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

What not to trade

Edges that decayed, never held up, or suit only institutions

A textbook should teach negative space. Most retail trading content focuses on the strategies in this chapter, because they are vivid, easy to backtest and frequently traded, yet they do not survive an evaluation by evidence tier. Knowing what not to allocate to is at least as valuable as knowing what to allocate to. Each rejection in this chapter is reasoned from evidence, not from opinion.

1 · Three reasons to reject a strategy

There are three reasons to reject a candidate strategy:

1. Decayed. The edge was real but has substantially compressed, so trading the historical effect now produces near-zero or negative net returns.
2. Never robust. The "edge" was always thin or specific to a methodology. The narrative is appealing, but the evidence is weak.
3. Institutional-only. The edge is real but requires infrastructure unavailable to retail: data feeds, low-latency execution, cross-margining or large capital.

The pattern matters more than the list. When you meet a new strategy that is not actually allocatable, you should be able to file it in one of these three categories. Each is a different way for the net expected value of Chapter 1C to fail. Write a strategy's expected result per trade as its gross edge less its costs. A decayed edge has lost its gross edge, and a strategy that was never robust never had one that survives honest testing. An institutional-only edge keeps its gross edge, but only an institution's costs are low enough to leave it positive.

2 · Decayed: the drift before FOMC announcements

The historical edge. Lucca and Moench (2015, Journal of Finance, JF) found that from 1994 to 2011, equities earned about 49 bps in the 24 hours before scheduled FOMC announcements. The effect was concentrated in the stocks with the highest beta: high-beta stocks gained 80 bps or more, and low-beta stocks close to nothing. The mechanism was that uncertainty before the announcement led to risk-off positioning, which reversed as the announcement approached.

With about eight scheduled meetings a year, 49 bps each would add up to about $8 \times 0.49\% = 3.9\%$ a year from eight days in the market. That arithmetic is why the pre-FOMC drift drew so much attention.

Why it is rejected. Kurov, Halova, Wolfe and Gilbert (2021, JFEcon) documented that the effect disappeared after 2015, because of the Fed's transparency reforms. They name the introduction in 2014 of the "dot plot" projections, the chart of each policymaker's forecast for rates, and the practice of forward guidance. Investors received more information between meetings, which reduced the premium for uncertainty on the day of the announcement. The realized drift from 2015 to 2024 was near zero.

The lesson. The decay risk of institutional and structural overlays is real, and it can be sudden. The mechanism has to persist for the edge to persist. When the mechanism changes, as the Fed's transparency practices did, the edge can vanish even with no rise in retail awareness.

The Dealer Mechanics Lab's own do-not-trade cards reject it too (Chapter 13F). The FOMC trade the Lab does teach comes after the announcement, as dealers unwind their hedges (Chapters 7B and 11K), and Chapter 10G sets it beside the Edge Book's view.

3 · Decayed: the S&P 500 inclusion premium

The historical edge. Stocks added to the S&P 500 historically rallied about 7% between the announcement and the effective date (Shleifer 1986; Harris-Gurel 1986). The mechanism was that index funds were forced buyers, and arbitrageurs front-ran the flow of their buying.

Why it is rejected. Greenwood and Sammon (2024, JF) documented that the S&P 500 inclusion effect has since fallen below 1%, for three reasons:

• Index funds have become smarter about positioning before the effective date.
• Inclusion announcements are integrated into the market's broader awareness.
• Active arbitrageurs have largely eliminated the obvious mispricing.

The 7% effect of the 1990s has shrunk to 50 to 100 bps, which is arguably statistically insignificant.

The lesson. Edges with a clear mechanism can decay almost completely when capital exploits them efficiently. An institutional edge has a half-life that depends on how discoverable the underlying flow is.

No chapter of the Dealer Mechanics Lab teaches this trade. Chapter 2A describes the index's rebalances, and Chapter 1G uses the inclusion effect's decline to show how institutional edges end.

4 · Never robust at retail: naive earnings short volatility

The narrative. Implied volatility typically rises into a company's earnings report, as traders price the uncertainty of the announcement, and then falls sharply after it. That IV crush seems to open a window for selling premium, and many retail traders commit significant capital to selling volatility systematically around earnings.

Why it is rejected for retail. The key paper is the Adams-Neururer study: Adams and Neururer (2020), "Volatility Risk Premium and Stock Returns Around Earnings Announcements", in the Review of Financial Studies (RFS). Its finding is that for single names, at a size a retail account can trade, naive short volatility around earnings has a negative mean despite a positive median. Most of the time you make a small amount on the IV crush. Occasionally you lose a large amount, when the move exceeds what was expected: a revenue miss, a guidance cut or a surprise warning. The rare large losses dominate the frequent small gains.

