Part XIII · Discipline · Chapter 13C

From signal to trade

The five-step pipeline and the kill rules

1 · One grammar for every action

Every action in your portfolio should pass through the same five steps: observation, interpretation, trade implication, sizing and invalidation. This grammar makes high-conviction trading operational rather than aspirational. Most retail trading fails by skipping steps, observing and then trading without an explicit interpretation, size or kill condition.

Each step is distinct and required: skipping any one of them produces undisciplined trading. The conviction score of Chapter 13B gates the third step. Observations and interpretations arrive whatever you do, while the score decides whether they become positions. If conviction is insufficient, you log the observation and watch it, but you do not trade it. That is what "selective" means in practice. Most observations never become trades, while the trades that exist have had all five domains checked, scored and survived.

2 · Step 1: observation

An observation answers "what did I see?" with a specific, measurable market state. One example is "SPX's RSI(2) closed at 5 today, with Stage 2 confirmed by a 30-day SMA slope above 0.05 and an ADX of 22" (Chapters 11C and 11E). "The market feels oversold" is not an observation. The Edge Book gives four more that qualify:

• The ratio of the VIX to the VIX3M crossed 1.05 today.
• The 21-day realized volatility of SPX is 32%, against an implied volatility of 26%.
• The TED spread widened from 25 bps to 95 bps over 5 days.
• MTUM has underperformed SPX by 8% over the trailing 6 months.

Four others fail, each for its own reason:

The pipeline starts with measurable inputs only, so if you cannot define the threshold, the observation step has failed.

Keep a written list of observation triggers, the specific measurable conditions that earn an interpretation. For most traders, about 5 to 10 triggers cover 95% of legitimate trade situations, leaving the rest to be researched as they arise. The Edge Book's list has eight, each with its role:

That is the whole list: about 8 observation triggers cover the systematic decision space.

3 · Step 2: interpretation

Interpretation answers "what does it mean?" by mapping the observation to a regime or a risk update, the bridge between data and action. For the swing observation above, the Dealer Mechanics Lab's interpretation reads as follows. RSI(2) below 10 in Stage 2 with positive gamma exposure (GEX) means that dealer hedging fades the pullback, providing mechanical buying support at the EMA. A good interpretation does three things:

1. It translates the observation into a characterization of the regime.
2. It refers to the body of evidence behind that reading.
3. It is specific about which risks are now higher and which lower.

Interpretation fails in three common ways:

• Over-interpretation. "The VIX is backwardated, so the market is about to crash" is prediction dressed as interpretation. Backwardation indicates elevated near-term volatility, not its direction, which is unpredictable.
• Under-interpretation. "The VIX is backwardated, and I do not know what to do." If you cannot translate a signal into a characterization of the regime, the signal is not useful.
• Narrative interpretation. "The VIX is backwardated because the AI bubble is popping." The story may or may not be true. The interpretation should rest on the regime, not on a story.

Write an interpretation for each trigger in advance, so that when a trigger fires you read its interpretation rather than composing a new one. The discipline lies in committing to what each signal means before it fires.

4 · Step 3: trade implication

The trade implication answers "what action does it imply?" with a specific entry, direction, instrument and timing, or with a decision to do nothing. This is where the conviction analysis of Chapter 13B happens.

Sometimes the right implication is to do nothing new:

• A filter fires while you hold no position it would change.
• The observation falls within the range of a normal regime.
• The conviction score is below 0.3.
• The action is already in place, so nothing more is needed.

"Do nothing" is a valid output of the pipeline. Most observations should produce no new trade. When the conviction score is below 0.3, the implication is "no new position": record the observation in your journal and do not act on it. Selectivity is operational: on most days, most observations produce nothing.

5 · Step 4: sizing

Sizing answers "how much?" from an explicit loss budget, never from intuition.

Risk is not the size of the position

"Risk 1% of the portfolio per trade" means that if the stop is hit, the realized loss equals 1% of the portfolio's value, not that you deploy 1% of your capital. If the stop sits a distance $d$ below the entry, the position is the dollar risk budget divided by that distance.

Whether that notional is acceptable depends on four things:

• Liquidity: can you exit cleanly under stress?
• Margin: do you have the buying power?
• Correlation: what is the aggregate risk with your existing positions?
• Gap risk: can the price jump through the stop (Chapter 5F)?

