What the Whole Book Risks

Heat is the only portfolio-level risk number

Every lesson so far has sized one decision. A portfolio needs a number that describes all of them at once, and the natural candidate — total dollars invested — is the wrong one, because it says nothing about where the exits are. Two books of identical value can have wildly different risk depending on how far their stops sit, which is why the useful measure is **heat**: the total the account loses if every stop is hit. Heat is computed by adding each position size multiplied by its own stop distance, which is the same arithmetic as R5 repeated across the book. It has three properties worth naming. It is measured on **stop distances** rather than on notional, so a large position with a close stop contributes little. It is **additive**, which is what makes a cap possible. And it makes the implicit assumption that stops are hit **together**, which in a bad market is the realistic case rather than a conservative one: positions in a correlated book are all exposed to the same shock, and a day that takes one stop takes several. The cap itself is a number the household sets and then obeys: commonly 4 to 6% of capital across the book, with a per-position limit inside it. The two work together. The per-position limit stops a single idea from being large; the heat cap stops the total from being large. A book can pass the first and fail the second — which is exactly what the four positions above do — which is why both have to be checked rather than either. The book, position by position — A: $50,000 at 8%: $4,000 risk · B: $62,500 at 5%: $3,125 risk · C: $37,500 at 12%: $4,500 risk · D: $25,000 at 6%: $1,500 risk · Total: 5.25% of capital: $13,125 against a 6% cap of $15,000 ← The largest position is not the largest risk: C carries $4,500 on $37,500 while A carries $4,000 on $50,000, because a 12% stop is a bigger slice than an 8% one. Position size and risk are separate questions.

What a heat cap changes about decisions

The practical effect of computing heat is that it converts a series of independent decisions into one budget, and budgets run out. When the book is at 5.25% against a 6% cap, the next idea has $1,875 to work with, which at an 8% stop is a position of about $23,000 — smaller than the household might have wanted and perfectly sized for the risk it is allowed to add. The alternative, taking the position at the size the idea deserves and letting heat drift to 8%, is how a portfolio becomes a single bet: the cap is what prevents the sum of several reasonable decisions from being an unreasonable one. It also gives the correct answer to a question that comes up constantly, which is what to do with a strong new idea when the book is full. There are three honest options and none of them is to exceed the cap. One is to make room by closing or trimming an existing position, which forces a comparison between the new idea and the old one on the same risk basis rather than on enthusiasm. Another is to take the new position at the size the remaining budget allows, accepting a smaller bet. The third is to pass, which for a full book is frequently right and is the option people find hardest because the idea feels free. A cap also has to state what happens when it is breached, because breaches happen through price movement rather than through new trades: a stop can be moved, a position can grow with the market, and overnight gaps can change the distances. The rule that works is the same shape as the drawdown policy in R17 — a defined action rather than a judgement, such as trimming to the cap by the next session — because a limit whose response is decided in the moment is not a limit. Room under the cap — Current heat: $13,125 — 5.25% of $250,000 · Room under a 6% cap: $1,875 · A fifth position at an 8% stop: about $23,000 of stock · The same idea at 8% heat instead: $20,000 of risk across the book ← A heat cap is measured against the current account value, so it shrinks after a drawdown on its own — and it must be recomputed rather than remembered, because every price move changes the stop distances.