If a trade wins an average of $g$ with probability $p$ and loses an average of $L$ otherwise, its expected result is

while its median is a gain whenever $p$ is above one half. The two can have opposite signs, which is the whole story of this strategy.

This is the structural shape of Chapters 1B and 1D, applied to one company at a time. Short volatility is short skew, and for a single name at retail size the skew tax is too large for the premium to cover.

Why it works for institutions. Cao, Han, Tong and Zhan (2022, RFS) document a profitable institutional version: portfolios that sell volatility across 200 or more names, built on the dispersion between them, with significant compensation on the short leg. It requires:

• The capacity to hold positions in 200 or more names.
• Cross-margining, so that positions offset one another across the book.
• Infrastructure to manage risk.
• Sophisticated analytics of the implied volatility surface.

Retail accounts cannot replicate this at scale. A single-name version does not capture the cross-sectional benefit; it takes only the negative-mean exposure. Diversification alone would not fix that. Two hundred bets with a mean of −$1.20 lose about $240 a round, only more steadily. The institutional version earns because it is built differently, and its breadth keeps one blowup from sinking the book.

The least-bad retail variant: calendar spreads on liquid names. A calendar spread that is long a longer-dated option and short a shorter-dated one sells the inflated front-month volatility while keeping exposure further out. It is defined-risk, its downside on a miss is limited, and it captures a moderate part of the IV crush. Even this is a small edge, better used for occasional small bets than as a sleeve.

The Dealer Mechanics Lab agrees. Its do-not-trade cards reject single-name earnings short volatility on the same paper (Chapter 13F). Its pre-earnings momentum trade closes before the report (Chapter 11I), and its selection matrix meets an earnings week with a diagonal spread rather than short premium (Chapter 12A).

5 · Never robust: scalping gamma-exposure levels

The narrative. Public GEX models, such as those of SqueezeMetrics and SpotGamma, identify "call walls" and "put walls": strikes where dealer gamma exposure is concentrated. Retail traders sell at the call wall and buy at the put wall, expecting dealer hedging to reverse moves at those levels.

Why it is rejected. Three reasons:

1. Public GEX models are model-dependent. Vendors give different numbers because they make different assumptions about who is long and who is short which strikes. The "ground truth" position cannot be known from public data.
2. The explosion of 0DTE options after 2022 changed the structure. Models built before 2022 assumed that monthly and weekly options dominated. With more than 40% of SPX volume in 0DTE options, on the Edge Book's count, models built for the old structure need not reflect dealer positioning.
3. A Cboe study of 2025 estimates 0DTE market-maker hedging at about 0.2% of SPX daily liquidity. That directly contradicts the narrative that dealer flows dominate the day. If hedging is 0.2% of liquidity, the level effects it creates are small relative to other drivers. The study is Mandy Xu's "0DTEs Decoded" (May 2, 2025). Because it measures same-day hedging only, reading it against walls built mostly from longer-dated open interest extends the measurement.

What the academic literature does support. Garleanu, Pedersen and Poteshman (2009, RFS) established that demand pressure on options affects their prices, the theoretical basis for thinking that dealer positioning matters. But the effect operates at the level of option prices, not the price of the underlying. Its magnitudes are small per unit of demand, and specific rules for trading levels are not validated in the academic literature.

The valid use is the directional sign of GEX, as low-confidence regime context (Chapter 10H). Negative-GEX days tend to have higher realized volatility, and positive-GEX days compressed realized volatility. Use the sign for sizing, not for trading levels. Chapter 13B scores level scalping with the conviction score and gets 0.005, effectively zero.

What the Dealer Mechanics Lab chapters claim

The Lab builds the walls and the gamma flip from open interest (Chapter 8C), and it reads them as more than context. Its dashboard treats the largest call wall above spot as potential resistance and a magnet for pinning, and the largest put wall below as potential support. It calls spot within 0.5% of the flip a transitional regime (Chapter 11A). Its conviction score rewards a long thesis whose target sits well below the call wall and marks down one with a call wall directly overhead (Chapter 13B). With positive GEX, it expects range trading to win and breakouts to fail (Chapter 8D).

The Lab also measures how small the effects are. It reports pinning of about 16.5 bps on expiration days, and calls published GEX "a useful prior on typical days". It also reports gamma effects of about 16 bps per standard deviation, a figure that appears in no version of the paper it cites: Barbon and Buraschi give about 12 bp for single stocks and more than 20 bp for indexes. Its own do-not-trade cards reject scalping level by level, and they use the sign as a regime indicator (Chapter 13F). So both traditions reject the scalp. They differ on whether the levels deserve any weight at all. The Lab uses them to judge a trade's path and targets, while the Edge Book lists open-interest concentration as descriptive, not predictive, and keeps only the sign.