For options the relation is structurally different, because the notional is bounded by the premium for a long option and by the collateral for a short one.

The two readings of "1%" differ by a factor you can compute. Sized by risk, the position holds $B/d$ shares, where $B$ is the dollar budget. Sized as 1% of capital, it holds $B/P$ shares at a price $P$. Their ratio is $P/d$, the price divided by the stop distance. A stop 1% below the price makes it 100, and a stop 2% below makes it 50.

The first figure separates the two numbers: set the portfolio, the share risked per trade, the entry, the stop, the contract multiplier and the margin available. The readouts give the dollar risk, the stop distance, the loss per contract, the position size and the notional it controls, with the notional as a share of the portfolio. The defaults risk 1% of $250,000 on a $600 entry with a $594 stop and a multiplier of 100: 4 contracts, a notional of $240,000, or 96% of the portfolio. A table prices a gap through the stop of two and three times the stop distance, while the verdict sets the notional against the margin available.

portfolio

$250k

risk per trade

1.00%

entry

$600

stop

$594

multiplier

100

margin available

$500k

dollar risk = $2,500

stop distance = $6.00

loss / contract = $600

position size = 4

notional deployed = $240,000

notional % of portfolio = 96.0%

planned risk / size ratio = 1.04%

Gap-through stress test	Loss	% of portfolio
2× gap	−⁠$4,800	1.92%
3× gap	−⁠$7,200	2.88%

Margin: GREEN. Within 70% of available margin.

Set the portfolio to $500,000 and the multiplier to 1 to reproduce the example: 833 shares and $499,800 of notional, which turns the margin verdict yellow. Moving the stop to $598 then grows the position to 2,500 shares and $1.5 million of notional, and turns the verdict red. The risk percentage controls what you lose if the stop holds. The notional percentage controls what happens if it does not.

The figure's gap test scales with the stop, so a tighter stop makes the gap look smaller. A real gap is a move in the price and does not care where your stop sits. At $1.5 million of notional, a 3% gap costs $45,000, 9% of the portfolio, or nine times the budget.

The loss-budget rule

The loss budget $B_{\exp}$ is the dollar loss you have decided in advance to absorb on a trade or a sleeve. If $L_{\max}$ is the loss on one unit (a share or a contract) at your stop, the number of units $n$ must satisfy

The brackets are the floor function, which rounds down to a whole number: $\lfloor 2.5 \rfloor = 2$. Rounding down keeps the rule conservative, since rounding up would breach the budget.

The example treats each contract as moving point for point with 100 units of XSP, as a deep in-the-money call nearly does. For an option with a lower delta, estimate its loss at the stop from its delta and its price (Chapter 5C).

The second figure computes (13C.1), with sliders for the portfolio, the risk per trade, the stop distance and the multiplier. It returns $B_{\exp}$, $L_{\max}$, $n$ and the planned risk, and prices gaps through the stop of two and three times its distance. Its verdict reads the 2× gap: under 3% of the portfolio is green, under 5% yellow, and beyond that red. The defaults, $250,000 at 1% with a $2 stop on a multiplier of 100, give 12 contracts and a planned risk of $2,400. Set the portfolio to $500,000 and the stop to $20 to reproduce the example.

portfolio size

$250k

risk per trade

1.00%

stop distance ($)

$2.00

contract multiplier

100

loss budget B_exp = $2,500

loss per unit L_max = $200

n = ⌊B_exp / L_max⌋ = 12 contracts

planned risk = $2,400

planned risk % = 0.96%

notional at a 2% stop = $120,000

Gap-through scenario	Loss	% of portfolio
2× gap	−⁠$4,800	1.92%
3× gap (severe)	−⁠$7,200	2.88%

GREEN. 2× gap-through still under 3% portfolio loss.

The example gives 2 contracts and a planned risk of $4,000, 0.80% of the portfolio. Its 2× gap costs $8,000 (1.60%) and its 3× gap $12,000 (2.40%), both green. The notional readout assumes, as the original widget did, that the stop sits 2% below the price. That places the price at 50 times the stop distance, $1,000 for a $20 stop, so read it as an illustration rather than as this trade's notional.

