen definition: turbulence-board admission ∧ attention z ≤ −1 ∧ MM §1.3 filters. A hypothesis in public, not a signal.

Priced-vs-Posted quadrants — 47-session panel priors

residual-ΔIV and single-strike program? columns await panel v2 (struck, not approximated).

S4 quadrant base rates below are computed panel priors — descriptions of 47 sessions, not claims.

Someone repriced this and nobody tweeted. The cell's historical forward stats are printed from our own panel; the pre-registered bar for promotion is t ≥ 2 day-clustered on the cell-minus-row spread after ≥60 sessions, both attention-extreme cells same ordering at f1 and f3 (accrual counter below). Until then: a hypothesis in public, not a signal.

Own-history stats are computed on at most N=53 sessions per name today. The finest honest claim at this depth is 'today is the max of its own history' (p ≈ 1/(N+1) ≈ 2%). Percentile claims sharpen as the panel ages; N is printed with every stat.

In plain English: these names have a scheduled event (usually earnings) inside their options' lifetime, so the insurance is priced for the event — the ± figure is the move the market has already penciled in. Priced is not the same as predictable: the actual gap lands on either side of it.

The event partition: names with a scheduled event inside the front expiry are excluded from every other board and appear here, with the move the market has already priced.

priced-move column: two-expiry event-variance extraction is audit-gated OFF (single front expiry per file); the fallback is atm_iv/√252 per day. Prior-cycle comparison ships with 52w chains.

Priced ≠ forecastable. Richness is a gap forecast the market already made — our 2-year backtest of selling it: NO MODEL (design Sharpe 2.0 → holdout −0.7). This board exists so tomorrow's 12% gap is never a surprise, on either side.

In plain English: who reports earnings over the next five sessions, with the size of move the options market has already priced for it — so tomorrow's gap, whichever way it goes, is never a surprise.

Scheduled earnings in-universe over the next five sessions, with the priced move where the front expiry covers the event.

dates are FMP-estimated (no confirmation feed — MM §1.6); the priced move is the front-expiry straddle where it covers the event, which includes non-event days (two-expiry isolation pending)

weekday calendar; exchange holidays not masked

Scheduled events in-universe over the coming sessions, with the move the market has priced where it is computable. Priced ≠ forecastable.

In plain English: these companies just announced they are selling new shares (or similar mechanics), which waters down existing holders. It is the one list on this page where history leans one way — such names tend to lag for about two weeks — so it works as a veto on buying, never as a list of names to bet against.

Offering-flagged names historically underperform matched controls over the next 10 sessions (−3.6% design / −7.3% holdout). Framed as a veto we apply, never a short-candidates list.

silenced-pump flag family: not yet wired (needs the pump-flag silence feed); dilution family live. Stated, not faked.

16 additional flagged name(s) beyond the 15-row cap (countdown continues off-page; re-qualification is by countdown, not by vibes)

The only rows on this page where history leans one way — and they lean down, both. Do not buy the dip; names re-qualify by countdown, not by vibes.

In plain English: an alarm for when professional quoting desks step back from a stock's options — quotes go wide or vanish. It doubles as the page's standing cost warning: with typical quoted round-trips near 29%, casually trading these options hands that toll to the middleman.

Spread blowouts, zero-bid withdrawal, repriced-on-no-volume.

Market-quality events: quoting desks arriving or defecting. Median ATM spread in this universe is 29% of mid; unconditional ATM buy-ask/sell-bid loses ~29%/day. This module is why every other module's temptation should route to stock, not options.

In plain English: open interest counts the option contracts that exist and are still open; this board lists the names with the most open contracts in LATER months, where slower-moving money tends to sit. The feed is brand new, so these are raw levels with no track record behind them yet.

All-expiry open-interest accumulation (new feed). No historical validation exists; the family accrues its record live in the graded ledger from its first edition.

day-over-day back-month ΔOI live (T+1 reported, from the OI history backfill); own-history percentiles remain unavailable until ≥20 sessions of feed history

LIVE-ONLY, UNVALIDATED — this family has no historical validation: the all-expiry open-interest feed is new, so there is no backtest behind it, favorable or otherwise. It earns its statistics in public through the same graded ledger as every other board, and until its pre-registered bar is met it makes no forward claim.

In plain English: ideas we are testing in public. Each tile states its rule in advance, counts its live sessions, and names the bar it must clear before it may be called a signal — so a pet theory cannot promote itself.

Pre-registered experiments with adjudication dates. Nothing here is a signal.

Each tile is its own experiment log: frozen definition, stated weakness, pre-registered promotion bar, live session counter, adjudication date. No color, no ranking, no prose inference.

In plain English: things we measured that have never predicted anything, kept on the page as a public record of what failed — including several metrics popular elsewhere.

OBSERVATIONS — none of this predicted anything out-of-sample. Logged for the record, not for trading.

IV−RV SPREAD list

RV-contamination flag marks a post-event RV leg (~20 sessions).

Skew / rr25 description

withheld: renders only post-salvage (parity-forward re-inversion, borrow extraction, HTB exclusion, two-sided distinct wings, spot-residualized ΔRR — MM §1.5); the current panel's rr25 is pre-salvage. Skew describes hedging demand; in our testing it predicted nothing.

Pin-risk flag into opex

withheld: needs strike-level size concentration (panel v2).

Implied borrow movers

withheld: parity-implied borrow ships with panel v2.

Largest iv_rank jumps

Epitaph: IV-momentum was killed three independent times (design t +4.5 → holdout t −2.2, plus two more sign-flips). A jump here is a fact about today, not a fuse.

OBSERVATIONS — none of this predicted anything out-of-sample. Logged for the record, not for trading.