Heat assumes the stops do not all fire at once

Heat is a good portfolio number and it rests on an assumption that is false in precisely the states where the number is used. It adds up the distance from each entry to its stop, which is the loss if that position alone is stopped out. In ordinary markets, that is close enough: positions are stopped out for their own reasons, on their own days, and the sum is a reasonable description of a bad week. In a genuine market event, the stops fire together, and the sum becomes the actual loss rather than an upper bound — with the additional problem that the fills are worse than the stops, since a gap means the exit happens below the level the heat calculation used. The gap complication is not a rounding error and it cannot be fixed by adding a fudge factor to the cap. Heat is computed from stop levels, and a gap through a stop means the realised loss was never available at the price the heat model assumed. So the honest version of a heat cap is a two-part rule: a limit on the sum of planned risks, and a separate stress test that replaces every stop with a plausible gap-through price and asks whether the resulting loss is survivable. That is the same logic as the gap lesson, applied at the level of the whole book rather than one position, and it is where a heat cap that looks comfortable can turn out to be thin. Correlation is the bridge to the sizing rules rather than an alternative to them. If the positions in the book are genuinely driven by different things, heat is a decent estimate of a bad day and the cap can be generous. If they are the same trade in different wrappers — five semiconductor names, three rate-sensitive REITs — then the heat number is not an upper bound at all, it is a forecast of every position being stopped on the same morning. Counting drivers, which the correlation lesson does, is what tells you which of those two books you are running; the heat cap tells you how much room is left. Both are required, and neither substitutes for the other. • Heat is the sum of planned risks, and it is an upper bound only while stops fire independently. • In a real event the positions are stopped together, and gaps make the fills worse than the levels used. • Pair the heat cap with a gap stress test that replaces every stop with a plausible gap price. • With correlated positions the heat figure is a forecast rather than a bound — check the drivers first.

The time dimension: the cohort you did not notice you built

Heat is computed on the current book, which means it is a statement about how much is at risk *now* rather than about how the book got there. That omission hides the most common way a heat cap is breached without a single rule being broken: positions added at different times in the same direction, each one passing the per-trade limit and each one staying under the portfolio cap at the moment it was opened, until the accumulated exposure is one trade wearing several names. The mechanism is that heat constrains the sum, and the sum is built incrementally — so the constraint that matters in practice is not only the cap but the **rate of addition**. Three entries in three days in the same driver produce the same risk as one entry three times the size, and only the second is visible to a rule that computes heat at the moment of entry. The instrument for that is a **cohort rule**: state how much new risk may be added in a given window, and count positions opened since the same signal as one. The reason it earns its place is the mechanics of correlated entry rather than a suspicion about discipline. Positions opened from the same setup in the same market on the same day share not only their driver but their timing, so their stops sit at structurally similar distances and their entries at similar prices; a shock therefore hits them simultaneously, which is the assumption the gap stress test already makes and which the cohort rule enforces in advance. A practical form is a limit on the number of new positions per week, or a requirement that a new position in an existing driver be half size, or a rule that total heat may not increase by more than a stated amount in any two sessions. All three are the same idea: heat is a stock, and the stock is built by a flow that has to be limited too. There is a second, subtler consequence of treating the book as a cohort. Positions held for a long time accumulate **unrealised profit**, and the conventional heat calculation uses the distance from the *entry* to the stop. Once a position has moved and its stop has been trailed, the risk it carries is no longer measured from the entry at all — it is measured from the current price to the stop, and the difference can be the whole of the position’s unrealised gain. A book whose heat is computed on original entry distances will overstate the risk of positions that have worked and understate the risk of positions that were added near their stops. The repair is to recompute heat on the current distance, and to add the unrealised profit to the risk estimate when the position is large enough that giving it back would itself be a loss the plan does not tolerate — which is the general principle this lesson is really about: heat is a measure of what can be lost from here, and “from here” moves every day. • Heat is a stock built by a flow — cap the rate of addition, not only the total. • A cohort rule counts positions opened on the same signal in the same window as one position. • Correlated entries share their timing and their stop distances, so a shock reaches them together. • Recompute heat on the distance from the current price to the stop, not from the original entry. • Where a large unrealised gain would be intolerable to give back, include it in the risk estimate. The pairing with the drawdown policy is exact and worth writing into a single page: the drawdown trigger decides what happens after the losses, and the rate-of-addition rule decides how much can be lost before it fires. A policy that constrains only the state and not the flow is a rule that reacts rather than a rule that prevents.

Sources

• Portfolio heat and total open risk limitsVan Tharp, "Trade Your Way to Financial Freedom"
• Correlated stops being hit together in a stress eventStandard portfolio-risk practice; 2008 and March 2020 evidence
• Exposure limits as a pre-committed rule rather than a judgementRisk-policy literature; CFA Institute readings on risk governance
• Why a full book should refuse trades rather than trim themElder, "Come Into My Trading Room"