A budget for a whole sleeve

The same rule sizes a sleeve that has no stop at all. There, $L_{\max}$ is the loss on one unit in a stress scenario rather than at a stop, and $B_{\exp}$ is the loss budget for the whole sleeve.

Compute $n$ from the formula, write down the $L_{\max}$ you assumed for the sleeve, and stress-test it against S3. If your intuition disagrees with the formula's $n$, in either direction, check which is wrong: it is usually the intuition.

The portfolio's budget

Beyond each sleeve, the whole portfolio has a budget. The expected shortfalls of the sleeves must not add up to more than the portfolio's loss budget:

Expected shortfall has a useful property: the expected shortfall of a combination is never more than the sum of its parts' expected shortfalls. The sum reaches the actual figure only when the sleeves lose together in the tail. So (13C.2) is conservative in calm markets and accurate in a crash, when the tail correlation of premium-selling sleeves approaches 1.0. That is why it binds more often than sizing each sleeve on its own suggests.

Kelly and the loss budget: one question, two units

Chapter 3D derived the Kelly criterion, the fraction of capital that maximizes the expected logarithm of long-run wealth. This section's rule, (13C.1), sizes from a dollar cap per trade instead. A reader of Chapter 3D may think Kelly is the answer, while a reader of this chapter may think the loss budget is. They are not in conflict. They answer different questions, and converge for honest retail-scale edges.

• Kelly maximizes expected log growth. Its output is a fraction of capital, under an assumed geometry of binary outcomes, a win $G$ or a loss $L$ per dollar staked. It assumes that you have estimated the edge correctly, and that you commit the full bet to a single opportunity.
• The loss budget bounds the dollar loss per trade. Its output is a position size, under the geometry of a stop. It makes no claim about long-run growth. It caps the drawdown that any one trade can contribute.

Where they converge. Chapter 3D showed that with symmetric payoffs the Kelly fraction of capital at risk is $2p - 1$ for a win rate $p$. For a typical retail edge, a 52% to 55% win rate with near-symmetric payoffs, that is 4% to 10% of capital. Half-Kelly halves it and quarter-Kelly quarters it, so at 52% quarter-Kelly risks 1%, exactly the loss budget, and at 53% it risks 1.5%. The two land in roughly the same dollar position because both rest on one anchor: size in proportion to your honest edge, not to your confidence.

Where they diverge. For a strong edge, Kelly is far more aggressive. Chapter 3D's worked example takes a 60% win rate with a $4 average win and a $3 average loss on a $50 share. Full Kelly holds five times the account, while even quarter-Kelly holds 6,250 shares, five times the 1,250 shares that the 1% budget allows. For that edge the loss budget binds by a wide margin and sizes each trade well inside the Kelly ceiling. It stops being enough only if the true edge is much weaker than the estimate: below a 52% win rate at even payoffs, 1% at risk exceeds quarter-Kelly.

Divergence one: diversification. Kelly is set per strategy, so five uncorrelated strategies, each at quarter-Kelly for a 55% edge, put $5 \times 2.5\% = 12.5\%$ of capital at risk at the same time. The loss budget is set per trade: 10 concurrent trades at 1% each put 10% at risk. The two agree on the ceiling once you do the arithmetic, which is exactly where retail traders fail to do it. The addition holds only for uncorrelated bets, since correlated trades behave like one larger bet. The single-bet Kelly fraction then applies to their sum (Chapter 13D, Example 10).

Divergence two: an uncertain edge. Kelly assumes the edge is known, so over-estimating it produces over-betting, whose cost Chapter 3D worked out. Betting 1.3 times the true Kelly fraction keeps 91% of the maximum growth, the same as betting 0.7 times it, but raises the chance of ever halving the account from 28% to 69%. An edge estimated from a backtest typically carries a standard error of about 30%, so betting the point estimate is meaningfully too aggressive. The loss budget is robust to this, because it caps each trade's loss whether or not the edge is real. Most traders, this book's intended reader included, cannot trust their edge estimates. For them the loss budget is the safer first-line tool, with Kelly as the check on the aggregate.

