Part XIII · Discipline · Chapter 13G

The portfolio

Allocation, stress tests, kill criteria and review

1 · The reference allocation

The Edge Book offers a reasonable default for a $1 million portfolio committed to the framework:

Three notes go with the table. The capital column is dollars, while the risk column weights each sleeve by its volatility. The Edge Book says trend has higher volatility per dollar than equity, so the same dollars give it more risk. The cash buffer is not dead capital. It is the option to buy in a stress, to meet a margin call without forced liquidation, and to absorb a drawdown in the tactical sleeve without compromising the foundation. And cross-asset carry is optional, given how few retail vehicles reach it.

The Edge Book summarizes the allocation in four lines:

• 70% to 80% in foundation premia, the durable risk premia of Chapters 10B to 10E.
• 5% to 10% in overlays, a modest enhancement and not a foundation (Chapter 10G).
• 10% to 15% in a cash buffer, as insurance against forced liquidation.
• Factor tilts inside the equity sleeve, not as a separate line (Chapter 10F).

The table and its summary do not quite agree. The table's foundation rows allow 70% to 95% of the capital and its cash row 5% to 15%, wider than the summary's 70% to 80% and 10% to 15%. The worked example later in this chapter holds 85% in foundation sleeves and 7.5% in cash, inside the table's ranges but outside the summary. The summary is the more cautious of the two, since it keeps more cash.

2 · Allocating by risk

Allocate by risk contribution, not by dollar weight. The Edge Book's illustration is that a 15% allocation to trend may contribute 18% to 20% of the portfolio's risk, while a 15% allocation to long bonds may contribute 8% to 10%. Dollar weights mean little without the risk weights beside them.

Chapter 1F wrote the variance of a portfolio as a quadratic form. Collect the sleeves' weights in a column vector $\mathbf{w}$, with $w_i$ the share of the portfolio in sleeve $i$, and their variances and covariances in the covariance matrix $\boldsymbol{\Sigma}$. The portfolio's volatility $\sigma_p$ then satisfies

The risk contribution of sleeve $i$ is its weight times the rate at which the portfolio's volatility rises as that weight grows:

The term $(\boldsymbol{\Sigma}\mathbf{w})_i$ is the $i$-th entry of the matrix $\boldsymbol{\Sigma}$ times the vector $\mathbf{w}$: the covariance of sleeve $i$ with the whole portfolio. So a sleeve contributes risk through its own volatility and through how it moves with everything else. The contributions add up exactly to the portfolio's volatility, which is what makes them a split of the risk.

Why the contributions add up to the whole

Write (13G.1) out as a double sum over the sleeves:

$$\sigma_p^2 = \sum_i \sum_j w_i\, w_j\, \Sigma_{ij}$$ (13G.3)

Differentiate with respect to one weight, $w_i$. The weight appears in row $i$ and in column $i$, and $\boldsymbol{\Sigma}$ is symmetric, so both give the same terms:

$$\frac{\partial \sigma_p^2}{\partial w_i} = 2 \sum_j \Sigma_{ij}\, w_j = 2\,(\boldsymbol{\Sigma}\mathbf{w})_i$$ (13G.4)

By the chain rule, the left side is also $2\sigma_p\,\partial \sigma_p/\partial w_i$. Dividing by $2\sigma_p$ gives $\partial \sigma_p/\partial w_i = (\boldsymbol{\Sigma}\mathbf{w})_i/\sigma_p$, the second line of (13G.2).

Now add the contributions: the sum $\sum_i w_i\,(\boldsymbol{\Sigma}\mathbf{w})_i$ is $\mathbf{w}^\top\boldsymbol{\Sigma}\mathbf{w} = \sigma_p^2$, so

$$\sum_i \text{RC}_i = \frac{\sigma_p^2}{\sigma_p} = \sigma_p$$ (13G.5)

Each sleeve's share of the risk is therefore $\text{RC}_i/\sigma_p$, and the shares sum to 1.

In practice, estimate the volatility of each sleeve's returns and the correlations between them, compute the risk contributions, and adjust the dollar weights until the contributions match your target. The example also shows that the table's note and its numbers can pull apart. With trend at 12% volatility and uncorrelated with equity, its share of risk falls below its share of dollars, the opposite of the note that trend has higher volatility per dollar. Which way it goes depends on the fund you hold and on its correlation with the rest of the portfolio, so compute it with your own estimates.

