A stop-loss on a single stock is on everyone’s lips. A few points below cost, and out comes the litany: discipline, cut, protect the principal. But what about total exposure? When the whole account slides from fully invested all the way down, who has ever set a stop for the account itself? A per-stock stop holds one position; it cannot hold a basket going down together. Everyone inside the account talks about stops — does the account itself have one?
One rule was written at the top of every column through 2018–2019: below 2638 on the Shanghai index, total position capped at 30% — watch more, trade less, hold only a minimal tracking position; once 2638 is decisively reclaimed, lean on the support, add on pullbacks, reopen the flexible allocation. Notice the direction: the deeper the fall, the lower the cap. The instinct of buying cheap goods points exactly the other way: the more it falls, the more one should be fully loaded. One clause caps, one clause reopens; the direction is written down in advance, not decided by the mood of the day.
What is the cap based on? Not on forecasting levels. On systemic risk. 2638 was the valuation floor where the market digested a full year of pessimism; below it there was no systemic rally to feed on, and no single stock’s independent run could last. Not one word of that judgment guesses tomorrow’s direction; it says one thing only: while systemic risk is in the room, whatever the holdings, the total must be capped. Quality governs what to buy; the cap governs how much — two rulers, and with either one missing the measurement fails.
Behind the rule stands a trade-off: give up uncertain potential gains, avoid irreversible permanent losses. What is given up can come back; what is permanently lost cannot — the cap carries that trade-off onto total exposure. Every capped day may miss a rebound, and a miss is uncertain; every uncapped day stands exposed to the fall, and that exposure is irreversible. That ledger runs one way.
First objection: the deeper it falls the cheaper it gets, so the more one should buy — a cap shuts the door on bargains. What the cap blocks is never cheapness; it is the systemic cheaper that follows the cheap. However low the valuation, it cannot stop every price from stepping down one more flight when everything sinks together. Cheapness decides whether a thing is worth looking at; the cap decides how much fits right now — two questions, and neither can answer for the other.
Second objection: a cap pinned to a level is seat-of-the-pants market timing, dead by tomorrow. This cap hangs on a falsifiable spot: breached, admit the error, stay capped; reclaimed, admit the call, reopen. A number written in advance does not predict tomorrow; it prescribes what to do when wrong. What expires is never the rule, only the forecasts that do not permit themselves to be wrong.
Third objection: per-stock stops are enough — if the portfolio falls, every stock is falling, so stop them one by one. In a systemic decline, stopping one by one is a queue at the butcher’s block: each position waits until it hurts before acting; after all the cutting, little cash remains and the tuition has been paid in full. A total-position cap cuts once. What it saves is not those few points but the difficulty of execution — one decision replacing twenty — and among twenty, one or two always land slow.
Why does nobody keep a stop on total exposure? A per-stock stop has a story to hold: the company stumbled, and the cut feels justified. A total-position stop has none — it cuts the whole account, not because a stock went bad but because everything did. A discipline with no story is the hardest to execute, which is why it must be written down in advance — a rule set beforehand recognizes no market mood, only the rule itself.
The hardest to replace are those who rode out 2018 fully invested: adding all the way down, until no bullets were left.
Most of the time, a cap looks like waste. In good years it presses down returns and makes no sound at all. It was never meant for ordinary days — the account’s airbag, opening only at the moment of impact.
