Part X · Edges, one by one · Chapter 10H

Regime filters

Signals that size trades rather than make them

Where we are

Chapters 10B to 10G examined edges you can hold: premia, factor tilts and overlays. This chapter takes the fourth category of Chapter 10A, regime filters: measurements of the market that decide how much of those edges to hold. A filter earns nothing directly, since its value is the losses it keeps out of other positions, and the classic error is to trade it.

What you need first

The VIX, the VIX3M and the term structure (Chapters 5E and 9A).
Realized against implied volatility, and why their gap decides a short option's P&L (Chapters 7A and 10C).
Gamma exposure and its sign conventions (Chapters 8A and 8B), and the two regimes of dealer hedging (Chapter 8D).
The market dashboard, which reads the same markers each morning (Chapter 11A).
A filter's hit rate as a positive predictive value (Chapter 10A).

Filters are not trades but variables that size other strategies up or down, and their "edge" takes the form of avoided losses on existing positions rather than direct returns. The classic retail error is to treat a filter as a strategy: buying VXX because the VIX is in contango, or shorting SPX because dealer gamma is negative. Filters modulate rather than generate. This chapter examines five of them and then stacks them into one reading.

In plain terms

A filter is a gauge on the dashboard, not a second engine. When it reads stress, you hold less of what you already own and open nothing new, and when it reads calm, you return to full size. Used that way, a gauge costs almost nothing, while trying to trade the gauge itself is usually expensive.

1 · What a filter is

A regime filter is a measurable market variable with three properties:

It has predictive content about the near-term distribution of returns: their mean, volatility or skew.
It does not itself constitute a tradable edge after costs.
It modulates the sizing of other strategies.

The distinction matters because the same information can be used in very different ways. Consider the ratio of the VIX to the VIX3M. When it is above 1.05, near-term volatility is priced meaningfully above longer-dated volatility, which indicates stress. Three uses of that information have very different results:

Trading the VIX directly through VXX, an exchange-traded product (ETP) that holds VIX futures, has lost 99.84% to roll decay since its inception in 2009.
Trading strategies that target the VIX, such as long-volatility funds, costs 5% to 15% a year in decay.
Using the ratio as a filter, to halt new put-writing, costs nothing and prevents real losses.

Same information, different operational use, very different P&L. The general rule follows from it. Suppose you can name a measurement, such as the term structure, dealer gamma, breadth or a credit spread, and you want to "trade" it. The answer is almost always not to trade it but to use it as a sizing input on something else.

A filter works through a sizing multiplier. Write $Q$ for the position you would hold in a normal regime and $x_t$ for the filter's reading at time $t$ . The filter sets a multiplier $m(x_t)$ between 0 and 1, so you hold

Q_t = m(x_t)\,Q

(10H.1)

A filter is worth having when the losses it avoids, where $m < 1$ in bad regimes, exceed the premium it forgoes, where $m < 1$ in regimes that turn out fine. Chapter 10A works that trade-off as a screening test, with the filter's hit rate as its positive predictive value.

2 · The VIX term structure

In plain terms

The VIX measures 30-day implied volatility on SPX, and the VIX3M measures it over 90 days, so their ratio captures the shape of the curve. In normal regimes, longer-dated volatility is higher, a contango with the ratio below 1, reflecting demand for longer-dated hedges. In stress, near-term volatility spikes and the curve inverts into backwardation, with the ratio above 1, reflecting acute risk aversion.

Worked example: A calm reading and a stressed one

Step 1: calm. The VIX is 14 and the VIX3M 17, so the ratio is $14/17 = 0.82$ , and the term structure is in contango, its "normal" shape.

Step 2: stress. The VIX is 32 and the VIX3M 26, so the ratio is $32/26 = 1.23$ , and the curve is in backwardation, pricing acute near-term risk.