The 30% line follows from the same arithmetic: five uncorrelated strategies with a 55% edge reach half-Kelly at $5 \times 5\% = 25\%$ at risk, so 30% is past it. If the open trades are correlated, they are a single bet, whose full Kelly at that edge is 10%. Then 30% is three times Kelly, where by Chapter 3D the growth rate is negative. Chapter 3D's worked example shows one trade's chain from end to end, while this rule keeps the aggregate honest.

6 · Step 5: invalidation

Invalidation answers "what would prove me wrong?" with the explicit conditions under which you close the position or reassess its thesis. It comes in two kinds, each with three triggers: trade-level invalidation says when this trade closes, while strategy-level invalidation says when the strategy is suspended or reassessed.

All three are written before the position is opened, since the discipline lies in committing to them in advance. The reason is behavioral: work in behavioral economics, by Thaler and by Sunstein, shows that decisions made beforehand in a calm state are more rational than decisions made afterward under stress. Stress activates loss avoidance and confirmation bias, which override consistent decision-making. Writing the invalidation down before the trade uses your calm self to constrain your stressed self, which is the operational meaning of discipline.

Every position therefore carries documented invalidation criteria. If you cannot say what would prove you wrong, the trade is not ready. Many bad trades fail this test because the entry rested on an unfalsifiable narrative, such as "the market should rally", rather than on a hypothesis that could be tested.

7 · Kill rules at three levels

Invalidation is written per trade, while the Dealer Mechanics Lab adds kill criteria at three levels: the setup, the sleeve and the whole portfolio.

The 25% drawdown and the 30-day cool-off

When the portfolio's drawdown crosses 25% from peak to trough, all discretionary tactical activity halts for 30 days. New entries are suspended, while any mechanical sleeves keep running (Chapter 13A). There are four reasons: drawdowns produce poor decisions, a forced halt prevents revenge trading, mechanical systems keep working through it, and knowing that the cool-off exists makes the drawdown less catastrophic psychologically.

The rule binds however good the next setup looks. It exists precisely to override your judgment during the period when your judgment is most compromised. By Chapter 1E, a 25% drawdown already needs a gain of $0.25/0.75 = 33\%$ to repair. Chapter 3D's illustration of drawdown discipline used a close-and-reset rule at 30%, while this book's portfolio rule halts at 25%.

Strategy, sleeve and portfolio

At the strategy level, each swing setup has its own kill rule:

The kill rules in the middle column come from the Dealer Mechanics Lab's chapter on the pipeline. The last column comes from its chapter on the setups, which Chapter 11H follows and the cards of Chapter 13F repeat. The two disagree on the Compression Breakout's time stop, 7 days against 8, and on the exit and time stop of the IBS + RSI setup, 3 days against 2. Whichever version you adopt, write it on the card and keep to it.

At the sleeve level, for each bucket of strategies:

• If 3 of the last 5 swing trades lost, pause new entries for 1 week and review the framework.
• If the sleeve's drawdown exceeds 15% from peak to trough, cut its sizing by 50% for the next month.
• If the sleeve's drawdown exceeds 25%, pause it for 30 days.

At the portfolio level, three rules apply:

• If the portfolio's drawdown exceeds 25% from peak to trough, halt all discretionary tactical trading for 30 days.
• If your personal behavioral state is degraded, halt new entries for at least 1 week.
• If the account's margin cushion falls below 30%, force a deleveraging by closing the smallest tactical positions first.

The margin cushion is the equity above the broker's requirement, as a share of the equity. In Chapter 1E's example, $100,000 of equity against a $40,000 requirement is a cushion of 60%. With the requirement at $75,000 the cushion is 25%, below the line, so the rule closes positions before a margin call can choose them for you.

"Personal behavioral state degraded" needs a measure, and the third figure supplies one. It scores seven factors, each from 0.1 to 1.0 by the guide under its name. They are last night's sleep, days since your last loss, any recent winning streak, whether the plan was written in advance, the fear of missing out (FOMO), today's screen time and external stress. Its readiness score is their product, which the figure labels S, a different S from the sizing factor of Chapter 13B. At 0.5 or more the verdict is green: trade your plan. From 0.20 it is yellow: reduce size by 30% to 50%, or skip new entries and manage existing ones only. Below 0.20 it is red: no new positions, and reduce existing exposure if possible. As with the conviction score, one bad factor sets the whole score, and the figure highlights it.

