Data Explorer

Cboe SKEW index: the implied probability of a large S&P 500 fall, scaled so 100 means a normal distribution. Higher is a fatter priced left tail.

10-year minus 2-year Treasury yield. Negative is an inverted curve.

10-year minus 3-month Treasury yield - the inversion measure with the longer record.

ICE BofA US High Yield Index option-adjusted spread, the FIRST value FRED published for each date - a later revision is not what was known then.

Total equity-option open interest across all OCC clearing members.

Equity put open interest over equity call open interest. OPEN INTEREST, not volume - it is a position measure, and the far more commonly quoted put/call VOLUME ratio is a different statistic with a different history.

1 uptrend-calm, 2 uptrend-stressed, 3 downtrend-calm, 4 downtrend-stressed; NULL when trend or volatility is unknown. A CODE, NOT A STATISTIC - never rank it. The thresholds live in app/core/regime.py, and every run rewrites the whole history, so a relabel re-derives it by the next evening.

Bitmask: 1 narrow (an uptrend carried by three sectors or fewer), 2 credit-stress (HY OAS at its 3y 90th percentile), 4 correlated (sector correlation at its 3y 90th). 0 is none; NULL when the label is unknown. A CODE, NOT A STATISTIC.

Reserve balances with Federal Reserve Banks, first-published vintage. The first leg of the liquidity axis; TGA, RRP and SOFR-EFFR follow when their daily ALFRED pulls are windowed (the full-vintage request is refused).

Overnight reverse repo take-up at the Fed, BILLIONS (reserves and TGA are millions).

Share of XLE XLF XLI XLK XLP XLU XLV XLY above their 200-session mean, in eighths. NULL unless all eight are known that day.

Mean of the 28 pairwise correlations among the eight sector funds' daily returns over the trailing 63 sessions. High means one factor is moving everything - the diversification that fails in a stress.

Sample stdev across the eight sector funds of their 21-session returns. High means sectors are going different ways, the opposite of correlated.

Secured over unsecured overnight funding, in basis points. It widens when collateral or reserves get scarce - the funding-stress signature of 2019-09.

The Treasury General Account at the Fed, weekly. A rising TGA drains reserves.

Cboe VIX close, the 30-day implied volatility of the S&P 500. Highest close on record 82.69 (2020-03-16); the bound catches a units error, not a crisis.

VIX divided by VIX3M. BELOW 1 IS CONTANGO - the normal curve, three-month vol dearer than one-month. ABOVE 1 IS BACKWARDATION, which is what stress looks like. The index-level twin of the per-symbol iv_term_slope_30_90.

Cboe 3-month VIX (formerly VXV); from 2009.

Cboe 9-day VIX. Moves more than VIX around events; from 2011.

Cboe VVIX, the implied volatility of VIX options; from 2006.

The strike carrying the most call gamma x OI across expiries, as level minus spot over spot. Level = the session's unadjusted close x (1 + value).

Split-, factor- and dividend-adjusted close as of the run date. A LEVEL, NOT A STATISTIC: its percentile against its own history only says where the price trend has carried it - ret_1d, px_vs_200dma and drawdown_252d are the ones to rank.

Mean close x volume over the trailing 21 sessions. A LEVEL, NOT A STATISTIC: price times shares, so its percentile against its own history rises with the price alone - a name that tripled reads 'high liquidity' on unchanged volume. Used across names at one date (tradability), never ranked against its past.

close / trailing 252-session high - 1. Zero at a new high.

Where net GEX changes sign, as LEVEL MINUS SPOT in 14-session ATRs. NEGATIVE means spot is above the flip, on the side the sign convention calls positive gamma. Found by repricing the book at hypothetical spots (sticky strike, r = 0); NULL when no flip lies within 25% of spot.

Where net GEX changes sign, in dollars on the session's own share basis. A LEVEL, NOT A STATISTIC: its percentile tracks the price and means nothing - gamma_flip_distance_atr is the one to rank. NULL under the same conditions as the distance.

WHY gamma_flip_level is or is not there. 1: a flip was found within 25% of spot. 0: the repriced book does not cross zero within 25% - a real reading, e.g. CRSR net long everywhere nearby. -1: the model and the vendor's gamma disagree about which side of the flip spot is on, which in practice means spot is near one. NULL: no book. A CODE, NOT A STATISTIC: never rank it or read its percentile.

30-day constant-maturity implied volatility. The canonical tenor: long enough to be immune to the 0DTE collapse that made v1's front-month column meaningless, short enough to respond to a catalyst.

90-day constant-maturity implied volatility.

SIGNED RUN LENGTH: sign is the state, magnitude is how long it has held. +3 means a third consecutive backwardated session; -12 means twelve in contango. One number rather than two because the pair is never read apart - a duration without its state is meaningless, and DESIGN.md:159 asks for 'contango/backwardation state AND consecutive days in it'. A slope of exactly zero counts as contango, the commoner state; it has not occurred in 9,250 symbol-days.

CM30 minus CM90. NEGATIVE IS CONTANGO - the normal upward-sloping curve, where longer-dated vol is dearer. POSITIVE IS BACKWARDATION: the near term is bid above the far, which is what stress looks like. Measured across the Core 40, 40.9% of symbol-days are backwardated, so this is not a rare state and a flag keyed to it alone would fire constantly.

Max pain of the nearest monthly expiry at least 7 days out, as level minus spot over spot. The 7-day floor keeps expiry-week pinning noise off the series and hands each monthly over the Friday before it expires.

Calendar days from the session to the monthly expiry max_pain_distance_pct was computed on: expiry = as_of_date + value. Between 7 (the roll floor) and 120 (the table's DTE limit). A LABEL, NOT A STATISTIC: it saws from ~35 down to 7 and resets every month; never rank it.

Dealer net gamma exposure, dollars of the underlying traded per 1% move to stay hedged. ASSUMES DEALERS ARE LONG CALLS AND SHORT PUTS - the common convention, and exactly what L exists to test. Positive means hedging damps moves. 0DTE contracts excluded: they are not held past the close.

The strike carrying the most put gamma x OI across expiries, as level minus spot over spot. Level = the session's unadjusted close x (1 + value).

close / 200-session mean - 1. Null until 200 sessions exist.

Annualised stdev of the trailing 21 session returns. COMPUTED BACKWARD, unlike UW's /volatility/realized which computes it FORWARD over ~20 sessions and is therefore not knowable on the date it is stamped with (DESIGN.md:139, the highest-priority provenance hazard in the document).

Total return over one session, dividends included.

10-delta risk reversal. Wider than the 25-delta on most days, as the tail of the smile should be.

25-delta risk reversal: put-wing minus call-wing implied volatility.

30-day implied minus the 21-session volatility subsequently realised. Positive means implied overpaid, which it did on 59% of measured rows. DATED BY WHEN IT BECAME KNOWABLE, not by the day it describes: the realised leg closes 28-33 sessions later, and a percentile over rows dated by their window start would read a month into its own future.

Share of the session's option volume in contracts expiring that session, denominator truncated at 120 DTE. ONLY SPY, QQQ AND IWM, which list a 0DTE expiry every session; NULL for every other name, whose Friday weekly would make it a once-a-week statistic under the same name.