When backwardation develops, near-term volatility is raised more than far-term volatility, which indicates concentrated short-term concern. Historically, that regime has come with sharp short-term moves in both directions, wider option spreads, higher correlation across assets and more margin expansion under portfolio margin (PM). The filter is the ratio itself:

F_t = \frac{\text{VIX}_t}{\text{VIX3M}_t}

(10H.2)

A common rule reads it in three bands: below 1.00 is a normal regime, 1.00 to 1.05 is elevated and transitional, and above 1.05, sustained, is a stress regime.

The evidence

Whaley (2009) on the VIX as a "fear gauge" established the framework of contango and backwardation.
ProShares "Term Structure" research notes backtest the term structure as a filter on equity exposure.
Bloomberg and Cboe practitioner literature uses it widely, but it is not academically formalized.

Why it works

The information comes from the demand-driven nature of VIX prices. When demand for near-term hedges spikes, as institutional accounts buy short-dated puts, the curve inverts. That spike in demand is itself information about how institutions perceive risk. Chapter 9A traces the mechanism in the VIX's own construction.

The canonical filter

The Edge Book's rule responds to backwardation in phases:

Trigger	Action
VIX/VIX3M above 1.00 for 3 closes	Halt new short-volatility entries
VIX/VIX3M above 1.05 for 3 closes	Halt all put sales; cut the existing premium-selling sleeve by 50%
VIX/VIX3M above 1.15 for 3 closes	Close non-essential short volatility, keeping only deep out-of-the-money (OTM) tails, whose assignment risk is small
VIX/VIX3M below 1.00 for 5 closes after stress	Re-enter in phases: 25% of target, then 50%, then 100%, over 3 weeks

Avoid binary on-off triggers, because they create whipsaw: being pushed in and out repeatedly by a signal that flickers across its threshold. Phase changes over several days instead. The persistence rule of 3 closes serves the same purpose, since a one-day spike does not trigger it.

Worked example: Why three closes

Step 1: one close. Suppose, hypothetically, that in calm markets the ratio pokes above 1.00 on 1 day in 20, at random. A rule that acts on one close then trades on a false alarm about once a month.

Step 2: three closes in a row. If those days were independent, three in a row would come about $0.05^3 = 0.000125$ of the time, roughly once in 30 years. Real false alarms cluster, so the true rate is higher, but the persistence rule still removes most one-day noise. The price is lateness: the rule acts 2 days after a genuine inversion begins.

The figure runs the rule through a made-up stretch of 130 trading days. Drag the closes required from 3 down to 1, and the calm blips halt or cut the sleeve. Drag them up, and the rule acts later in the inversion.

closes required 3 closes

halt line 1.00

false alarms: 0

calm days halted or cut: 0

lag: 2 days to halt, 2 to halve

Waiting for closes trades whipsaw for lateness. A made-up path of the ratio of the VIX to the VIX3M runs through a calm stretch with a one-day spike and a two-day cluster, then a sustained inversion. Below it, a put-writing sleeve follows the chapter's phased rule, and each step waits for the chosen number of closes above its line: the halt line, 1.05 or 1.15. With one close required, the two calm blips halt or cut the sleeve for 21 days. With three, neither fires, and the rule acts 2 days after the inversion begins.

The market dashboard of Chapter 11A reads the same ratio each morning, with 1.0 as its stress line. Its same-day read and this filter's persistence rule answer different questions: whether to open a trade today, and whether a running strategy should stand down. Chapter 13E works the filter through an open put-writing position, where the rule's 50% cut means closing positions, not only stopping new ones.

When it fails

Sometimes backwardation is brief, a single-day liquidity event that reverses without any action being needed. At other times it persists longer than expected, as in 2008 and in March 2020, when it lasted for weeks. The rule has a real opportunity cost during false stress signals.