The pinning figure needs its scope. It is Ni, Pearson and Poteshman's (2005) lower bound for optionable single stocks from 1996 to 2002, not an index figure. For the index, Golez and Jackwerth (2012) found S&P 500 futures pulled toward the at-the-money strike when options on them expired, shifting at least $115 million of notional per expiration day.

6 · Never robust: systematic iron condors

The narrative. An iron condor is "defined-risk": it sells an out-of-the-money (OTM) call spread and an OTM put spread, which caps the loss on each side, and it profits in range-bound markets. Markets are usually range-bound, so condors usually win, and many retail "income" strategies are built on them.

Why it is rejected. The Cboe CNDR index returned about 690% from 1986 to 2024, while the SPX Total Return index (SPX TR) returned about 5,500% over the same period. That is an empirical rejection spanning decades: systematic condors underperform massively over long horizons.

A total return turns into a compound annual rate $c$ through the growth factor of Chapter 1A. If a value grows from $V_0$ to $V_T$ over $T$ years,

Why condors fail.

• Each leg has negative skew, and combining two short legs does not offset the skew: it adds to it.
• Condors are vulnerable to one-sided breaks. When SPX trends, one wing gets blown through and the "defined" maximum loss triggers.
• Tail risk is asymmetric, since both wings can be tested, in different scenarios.
• Theta accumulates slowly, and a few wing breaks a year erase a year of premium.

The deeper lesson. "Defined risk" is an operational property, in that you know your maximum loss, but it is not a property of the P&L. Systematic condors with explicit caps on their losses still underperform, because wing breaks happen often enough to outweigh the theta collected.

What the Dealer Mechanics Lab chapters claim

The Lab teaches the condor as an intermediate neutral structure, a pure short-volatility bet (Chapter 12C). It is best when realized volatility is expected to come in meaningfully below implied, and worst when the index breaks out. It reports the CNDR index's underperformance itself, over the index's history from a backtest starting in 1986 and live data from 2007, and it draws the index against SPX with illustrative, synthetic series. It reads the lesson as specific to the condor's structure, since the PUT index's systematic put-writing has competed well on risk-adjusted terms (Bondarenko 2019). Chapter 5E, from the same tradition, counts condor writing among the short-premium strategies with positive long-run returns. Its strategy matrix picks a defined-risk condor when implied volatility is rich and the market is directionless, and sizes it at 0.25% of the portfolio at risk rather than 1% (Chapter 12A).

The two traditions therefore agree on more than they seem to. The Lab's own do-not-trade cards reject systematic condors on the same CNDR record (Chapter 13F), and the Edge Book allows the occasional discretionary condor. They differ in emphasis: the Lab teaches the condor as a regular tool, chosen for a regime and sized small, while the Edge Book treats it as an exception. The open question, as Chapter 10A puts it, is whether a condor chosen for the regime escapes the record of the mechanical index.

7 · Leverage with extra steps: the put spread and LEAPS call

The narrative. "Sell 1-year SPX put spreads to fund LEAP calls. The premium pays for the LEAP exposure. Portfolio margin makes this efficient. The result is a leveraged equity exposure with positive edge from VRP harvest." This is a popular framing in retail content, especially among traders with portfolio margin accounts. A LEAP is a long-dated listed option, one of the LEAPS.

Why it is rejected. The structure is economically equivalent to about 1.4 times SPX exposure, depending on the specific deltas, with worse path properties. Chapter 13E works the example in full, with a decomposition of its Greeks and its P&L. Decomposed, it is three things:

• The long LEAPS call is roughly long SPX with a financing cost built in: 50 to 150 bps a year of embedded leverage drag (Frazzini-Pedersen 2012, 2022).
• The short put spread is a small harvest of the variance risk premium with a bounded loss, "−2 to −3 vol points realized after costs" in the Edge Book's words.
• Net: long about 1.4 times SPX, with embedded drag and margin requirements that depend on the path.

Three simpler alternatives give equivalent or better economic exposure, with less operational complexity and better path properties:

• ES futures: 1.4 times exposure with cleaner financing, at a basis of about SOFR plus 10 to 20 bps.
• Box spread financing plus SPY: similar economics, with no dependence on the mark-to-market (MTM).
• A 2 times SPX leveraged ETF (SSO): an expense ratio of 0.85%, but no LEAPS to roll and no put spread to manage.

Why the "edge" claim is wrong. The argument is that the VRP makes the structure profitable. But:

• The VRP from 1-year put spreads, after retail costs, is roughly 1 to 3 volatility points net.
• The embedded leverage drag in LEAPS is roughly 50 to 150 bps a year.
• The net "edge" over simple leverage is approximately zero, or small.
• The path properties are worse: margin expands under stress, and the position can be force-liquidated.

