The rule you wrote after the one time it happened

In January 2018 a colleague mentions a name over lunch. By the end of the week ₹3,00,000 has gone into it, out of a portfolio then worth ₹13,50,000, and ₹1,00,000 of that ₹3,00,000 is drawn on a credit line because the salary had not come in. Over the following fourteen months it falls 68 per cent. In April, with the position already halved and no floor in sight, one line goes into a notes app: never buy small-caps. Eight years later that line is still there and still obeyed. It has never been reread with any care, because it does not need to be — it is short, it is unambiguous, and it was earned. It is also aimed at the wrong thing, and it has been aimed at the wrong thing every day since.

Think of it like this

The road you never take again

A driver skids badly on a wet flyover on the way to the airport and afterwards avoids that flyover for years, taking a longer route on dry days and wet ones alike. The flyover was not the cause. The speed on a wet surface was. Avoiding the road feels like having learnt something, and it leaves the driver equally exposed on every other wet road in the city.

In the market

A rule written after a loss almost always names the place rather than the cause, because the place is what the memory is made of. It then protects you from that place and from nothing else, at the cost of a route you now never take.

What the rule names, and what actually did the damage

Take the 2018 event apart. The loss was ₹2,04,000 — 68 per cent of ₹3,00,000 — which is 15.1 per cent of the whole portfolio in one position. Now list what produced a figure that size, and check each item against the rule that was written.

What was actually true on the day	How much it contributed	Does the rule address it?
The position was 22 per cent of the portfolio	This is what converts a bad holding into a bad year. The same fall in a 5 per cent position is a 3.4 per cent portfolio loss	No. The rule says nothing about size
₹1,00,000 of it was borrowed	Puts a deadline on a drawdown, so the decision about when to sell stops being yours	No
It was bought on a tip, with no thesis	With no stated reason for owning it there was no condition that could tell you it had failed, so the holding could only be ended by pain	No
The company was a small-cap	Widened the range of outcomes. Real, and the least of the four	Yes — and only this one

Worked example

The rule that was written against the rule that was not

The same portfolio of ₹13,50,000, the same three behaviours, tested against different rules

What actually happened15.1 per cent of the portfolio, of which part was borrowed₹3,00,000 position, down 68% — a loss of ₹2,04,000

With the rule that was written, applied to a large-cap insteadIndividual large-caps fall 45 per cent in bad years too. The loss is ₹1,35,000Same 22% position, same borrowing, a 45% fall

That loss as a share of the portfolioPlus interest on the ₹1,00,000 for as long as it was carried. The rule reduces the damage and does not prevent it10.0 per cent

With a rule that was not written: no position above 8% of the portfolioA loss of ₹73,440₹1,08,000 in the same small-cap, down 68%

That loss as a share of the portfolioAnd with the position that size there is no reason to borrow, so the drawdown carries no deadline5.4 per cent

What each rule costs when it is not neededThe size cap constrains every purchase slightly. The category ban removes an entire category from consideration for eight yearsOne excludes an asset class permanently; the other caps a position

The rule that was written turns a 15.1 per cent hit into a 10 per cent hit, and only if the next mistake happens to be in a different category. The rule that was not written turns it into 5.4 per cent regardless of category, and would have applied to every position taken since. The household has spent eight years enforcing the weaker of the two, believing it had learnt the lesson — and it has never compared them, because the comparison requires writing down what the loss was made of, which is precisely the exercise nobody wants to do in April with the position still falling.

Why a rule like this is never re-examined

It appears to work. The event has not recurred in eight years. But it was never going to recur annually — the [[base rate]] of a 68 per cent collapse in a position you hold is nothing like once a year, so eight quiet years are exactly what you would expect with the rule and without it.
Its cost leaves no record. Every purchase the rule blocked is absent from the account, so the ledger shows all of the rule’s benefits and none of its price. This is the [[silent evidence]] problem applied to your own constraints.
It carries the authority of pain. A rule written after a loss feels earned in a way that a rule written on a calm Sunday never does, and reopening it feels like inviting the loss back rather than like ordinary maintenance.
It is one sentence long. The reasoning was never attached, so there is nothing to disagree with. A rule you cannot argue with is not a strong rule; it is an unfalsifiable one.
It was fitted to a [[sample size]] of one. Tuning a rule until it explains a single observation is [[overfitting]] in the ordinary sense — it fits that afternoon perfectly and generalises to nothing.