Own-history stats are computed on at most N=53 sessions per name today. The finest honest claim at this depth is 'today is the max of its own history' (p ≈ 1/(N+1) ≈ 2%). Percentile claims sharpen as the panel ages; N is printed with every stat.

ATM IV

The price of options insurance on a stock, quoted as the yearly percent move that price implies; 'ATM' (at the money) means it is measured at strikes near the current stock price. A 60% ATM IV prices the stock as if a typical day were roughly a ±4% day.

IV vs RV20

IV is what the options market charges for FUTURE movement; RV20 (realized volatility) is how much the stock ACTUALLY moved over the last 20 sessions. IV far above RV20 means the insurance costs more than the stock's recent movement would justify; far below means the opposite.

book churn / quote-book tilt

Market makers keep standing buy and sell quotes on every option; that wall of quotes is the 'book', and tilt measures which side of it is heavier. Churn is how much the wall got rearranged since yesterday. It is quote traffic, NOT open interest — no contract needs to have traded.

T1 / T2 / T3

Churn tiers from two cut points frozen when the signal was validated: T1 quietest, T3 the top tier. T3 is the only tier this page forecasts with — historically about 1.5x the base odds of a >5% move the next session, direction unknown.

spread bands A / B / C

The bid-ask spread is the gap between the price you can buy at and the price you can sell at — the toll for a round trip. Band A: toll under 10% of the option's price; B: under 25%; C: worse. On a $1.00 option, a 5% spread costs about 5 cents to get in and out; a 25% spread surrenders a quarter of the position before anything happens.

straddle-implied move

A straddle is a call plus a put at the same strike, so it pays on a big move in either direction. Its cost divided by the stock price — the ± figure on this page — is the size of move the options market has priced in by that expiry. A forecast of movement, never of direction.

open interest (OI) and T+1

Open interest counts the option contracts that exist and remain open — real positions, unlike quote churn. Exchanges report it T+1, one day late, so tonight's OI describes yesterday's positions.

back-month OI

Open interest in expiries beyond the nearest one. Longer-dated contracts are where slower, larger money tends to sit, since they do not decay away within days.

percentile claims and (n=…)

'≥ p80 (n 502)' means today's value sits in the top 20% of that stock's own last 502 sessions. The sample size prints with every claim because a percentile over a handful of sessions means little — short histories only support coarse claims, and the page refuses to sharpen them.

Poisson flag / run rate λ

λ (lambda) is how many names would trip a board's thresholds on an ordinary day purely by chance, measured on our own history. The Poisson flag turns on only when today's count exceeds λ by more than two standard deviations of that chance process — the page's only definition of a hot tape.

the badges

VALIDATED-SIGNAL: survived out-of-sample testing and is graded nightly in public. OBSERVATION: a fact about today; no predictive claim. CANDIDATE: scored from day one, not yet trusted. WATCH ITEM: a pre-registered experiment with a stated promotion bar. LIVE-ONLY: brand-new data with no history behind it, earning its record in public.

Plain definitions of the page's working vocabulary. Reference, not signals — nothing here adds information to the boards above.

In plain English: the page's contract, unchanged every night: it forecasts how MUCH names might move, never which WAY, because every direction signal we tested failed out-of-sample.

This page reports what options markets DID, and forecasts only turbulence (bigger moves, either direction — validated t +11.5/+7.5 across two regimes). It never forecasts direction: we tested direction from option state ~12 independent ways across ~1,100 screened cells and killed all of them, including IV-momentum (three sign-flips) and a premium-selling model whose design Sharpe of 2.0 became −0.2..−0.9 on untouched holdout. Own-history stats run on ≤47 sessions today — the finest claim that depth supports is p ≈ 2%. Scoreboard above grades this page's only forecast.

The Day · Prospects · Tickers

In plain English: what tonight's data can and cannot support, said out loud. Every missing input is named, and nothing is estimated to paper over a hole.

• Open interest IS in tonight's feed (settled EOD pull: per-side front-expiry OI plus all-expiry back-month totals, T+1 reported, position date 2026-07-13). Still no gamma / GEX / vanna / charm: 'unsigned T+1 OI on ≤2 expiries cannot support them' (EX §2.3) holds with this OI too — publishing the refusal is worth more than the metric.
• book_imbalance is bid-size quote-book tilt, NOT open interest (the two now coexist in the feed and are labeled separately); per-contract volume is in the EOD pull but not yet on any validated board.
• Struck this edition (absent inputs, never approximated — C3): Repriced — fixed-strike ΔIV residual. Annex families still withheld: pin-risk, implied borrow, salvaged skew, two-expiry event extraction, strike-adds column.
• Calibration: Phase-0 S1–S5 pack of 2026-07-06 (frozen thereafter; refresh only via the quarterly ritual, MM R6, with a page changelog line).
• 2-year own-history baselines: LOADED — 959 DEEP / 96 MEDIUM names carry depth-tiered percentile claims (n printed per claim; SHALLOW names print none); grid-snapshot history is cited as '10:30 hist' — regime-comparable to tonight's EOD row, not identical.
• Halts/LULD are not covered by the plumbing pre-pass; ex-div and earnings flags are best-available (FMP) and their absence is stated, never faked.
• Silenced-pump veto family: not yet wired; dilution family live (edition-2 precision pass: announcement facets only, primary-ticker attribution, author corroboration, ETF exclusion, mega-cap sanity gate — see the veto board's own notes).
• splits/strike-grid shifts and halts/LULD are not covered by the current pre-pass (needs strike-level chains); the legend states it

plumbing-calendar sources:

• opex-week: deterministic (3rd-Friday rule)
• expiry-roll: deterministic (observed chain expiry change d/d)
• ex-div: FMP dividend calendar (ex-date ±1)