A simpler heuristic

For most retail portfolios the Edge Book suggests equal risk contributions across the foundation sleeves, the idea behind risk parity. If equity's volatility is 16%, trend's 12% and the VRP's 10%, equal contributions need more dollars in trend and the VRP than in equity. With zero correlations, each contribution is proportional to $w_i\sigma_i$, so the weights must be proportional to $1/\sigma_i$. That gives 1/16, 1/12 and 1/10, or 25%, 34% and 41% of the foundation's dollars. Each sleeve then has the same product of weight and volatility, $0.254 \times 16 = 0.339 \times 12 = 0.407 \times 10 = 4.07$, and so the same contribution, $(w_i\sigma_i)^2/\sigma_p$.

But equal risk contribution often leaves too little in equity, so the Edge Book offers a pragmatic target instead:

These shares are approximate, and as given they sum to 90%, not 100%. The Edge Book does not assign the last tenth, so treat the figures as rough proportions: about half the risk in equity, and more in the other foundation premia than in the overlays.

The figure splits three foundation sleeves both ways, by dollars and by risk. Start from the worked example's 60/40, then set trend's correlation with equity to −0.2 to reproduce the self-check, and switch to equal risk to see the heuristic's 25%, 34% and 41%.

dollars 60/40 equal equal risk

trend's volatility

12%

trend with equity

0

VRP with equity

0

equity: dollars 60%, risk 80%

trend: dollars 40%, risk 20%

VRP: dollars 0%, risk 0%

The portfolio's volatility is 10.7%: equity contributes 8.6 points, trend 2.1 and the VRP 0.0.

Sizing from a stress loss

A more direct rule sizes from the loss you can accept in a stress. The Edge Book notes that Ahmed's earlier framework for SPXW options used it:

Here $B_\text{exp}$ is the loss budget for the period, set explicitly in advance, and $L_\text{max}$ the maximum loss per trade or per position. The rule turns a budget into a contract count, mechanically and without intuition, as Chapter 13E's Example 10 works through.

The same logic works for a whole sleeve. If the sleeve's S2 stress loss is $X\%$ of the portfolio's net asset value (NAV) and your budget is $Y\%$, resize the sleeve to $Y/X$ of its current size. This is mechanical resizing from the stress test. In the table below, the VRP sleeve of $125,000 loses $50,000 in S2, 5% of NAV. With a budget of 3%, it shrinks to $3/5 \times \$125{,}000 = \$75{,}000$.

3 · Stress-testing the whole portfolio

Test the full portfolio under four stress scenarios, the three of Chapter 1D and a fourth for a gap like 1987's:

For a $1 million portfolio in the reference allocation, the Edge Book's worked example gives this table, in thousands of dollars:

The equity row is the sleeve times the index's fall: 10%, 20%, 35% and 22% of $550,000. The columns add up: in S3, $-193 - 110 + 30 - 15 - 15 = -303$. The premium-selling sleeves fall together with equity, and they should, because in liquidity events all such sleeves correlate at about 1.0 (Chapter 10H).

The other rows are the Edge Book's allowances, and they do not all follow from the cards of Chapter 13F. The VRP, trend and carry rows repeat, dollar for dollar, the foundation table of Chapter 10E, whose sleeves were larger: $150,000, $150,000 and $100,000. Checking them against the cards shows which way each one errs.

On balance the table is conservative, so a portfolio that survives it survives the card-based arithmetic with room to spare. But the trend row is not conservative, and it matters most exactly when you need it. If trend merely went flat in S3, the table's total would be −$333,000, a third of the portfolio.

What "survivable" means

The Edge Book reads each column in turn:

• S1 is survivable, absolutely, since a loss of 9% is a normal year.
• S2 is survivable too. A loss of 17% is painful but not catastrophic, and recovery is feasible in 2 to 3 years.
• S3 is survivable with discipline. A loss of 30% is severe, so do not add risk, do not capitulate, and hold the foundation.
• S4 is survivable. A loss of 22% on an overnight gap is recoverable, and the cash buffer prevents forced liquidation.

If your own stress table makes any of these "borderline" or "no", your sleeves or your tactical bets are too big, so resize them.

Chapter 1E turns each loss into a recovery. A drawdown $D$ needs a gain of $D/(1-D)$, and at a steady growth rate $g$ it takes $-\ln(1-D)/\ln(1+g)$ years. At a hypothetical growth rate of 8% a year, the four losses take the times in the table:

The S2 row agrees with the Edge Book's 2 to 3 years. S3 takes nearly twice as long as S2 to repair, though its loss is less than twice as large.