In practice

Do	Do not
Use the ratio to halt new short-volatility entries	Buy VXX or VIX futures directly: roll decay and the basis between futures and the index work against you
Use it to modulate the size of existing positions	Treat backwardation as a directional signal on SPX: it correlates with stress, but the direction after stress is unpredictable
Use it to phase your return after stress	Use a single threshold: phase the response

Common mistake: "VIX in backwardation means stocks are about to fall"

Backwardation indicates elevated near-term volatility, not a negative direction. Many episodes of backwardation resolve through mean reversion upward, in sharp rallies as fear subsides. The signal is about magnitude, not direction: use it to size, not to bet on direction.

3 · Realized against implied volatility

In plain terms

Selling implied volatility assumes that future realized volatility comes in lower. If realized volatility is running at 30% while implied is 22%, selling premium is structurally negative in expected value (EV) until realized volatility falls back below implied.

Worked example: Selling 22 while the market moves at 32

Step 1: the readings. SPX's 21-day realized volatility is 32%, and the VIX, its implied volatility, is 22%.

Step 2: the daily budget. By the rule of 16 (Chapter 5E), a VIX of 22 prices a typical day of about $22/16 = 1.4\%$ , while 32% volatility delivers days of about $32/16 = 2.0\%$ . Selling puts collects premium for 22% volatility while the index moves at 32%. The position is underpriced relative to the actual movement: this is the "negative VRP" regime.

The Edge Book measures the gap in volatility points, with realized volatility over the past 21 trading days:

\text{VRP}_t = \sigma_{I,t} - \sigma_{R,\,t-21,\,t}

(10H.3)

Here $\sigma_{I,t}$ is today's implied volatility and $\sigma_{R,\,t-21,\,t}$ the realized volatility (RV) over the last 21 days. In the example, $\text{VRP}_t = 22 - 32 = -10$ points, which signals an unfavorable regime for selling premium. The gap is a backward-looking proxy: the variance risk premium of Chapter 10C compares implied variance with the variance expected to follow, which no one can observe in advance.

Decision rule

Realized above implied over 21 days (a negative gap): halt new short-volatility entries.
Realized above implied plus 5 volatility points: close existing short volatility if the position's size requires it. In the example, $32 > 22 + 5$ , so this step applies.
Realized back below implied: permit re-entry over 5 sessions.

When it fails

Realized volatility can stay elevated for weeks: in March 2020 it stayed above 40% for more than 6 weeks. The timing of mean reversion is unpredictable. And some episodes of negative VRP resolve through implied volatility jumping, which makes your short puts worse, rather than through realized volatility falling.

Combine this filter with the term structure as a layered filter. Contango and realized below implied is a green light for selling premium; either condition violated is a pause. Chapter 11A's dashboard reads the same gap, the implied minus realized spread, as the variance premium available to a seller.

4 · The liquidity regime

In plain terms

Funding stress propagates non-linearly: when banks become reluctant to lend to one another, leveraged players cut their positions and risk assets sell off together. The MOVE index measures the volatility of the bond market, and its spikes mark stress in fixed income, which often precedes stress in equities.

The TED spread measures that reluctance. It was the 3-month LIBOR rate, at which banks lent to one another, minus the 3-month Treasury bill (T-bill) rate. The Edge Book now reads it as SOFR minus the T-bill rate. The third gauge is the high-yield option-adjusted spread (HY OAS), the extra yield that high-yield bonds pay over Treasuries.

Worked example: Three readings of the TED spread

Step 1: normal, 20 to 30 bps. Banks lend to each other freely, and funding costs barely register in other markets.

Step 2: elevated, 100 bps. Funding strain is visible, and leveraged funds are reducing their positions.

Step 3: crisis, 250 to 400 bps, as in 2008 and March 2020. Funding stress is acute, and everything correlates.