Factor and scoring guide	Score sᵢ
Sleep last night ≥7 hrs = 1.0; 6h = 0.7; <5h = 0.3	1.00
Days since last loss >3 = 1.0; 1-3 = 0.6; same day = 0.2	1.00
Recent winning streak none = 1.0; 3-5 wins = 0.7; >5 = 0.5	1.00
Plan written in advance? yes-clean = 1.0; rough notes = 0.6; in head = 0.2	1.00
FOMO signal low = 1.0; moderate = 0.6; high = 0.2	1.00
Screen time today under 2h = 1.0; 2-4h = 0.7; over 4h = 0.4	1.00
External stress low = 1.0; moderate = 0.6; high = 0.3	1.00

product score S = 1.000

average factor = 1.000

binding = —

GREEN. Behavioral state supports normal sizing. Trade your plan.

Set "days since last loss" to 0.2, a loss earlier the same day: the score can then be no more than 0.2, however well you slept. With a short night added, 6 hours at 0.7, the product falls to $0.2 \times 0.7 = 0.14$: red, no new positions. The figure's premise is that trades taken in a low state are systematically worse. Its rule sets the cost of missing a day against the cost of trading through it. In this state it typically judges the first to be smaller.

8 · The pipeline in action

The Edge Book walks the whole grammar through three situations: a tense morning, a slow allocation and the portfolio itself.

The result is no action today, plus conditions committed in advance for tomorrow. That is what the pipeline produces in an elevated regime that is not yet stressed. Most of the discipline shows up in the non-actions: not entering new positions because the filter is yellow.

The morning's rules use two thresholds, 1.00 for the first steps and 1.05 in the invalidation example above. The VRP harvest's card in Chapter 13F layers them: halt new entries above 1.00 for 3 closes, and cut the sleeve by 50% above 1.05. Whatever levels you choose, write them on the card before the morning they matter.

Running the pipeline even on calm, slow allocations is what prevents impulsive exits during their drawdowns. Chapter 10D's kill rules for the same sleeve reassess sooner, after three consecutive negative years rather than five, but both review rather than sell.

9 · Selectivity

Retail traders lean toward action, because action feels productive and doing nothing feels lazy. The data systematically rewards the opposite: on most days, doing nothing is the move with the highest expected value (EV). A competent swing trader running this framework should expect:

• Most days, no new positions. On 60% to 70% of days no setup qualifies, a share that Chapter 11N puts at 70% to 80%.
• Most weeks, 1 to 3 new positions across the whole watchlist.
• Most months, 6 to 12 positions, each with an explicit conviction score of 0.5 or more.

The weekly and monthly counts agree: 1 to 3 a week over about 4.3 weeks is 4 to 13 a month. The pipeline imposes this selectivity by its structure: most observations do not pass interpretation, most interpretations do not pass conviction, and the trades that remain have cleared every gate before sizing. This is the system working, not failing.

10 · How the pipeline fails

Five failures recur:

1. Skipping interpretation. "The VIX is up, sell calls" never says what a rising VIX means for the regime. Without an interpretation, the action is arbitrary.
2. Skipping sizing. "This is a great trade, I am going big." How big, and why? Without an explicit size from a loss budget, you have abandoned the discipline of survival.
3. Skipping invalidation. Saying "if it works, great, and if not, I can decide then" leaves every decision to your stressed self. Commit to the invalidation before you enter.
4. Re-running the pipeline mid-trade with new biases. You enter under one set of conditions, then under stress re-run the pipeline with a biased interpretation that justifies holding. The original interpretation should govern unless objective conditions change. Rationalizing as you go is the subtlest failure of all.
5. Running the pipeline without writing it down. Memory degrades, while a written record preserves the discipline. Even a brief journal entry for each run is enough.

11 · In practice

12 · Where this fits

Chapter 13D walks ten swing trades through this grammar, and Chapter 13E ten portfolio decisions. Chapter 13F compresses each strategy onto an operational card, and Chapter 13G takes the longest view: the portfolio's allocation, its kill criteria and the cadence of its review.

13 · Check yourself