The gut check

The Edge Book calls one question the most important of all. If the portfolio is at S3, down 30%, and another 10% fall comes tomorrow, are you still rational, or do you exit in panic? The second fall takes the portfolio to $0.70 \times 0.90 = 0.63$ of its peak, a loss of 37% that needs a gain of 59% to repair.

If you would exit in panic, your behavioral sizing is wrong even when your mathematical sizing is fine. Reduce by 25% from what the math says is acceptable: a VRP sleeve the stress test allows at 4 contracts runs at 3, and a tactical sleeve allowed $100,000 runs at $75,000.

4 · Kill criteria at three levels

Kill criteria operate at three levels, and all three must exist. The first protects against a single strategy failing, the second against a whole category failing at once, and the third against the portfolio, or its owner, failing.

Level 1: each strategy

Each strategy has its own kill criteria, from the chapters that examine it:

• VRP harvest: halt new sales when VIX/VIX3M stays above 1.05 for 3 days, cut the size by 50% when realized volatility exceeds 30%, and close the sleeve when its drawdown exceeds 20% peak to trough (PTP). These are the three criteria of Chapter 10C. Card 02 and the phased filter of Chapter 10H halt new entries earlier, above 1.00.
• Trend: a drawdown in the vehicle above 25% PTP, or tracking error above 4% a year (Chapter 10D).
• EOQ rebalance: suspend it if the hit rate over 4 quarters is below 50% (Chapter 10G).
• Factor tilts: a 24-month rolling Sharpe ratio below 0 cuts the tilt by 50% (Chapter 10A).

Level 2: each category

Each category has its own rules, in case several strategies in it fail at once:

• Foundation: if 2 of the 4 foundation sleeves trigger their own kill criteria, reduce all foundation sleeves by 25%.
• Factor tilts: if 3 of 5 factors are in deep drawdown, suspend new factor allocations for 6 months.
• Overlays: if their combined P&L is negative for 2 years, simplify by dropping the costliest.
• Tactical: a sleeve drawdown above 30% suspends all tactical trading for 3 months.

The swing rules of Chapter 13C act sooner for the swing part of the tactical sleeve. Three losers in the last five trades pause new entries for a week, a sleeve drawdown above 15% halves the size for the next month, and one above 25% pauses the sleeve for 30 days.

Level 3: the portfolio and its owner

The Edge Book calls this level the most important. If the portfolio crosses these thresholds, automatic actions take effect regardless of the logic of any single strategy:

• A portfolio drawdown above 25% PTP: halt all discretionary tactical activity for 30 days. The foundation sleeves continue mechanically, and only the tactical bets are frozen.
• A degraded personal behavioral state: if you have been making impulse decisions, sleeping poorly or trading outside your written rules, halt new entries for at least a week.
• An account margin cushion below 30%: force the account to deleverage, closing the smallest tactical positions first.

The cushion is the equity above the broker's requirement, as a share of the equity (Chapter 13C). With $400,000 of equity against a requirement of $300,000 it is $(400{,}000 - 300{,}000)/400{,}000 = 25\%$, below the 30% line, so the smallest tactical positions go first.

The behavioral cooling-off period

The Edge Book insists that the cooling-off period, a 30-day halt after a 25% drawdown, is not optional. It calls it the single most valuable behavioral safeguard in the framework, for four reasons:

1. Drawdowns produce poor decisions, and the 30-day halt moves decisions to a later date, when your emotional state is better.
2. A forced halt prevents revenge trading and trying to "make it back".
3. The mechanical foundation keeps running, so you give up only the discretionary overlay, not the long-term anchor.
4. Knowing the halt exists in advance makes the drawdown less catastrophic psychologically, because you have a rule to point at.

This is the structural defense against the behavioral haircut of Chapter 13A.

5 · The monitoring cadence

The framework is reviewed on six clocks:

Added up, the cadence costs real time. Take 252 trading days, 52 weeks and working days of 8 hours: the daily checks take 21 to 63 hours a year, the weekly 26 to 52, the monthly 24 to 48, the quarterly 16 to 32 and the annual 8 to 16. That is about 95 to 211 hours a year, roughly 2 to 4 hours a week, before any trading.

6 · The five-year revalidation

Every 5 years, re-examine the entire framework with six questions:

1. Are the foundation premia still real? Has new evidence shown that the ERP, the VRP, trend or carry has changed structurally?
2. Has any factor decayed below your threshold? Should quality, momentum, value or low volatility be replaced or removed?
3. Has any overlay decayed? The FOMC cycle, OPEX, the end-of-quarter trade?
4. Has any rejected strategy been validated again? Sometimes the literature reverses, as with value adjusted for intangibles in 2024.
5. Are the kill criteria still calibrated? Perhaps the 25% drawdown threshold should be 20%, or 30%.
6. Is the conviction score still working? Are high-scoring trades actually beating low-scoring ones in your account?