The filter is a composite, triggered by any one of three measures of stress:

\begin{aligned} L_t = \mathbb{1}\{&\text{TED} > \tau_1 \text{ or} \\ &\text{MOVE} > \tau_2 \text{ or} \\ &\text{HY OAS} > \tau_3\} \end{aligned}

(10H.4)

The indicator $\mathbb{1}\{\cdot\}$ equals 1 when the condition inside it is true and 0 otherwise, so $L_t = 1$ as soon as any one gauge crosses its threshold $\tau$ . The Edge Book's rough thresholds are a TED spread above 200 bps, a MOVE index above 150 and an HY OAS above 600 bps. They are not fixed, so recalibrate them on rolling history. With a MOVE of 160 and the other two gauges calm, $L_t = 1$ .

Decision rule

When liquidity is stressed, cut all leveraged and short-volatility exposure by 50%, cut equity beta by 25%, and halt new tactical positions until the gauges normalize.

This filter matters most for selling premium, because in liquidity events all premium-selling sleeves correlate at about 1.0 (Chapter 13G). The diversification you thought you had between put credit spreads (PCS), covered calls (CC), 0DTE trades and condors evaporates, since all of them are short volatility in different wrappers. Liquidity filters are the strongest defense against this.

The dashboard of Chapter 11A reads the same instruments with different lines: HY OAS above 450 bps as elevated and above 700 as stress, and a MOVE above 120, sustained, as stress in rates. The Edge Book's full-size conditions, later in this chapter, are stricter still for calm: a MOVE below 100 and an HY OAS below 500 bps. Thresholds on these gauges are conventions to calibrate, not constants.

When it fails

Liquidity indicators can be late: the damage to equities is often done before credit and funding markets reflect it. Use this as one layer of the filter, not as the only one.

5 · The sign of dealer gamma, at low confidence

In plain terms

Dealers who have sold options hedge to stay neutral to direction. When they are net long gamma, their hedging is stabilizing: they sell into rallies and buy into dips. When they are net short gamma, it is destabilizing: they buy into rallies and sell into dips, amplifying moves.

Worked example: Ten billion dollars of gamma

Step 1: long gamma. Suppose dealers are net long gamma worth $10B per 1% move, in the dollar-gamma units of Chapter 8A. A 1% rally raises their delta by about $10B, so they sell about $10B of index exposure to stay hedged. That selling damps the rally: implied volatility falls and range trades develop.

Step 2: short gamma. If dealers are net short the same amount, the same 1% rally lowers their delta by $10B, and they must buy. The buying amplifies the move: implied volatility rises and trends extend.

Gamma exposure (GEX) adds up this hedging across every strike:

\text{GEX} = \sum_i \Gamma_i \cdot \text{OI}_i \cdot M_i \cdot d_i

(10H.5)

Here $\Gamma_i$ is the gamma of the options at strike $i$ , $\text{OI}_i$ their open interest, $M_i$ the contract multiplier and $d_i$ the direction, +1 or −1, for whether dealers are long or short them. The direction is the part that depends on an assumption about who holds which strikes. Public GEX models, such as those of SqueezeMetrics and SpotGamma, make assumptions that may be unreliable. Written this way, GEX counts the shares dealers trade for each point the index moves. Chapter 8A builds the full formula, in dollars per 1% move, and shows how the sign flips with the convention.

The evidence

Garleanu, Pedersen and Poteshman (2009), "Demand-Based Option Pricing", in the Review of Financial Studies (RFS), established the theoretical link between dealer positioning and option prices.
A practitioner narrative about dealer hedging effects grew from 2017 to 2022.
A Cboe study of 2025, Mandy Xu's "0DTEs Decoded" (May 2, 2025), estimates that 0DTE market-maker gamma hedging is approximately 0.2% of SPX daily liquidity, much smaller than retail GEX narratives suggest.