It is leverage, dressed up; calling it "edge" overstates what it adds. Chapter 1E warned against this structure as the leverage trap.

No chapter of the Dealer Mechanics Lab teaches this combination. Its nearest relative in the strategy atlas, the poor man's covered call of Chapter 12C, pairs a LEAPS call with short calls instead, and its long leg carries the same embedded drag.

8 · Catastrophic tails on both sides: VIX exchange-traded products

Long VIX ETPs. VXX has lost 99.84% from its inception in 2009 through 2024. The losses come from continuous roll decay. The VIX term structure is in contango most of the time, so the exchange-traded product keeps selling expiring futures for less than it pays for their replacements. Chapter 10A works the arithmetic. The yearly roll decay is 5% to 15% in normal times, and much higher in extended periods of contango.

As a portfolio hedge, the carrying cost dominates. You would need very specific timing, entering near an inversion of the term structure and exiting before it mean-reverts, to come out ahead after costs, and most retail attempts fail.

Short VIX ETPs. XIV was terminated after losing 96% in a single day, in February 2018. The Bank for International Settlements puts the loss at 84% (Chapter 8H). SVXY has since been de-leveraged but is still subject to catastrophic single-day losses in volatility shocks. Chapter 8H tells the story of that day. The "short vol" trade through an inverse VIX ETP has negative skew even more extreme than that of direct put-writing, for three reasons:

• VIX moves are highly non-linear: a 5-point spike in the VIX is much more than a 5% move in SPX.
• The ETP's daily rebalancing forces it to buy volatility after a stress day, selling at the worst time.
• Termination clauses can wipe out the position.

The Dealer Mechanics Lab chapters agree that the VIX is a gauge, and they put VXX's roll drag at about 10% a year (Chapter 9A). They do use one VIX derivative: long-dated VIX calls, which Chapter 13H names the cleanest hedge for a book of short vega and Chapter 12F lists as the first choice for that purpose. Those are options on the VIX, not ETPs, so they fall outside this rejection, though the Edge Book's own advice for volatility exposure points to SPX and XSP options instead.

9 · What looks like edge but is not

When you evaluate any candidate strategy, these are red flags:

The textbook rule is short: a real edge survives critical evaluation, and one that does not is narrative.

10 · The complete map of rejection

Each item has at least one specific empirical or academic rejection, and none should be allocated to.

The figure turns the map into practice. File each card in a drawer before you look back at the map. For each one, the figure gives the evidence and, when a drawer is wrong, the reason it does not fit.

card 1 of 9 · filed 0 · right 0

Pre-FOMC drift. Buy the index for the 24 hours before each scheduled FOMC announcement.

file as decayed never robust

or institutional-only catastrophic

Choose a drawer. Decayed, never robust, institutional-only, or catastrophic costs or tails?

previous card

next card

Strategy	Your drawer
Pre-FOMC drift	—
Systematic iron condors	—
Long VIX ETPs	—
Earnings volatility across 200 names	—
GEX-level scalping	—
S&P 500 inclusion	—
Naive earnings straddles	—
Short VIX ETPs	—
Put spread plus LEAPS	—

11 · What this leaves you with

After Chapters 10B to 10I, the universe of allocatable strategies is manageable:

• Foundation premia: the equity risk premium (ERP), held as long equity (Chapter 10B), and the VRP, through cash-secured at-the-money (ATM) XSP put-writing only (Chapter 10C). Trend following, through funds such as DBMF or KMLM (Chapter 10D), and cross-asset carry, a small allocation where you can reach it (Chapter 10E), complete them.
• Factor tilts: quality (QUAL), momentum (MTUM), value (a RAFI fund or AVUV), low volatility (USMV) and size with quality (AVUV) (Chapter 10F).
• Structural overlays, implemented cheaply only (Chapter 10G): the FOMC even-week tilt through the sizing of equity, end-of-quarter (EOQ) rebalancing at a very small size with an explicit direction, and Halloween through the annual rebalance.
• Regime filters: the VIX term structure, canonical and phased; realized against implied volatility; the liquidity composite; and the sign of GEX as a low-confidence supplement (Chapter 10H).

The universe is manageable for a reason: most of what is marketed as "edge" does not survive evaluation. The taxonomy and the conviction framework of Chapter 13B filter aggressively.

Chapters 10B to 10H covered the strategies you can allocate to, and this chapter those you should specifically not trade. Together they complete the taxonomy. Part XIII turns the concepts into judgment: ten worked examples in Chapter 13E walk the full chain of reasoning end to end. Chapters 13F, 13C and 13G then give the setup cards, the pipeline from signal to trade and the kill criteria, and the reference chapters, R1 to R4, collect the tools.

12 · Check yourself