1. Writing the limits before you need them

Risk 9: A risk system of your own

The previous eight levels have assembled the complete toolkit: measuring, sizing, surviving the drawdown, understanding leverage, building the portfolio, knowing the limits of the tools, seeing the risks that never show up on the chart and ordering money through time. What is missing is the piece that turns all of it into something that works on an ordinary Tuesday: a written system of your own. It starts with the dullest and most profitable thing in the course: writing the limits before you need them.

The same limit decided at two different moments

A limit is a short sentence. No position will go above 8% of the portfolio. Written on a quiet Sunday, it is drafted by somebody who feels nothing. Written on the third day of a 25% fall, it is drafted by a different person: the same head with a different system in charge.

Daniel Kahneman described the mechanism. Faced with a loss in progress, judgment stops leaning on slow reasoning and gets governed by the fast reaction, the one that answers to pain rather than to arithmetic. Add loss aversion on top: losing a sum hurts roughly twice as much as gaining it pleases. The result is not a slightly worse decision-maker but one skewed always in the same direction, and precisely when that costs most.

There is a second effect, more treacherous still: while you are calm you cannot predict what you will feel in the fall. Almost everybody, serene, believes themselves capable of riding out a 30% decline; a minority manages it. That inability to imagine your own future state is why the mental promise is useless and the written sheet is not.

Annie Duke puts it in decision theory terms: the moment to decide what you will do when something happens is before it happens, because afterwards you no longer decide, you react. Hence her insistence on setting exit criteria in cold blood.

What a limit is and what it is not

A limit is not a forecast. It does not claim the market is going to fall, or when. It is a rule about you: it describes how far your capacity to bear an outcome goes without taking decisions you would not defend later. That is why two reasonable investors can have different limits without either being wrong; what cannot differ is whether the sheet exists.

The four worth having in writing

• Maximum exposure per position and in total. Two figures, not one. The first caps what it costs to be wrong about one idea: with 8% per position, the worst imaginable thesis, the one that loses half, costs 4% of the portfolio. The second caps the common bet: if your total limit per source of risk (a sector, a currency, an interest rate) is 40%, you have already decided that no story, however good it sounds, gets to become the whole portfolio. Level 7 explained why hardly anybody has that second figure.
• The maximum loss you are prepared to bear before revisiting anything. The most uncomfortable to write and the most useful. It is not a sell trigger: it is the threshold below which you oblige yourself to reread the thesis in writing. For example: at 20% from the highs, a documented review becomes compulsory; at 35%, review and reduction of the leveraged part to zero.
• Maximum leverage. Gross, not net, and with the small print from level 4: the figure that matters is not today's, but the one your portfolio will have when the asset falls and the debt stays intact. Write it down as a sentence about margin too: what fall you can take before the first call.
• Minimum liquidity. In months of spending and in euros, and outside the portfolio. It is the only line on the sheet that decides whether a fall is a bad year or a final outcome, and level 8 showed that the market does not draw that distinction, you do.

The exact number matters less than you think

Here is the counterintuitive part. None of those four numbers is optimal, and there is no way of calculating one: they depend on an unknown future distribution and on a tolerance of yours that can only be measured under conditions that have not yet occurred. 8% per position and 10% produce almost identical portfolios.

What does change the outcome is that the number exists and came first. Howard Marks has spent decades repeating the same idea: risk control does not consist of getting the level right, but of not letting the market choose your size for you. Antti Ilmanen adds the economic reason: if risk premiums are the payment for bearing pain in the bad moments, whoever cuts at exactly the worst moment pays the full price and never collects the reward.

How to write it so that it holds

• With a date, the only proof that the limit predates the fright.
• Observable, not open to interpretation: "no 40% of the portfolio in one sector" can be checked; "I will not overexpose myself" can never be checked.
• Saying what happens when it is crossed, not just where the threshold sits. A limit with no written consequence is a wish.
• With a rule for changing the rules: every amendment is written down, dated and takes effect thirty days later. That does not stop you improving the system, it stops you improving it on the Tuesday that hurts.

A worked scenario: the same 200,000 euros

Two investors with 200,000 euros and the same portfolio of global equities. The first has a sheet dated in January: maximum 8% per position, compulsory review at a 20% fall, no leverage and 18,000 euros of liquidity outside the portfolio. The second has the same intentions, but only in his head.

February 2020 arrives. Between February 19 and March 23 the S&P 500 falls around 34% in a little over a month, and the volatility index closes above 82 on March 16, its highest close on record. Both portfolios lose something of the order of 60,000 euros.

The first crosses his threshold, does what the sheet says (reread the thesis and write it out) and finds no reason to sell: his liquidity covers the year, there is no margin involved and nothing forces him to decide today. The second, with no sheet, decides at the worst possible moment with the worst version of his own head. The index recovered its February high in August. The difference was not in the analysis: it was in a piece of paper written five weeks earlier.