Why the Edge Book downgrades it

Three reasons move GEX from an edge to a low-confidence filter:

Public GEX models are model-dependent. Vendors give different numbers because they make different assumptions about who is long and short which strikes.
The explosion of 0DTE options after 2022 changed the structure. GEX models built before 2022 assumed that monthly and weekly options dominated, which is no longer true.
The Cboe finding that 0DTE market-maker hedging is about 0.2% of liquidity directly contradicts the narrative that dealer flows dominate the day. The finding covers same-day hedging only, so applying it to GEX and its walls, built mostly from longer-dated open interest, extends the measurement.

Decision rule

Use the sign of GEX, positive or negative, as low-confidence regime context only.
With GEX negative, expect higher realized volatility and wider intraday ranges, and consider trimming put-writing modestly.
With GEX positive, expect compressed realized volatility and more pinning; sizing put-writing normally is fine.
Do not trade specific levels, such as the "call wall" and the "put wall".
Do not size on the magnitude of GEX, which is too model-dependent.

GEX is a soft input: ignore it if it disagrees with the term structure or the realized-volatility filter, and take slight confidence from it if it agrees. It is worth 10% to 15% of one filter's weight, not a filter of its own.

Where this book's two traditions disagree

The chapters drawn from the Dealer Mechanics Lab give the sign far more weight. Chapter 8J makes it, with the term structure, the decision boundary for choosing a strategy: premium selling and mean reversion with positive GEX and contango, directional and long-volatility trades with negative GEX and backwardation. Chapter 9A's regime table and the classifier of Chapter 9C rest on the same pairing. The dashboard of Chapter 11A also reads the distance to the gamma flip and the nearest large strike, and the conviction score of Chapter 13B asks whether dealer positioning supports a trade's path.

The Edge Book's reasons for giving it less are the three above. Both traditions read the sign as regime context rather than a trade, and Chapter 8B warns that the sign itself rests on a convention. Where they differ is weight: a central input for the Lab, a nudge worth a tenth of a filter for the Edge Book. Both reject scalping level by level (Chapters 10I and 13F).

Common mistake: "I can sell at the call wall and buy at the put wall"

Rules for trading specific levels are not retail edges. The levels shift during the day as positioning changes, and the Cboe finding that 0DTE market-maker hedging is about 0.2% of liquidity directly undermines the size of these effects. Use the sign of GEX as low-confidence context, not as a set of level-trading rules, and see Chapter 10I for the full case.

6 · Breadth and internals

In plain terms

An index's price can hide deterioration beneath it. When fewer and fewer stocks take part in a rally, the rally is fragile, and when breadth improves while the index is flat, the regime is bullish.

Worked example: Two divergences

Step 1: a narrow high. SPX is at an all-time high, but only 30% of its stocks, about 150 of 500, are above their 50-day moving average (MA). The high is being driven by a few mega-caps, which is fragile: when the few stocks holding it up rotate, the index falls quickly.

Step 2: a broad base. SPX has been flat for 3 months, but 65% of its stocks, about 325 of 500, are above their 50-day average. Beneath the surface, participation is broad and improving, which is bullish although the index does not show it yet.

The usual measures are the shares of stocks above their 200-day and 50-day moving averages, new highs minus new lows, and the advance-decline line. That line is the running total, day after day, of the number of stocks that rose minus the number that fell.

Decision rule

Breadth divergences are warning signs, not direct filters. Use them as context that confirms or disconfirms other signals:

A new high in SPX with deteriorating breadth: cut equity beta by 10% to 15%.
A flat SPX with improving breadth: maintain equity beta or raise it slightly.

When it fails

Breadth can stay weak for months, as in the era of mega-cap concentration from 2018 to 2024. The signal often runs 6 to 12 months ahead of price, so acting on it early costs opportunity.

Chapter 11A's dashboard uses the same measure with its own lines: under 40% of members above their 50-day average during a rally is narrow, and over 55% is part of its reading of a healthy uptrend.