Without revalidation, the framework calcifies. The pre-FOMC drift was a real edge until 2015, and it is not now. People who did not revalidate kept allocating to it long after it stopped working, and the same risk applies to every edge in this book. The Edge Book chose 5 years as roughly the pace at which structural changes in the market accumulate enough to matter, without being dominated by short-term noise.

In practice, revalidation means five tasks:

• Read the recent academic literature on each foundation premium.
• Track 5-year rolling Sharpe ratios against the longer-term expectation.
• Test alternative implementations, for example whether AVUV beats IVE for the value tilt.
• Revisit your assumptions about haircuts and execution costs.
• Update the framework, this book included, where the evidence has changed.

If revalidation shows that 2 of the 4 foundation premia have changed meaningfully, the framework needs serious revision, not a tweak. Be honest with yourself.

7 · Behavioral safeguards

Chapter 13A documented that the behavioral haircut is large, and this section specifies the structural defenses against it. Five safeguards are non-negotiable:

1. Written rules for entry, exit and sizing: every strategy has a setup card (Chapter 13F).
2. Kill criteria committed in advance, at all three levels of this chapter.
3. A monthly strategy diary, recording what you actually did against what the rules said.
4. A quarterly variance review: if your account runs consistently 200 bps or more behind its benchmarks, the gap is behavioral.
5. The 30-day cooling-off after a 25% drawdown, the single most important safeguard.

Four more are optional but recommended:

• Mechanical execution where possible, such as scheduled rebalances and conditional orders.
• Review by a trading partner or spouse: someone else seeing the trades reduces decisions driven by ego.
• A delay on large new positions: wait 24 hours, and take the trade only if you still want it then. Most impulse trades fail this test.
• Capacity discipline: do not trade beyond your operational capacity. The Edge Book's example is 68 trades in a day, beyond any retail trader's capacity, and it reports historical evidence that such overtrading destroys returns.

What good looks like

A trader following the framework should find about 85% to 95% of the variance in their P&L explained by the intended strategies. About 5% to 15% comes from execution friction and minor errors, and almost none from "trades I did not plan to make". If that third share is large, behavioral discipline is not working, so add safeguards.

8 · The selectivity principle

The Edge Book counts one principle among the highest-yield insights of an earlier portfolio analysis: selectivity beats activity. In that analysis, a day of 68 trades was the most informative single observation about overtrading destroying returns. The BTF-style bot strategy that succeeded did so because of extreme selectivity: it skipped entire weeks when its criteria were not met.

Selectivity has a recognizable rhythm:

• Most days, no new positions.
• Most weeks, 1 to 3 new positions across the entire portfolio.
• Most months, 4 to 8 positions, each with an explicit conviction score.
• Most quarters, 1 or 2 deep dives that produce changes of thesis.

The swing chapters of the Dealer Mechanics Lab expect more trades from their narrower book. Chapter 13C puts the swing trader at no qualifying setup on 60% to 70% of days, 1 to 3 new positions in most weeks, and 6 to 12 positions a month with conviction scores of 0.5 or more.

Retail traders bias toward action, because action feels productive and doing nothing feels lazy or like missing an opportunity. The data systematically reward the opposite: doing nothing is the highest-EV move on most days. The pipeline of Chapter 13C imposes selectivity by its structure. Most observations do not pass interpretation, most interpretations do not pass the conviction score, and most ideas that pass the score do not pass sizing. The trades that remain have cleared all four gates, which is why they should be larger and more durable than the trades you would otherwise make.

9 · What this chapter is really about

Three things hold the portfolio together:

1. Capital allocation across sleeves that respects risk contribution and survivability in stress.
2. Kill criteria at several levels that protect the portfolio when a strategy, a category or your own behavior fails.
3. A revalidation cadence that keeps the framework alive over horizons of 5 years and more.

Without these, the success of any single strategy is irrelevant, because the portfolio dies in stress, the framework decays, or behavioral haircuts pull it into mediocrity. The Edge Book calls this structure the deliverable of the whole book. Foundation premia, factors, overlays, filters, conviction and the taxonomy all sit upstream of it, and the portfolio framework is where the abstract becomes actual.

10 · Check yourself