7 · Stacking the filters

The real value comes from stacking filters, because each one alone is noisy and conviction is highest where they agree. The Edge Book's conditions for full short-volatility exposure are these:

Gauge	Green when
VIX/VIX3M	Below 1.00
21-day realized volatility	Below implied
TED spread	Below 100 bps
MOVE index	Below 100
HY OAS	Below 500 bps
GEX, a low-confidence supplement	Positive

All green means full size, one yellow means cutting size by 25%, and two or more mean cutting it by 50% or more. This layered filtering is the operational discipline that turns individual signals, each with a mediocre signal-to-noise ratio, into a portfolio-level regime indicator with much higher confidence.

Worked example: One morning's stack

Step 1: the readings. Suppose, hypothetically, the VIX is 18 and the VIX3M 19.5, a ratio of 0.92, with realized volatility at 14% against implied 18%. The TED spread is 40 bps and HY OAS 430 bps, but the MOVE index is 112, and GEX is negative.

Step 2: count the lights. The term structure, the volatility gap, the TED spread and credit are green, while the MOVE index is yellow, above 100. GEX disagrees with the term structure and the realized-volatility filter, so by its own rule it is ignored.

Step 3: the size. One yellow means a 25% cut: a put-writing sleeve with a target of $100,000 runs at $75,000.

Familiar ground

Requiring every light to be green before full size is testing in parallel, in which any positive result stops you, and parallel testing raises sensitivity while lowering specificity. With two independent tests of 70% sensitivity and 90% specificity each, the pair misses stress only when both miss, so its sensitivity is $1 - 0.3^2 = 91\%$ . It passes a calm market only when both pass it, so its specificity is $0.9^2 = 81\%$ . Missing a crash is far costlier than a paused sale, which is the case for accepting the extra false alarms.

The cumulative impact of well-implemented filters is large. The biggest single contribution to a retail short-volatility account's survival is not entering during high-stress regimes. A disciplined stack of filters can prevent a drawdown like the PUT index's 33% in the global financial crisis (GFC) of 2008, by halting new entries before the worst losses. The cost is opportunity cost during false stress signals, the premium you do not capture. Empirically, the Edge Book finds that this opportunity cost is much smaller than the drawdown avoided.

8 · What is not a filter

The negative space clarifies what a filter is. These are not filters:

Not a filter	Why
Specific GEX levels, such as a "call wall at 5500"	Too model-dependent
Single oscillator readings, such as a relative strength index (RSI) at 30	Too noisy for sizing decisions
Headline news, such as a CPI print or a geopolitical event	Unpredictable and reactive
Concentrations of options open interest	Descriptive, not predictive
Sentiment indicators, such as the AAII survey or the Fear & Greed index	Noisy, and useful only as contrarian signals

A filter needs continuous measurement, demonstrated value in predicting the regime, and a basis that is not a narrative. If a "filter" relies on a story to justify its use, it is probably not a filter.

In practice

Write your filters down as rules before you need them: the gauge, the threshold, the number of closes, the size it sets and the condition for coming back. Read them at the same time each day, as the dashboard of Chapter 11A suggests. When a rule fires, follow it even while your intuition says the position can recover, because the rule exists for exactly that moment (Chapter 13E). Keep a log of each trigger and what followed, so that you can recalibrate a threshold that produces persistent false alarms (Chapter 10A).

When it fails

Filters cry wolf, as brief backwardations and weak breadth do, and each false alarm forgoes premium. They arrive late, as liquidity gauges often do. Their thresholds also drift as a market's structure changes, so a level that worked in one decade misfires in the next, and the layered stack reduces each failure without removing it.

Decision rule

Size short volatility by the stack: full size only when every light is green, 75% with one yellow, 50% or less with two or more.
Act on persistence, not on a single close: 3 closes to step down, 5 to begin coming back, and 3 weeks to return to full size.
Never trade a filter directly. Hold no VXX, UVXY or VIX futures position because of a filter reading.
Give GEX only 10% to 15% of one filter's weight, and ignore it when it disagrees with the term structure or the realized-volatility gap.

Common mistake: "if a filter has predictive value, I should trade it directly"

Predictive value about volatility is not a tradable directional edge. The VIX has predictive value for stress, but trading it through VXX or UVXY costs 5% to 15% a year in roll decay. The information is real; the implementation is what loses money. Filters modulate other strategies, and they are not strategies themselves.

The negative space matters as much as the positive, and Chapter 10I closes Part X with the edges that are rejected outright, explaining why each one fails.

What to carry forward

A filter sets a multiplier on positions you already hold. Its value is the losses it avoids, net of the premium it forgoes, and it gets no allocation of its own.
The VIX/VIX3M ratio is the canonical filter: halt new short volatility above 1.00, halve the sleeve above 1.05, and close all but deep tails above 1.15, each after 3 closes.
Realized volatility above implied, liquidity stress in the TED spread, the MOVE index or HY OAS, and narrow breadth each cut size further.
The sign of dealer gamma is low-confidence context in the Edge Book, and a central input in the Dealer Mechanics Lab chapters.
Stacked filters catch more stress than any single one, at the cost of more false alarms.

9 · Check yourself

Question 1

State the canonical VIX term-structure filter, with its phased response.

Answer

Watch the ratio of the VIX to the VIX3M. Above 1.00 for 3 closes, halt new short-volatility entries; above 1.05 for 3 closes, halt all put sales and cut the premium-selling sleeve by 50%; above 1.15 for 3 closes, close non-essential short volatility. After stress, once the ratio has been below 1.00 for 5 closes, re-enter at 25%, then 50%, then 100% of target over 3 weeks.

Question 2

Why is GEX downgraded from an edge to a low-confidence filter?

Answer

Public GEX models are model-dependent, because vendors assume different holders for each strike. The explosion of 0DTE options after 2022 changed the structure those models assumed. And a Cboe study of 2025 put 0DTE market-maker hedging at about 0.2% of SPX daily liquidity, against the narrative that dealer flows dominate. So the sign is worth 10% to 15% of one filter's weight, and is ignored when it disagrees with the other filters.

Question 3

What three indicators define the liquidity filter, and what does it do when it fires?

Answer

The TED spread, the MOVE index and the high-yield option-adjusted spread, with rough thresholds of 200 bps, 150 and 600 bps, any one of which fires the filter. You then cut leveraged and short-volatility exposure by 50%, cut equity beta by 25% and halt new tactical positions until the gauges normalize.

Question 4

Why does layering filters give a better signal than any single filter?

Answer

Each filter alone has a mediocre signal-to-noise ratio. Requiring all of them to be green before full size is parallel testing, which catches stress that any single gauge misses: it raises sensitivity at the cost of more false alarms. In the parallel-testing example, two tests of 70% sensitivity reach 91% together. Since missing a crash costs far more than pausing a sale, the extra false alarms are worth paying for.

Question 5

Hypothetically, the VIX is 24 and the VIX3M 22.5, realized volatility is 26% against implied 24%, the TED spread is 60 bps, the MOVE 135, HY OAS 520 bps, and GEX is negative. What size does the stack allow a put-writing sleeve?

Answer

The ratio is $24/22.5 = 1.07$ , so the term structure is off green. If it has held above 1.05 for 3 closes, its own rule halts all put sales and cuts the sleeve by 50%. Realized volatility above implied, the MOVE above 100 and HY OAS above 500 bps are three more lights off green. With two or more, the stack says 50% or less, and with the term-structure rule firing you open nothing new. GEX agrees, which adds only slight confidence.

Question 6

A newsletter proposes the AAII sentiment survey as a filter. Does it meet the three requirements?

Answer

It is measured weekly, not continuously. Its predictive value is noisy, and the Edge Book allows sentiment indicators only a contrarian use. Its use usually rests on a story about crowds being wrong at extremes. It fails the tests, so it is at most context, never a sizing rule.