The comparison nobody makes

A rule is a trade. It prevents some harm and it blocks some things that would have been fine, and it is worth keeping when the first exceeds the second. Both sides can be estimated roughly, which is enough, and only one side is ever estimated at all.

Keep the rule while A > B, where A = (how often the harm happens × what it costs each time) and B = (how often the rule blocks something harmless × what each block costs)

A: What the rule saves you. Vivid, remembered, and the only side anybody ever estimates
B: What the rule costs you. Every wrong firing leaves no trace anywhere, so this side is not merely unknown — it feels like zero

Example: Here: the harm has occurred once in eight years. The rule fires on every small-cap that would otherwise have been considered — perhaps ten a year, so roughly eighty blocked considerations per real avoidance. Whether eighty refusals cost more than what the rule actually saved is a genuine question with a genuine answer, and it turns on how many of those eighty you would have sized sensibly — noting that what it saved is not the whole 15.1 per cent, but the five points between that and the 10 per cent the same behaviour cost in a large-cap. The point is that it has never been asked, and a rule whose cost is invisible will always feel free.

Three lines that turn a reflex into a rule

Written when the rule is made, or added to a rule you already have

1
The mechanism, stated so it would still make sense if the original event had never happened
"Never buy small-caps" fails this test — it is only intelligible with the 2018 story attached. "No single position above 8 per cent, because one holding must not be able to cost me a year" passes, and would have been written the same way by somebody who had never lost anything.
2
What it costs when it fires wrongly
One line, in rupees or in plain words. "This will keep me out of some good businesses I cannot size properly" is a complete entry. Writing it down is what converts the cost from invisible to merely accepted, which is all that is needed.
3
A date, and the evidence that would retire it
Not a market level and not a feeling — a date in the calendar, and a sentence describing what you would have to see to conclude the rule is no longer earning its keep. A rule with no expiry and no falsifier is a superstition with a good origin story.

◆ Your call

It is eight years later and the rule is still in the notes app

Your portfolio is now ₹68,00,000 rather than ₹13,50,000. You have a written process, you size positions deliberately, and you have not used borrowed money since. The line "never buy small-caps" is still there and you have never broken it.

Check yourself

Why is "the rule has not been broken in eight years and nothing bad has happened" such weak evidence that the rule is working?

◆ Checkpoint

Module checkpoint: the decisions nobody took

4 questions. Answers are revealed once you submit all of them.

1.A household holds ₹6,80,000. Its emergency plan needs ₹6,80,000 and its down-payment plan needs ₹6,80,000. Why can no review of either plan detect the problem?

2.A 22-year projection of ₹40,000 a month reads ₹5,13,22,785 at 12 per cent and ₹2,86,71,513 at 8 per cent. What should the household take from that?

3.Why does ordinary scepticism fail against a claim that has been forwarded four times?

4.A rule written after a single large loss has been obeyed for eight years without incident. What is the strongest reason to reopen it?

0 of 4 answered

Simple bhasha mein

Wahi flyover, jispe ek baar phisle the

2018 mein colleague ki baat pe ₹3,00,000 ek small-cap mein daal diye — portfolio tha ₹13,50,000, yaani 22% ek hi naam mein — aur usme ₹1,00,000 udhaar ka. Stock 68% gira: nuksaan ₹2,04,000, poore portfolio ka 15.1%. April mein ek line likhi — "kabhi small-cap nahi" — aur aath saal se woh line chal rahi hai. Par nuksaan bana kis cheez se tha? Size, udhaar, aur bina wajah ke tip pe khareedna. Rule in teeno mein se ek ko bhi nahi chhoota. Hisaab dekho: wahi 22% aur wahi udhaar agar ek largecap mein hota jo bure saal mein 45% gira, toh nuksaan ₹1,35,000 — 10%, yaani bacha nahi. Aur agar rule yeh hota ki "ek position 8% se zyada nahi", toh usi small-cap mein ₹1,08,000 jaate, nuksaan ₹73,440 — 5.4%, aur itni chhoti position ke liye udhaar ki zaroorat hi nahi padti. Yaad ghatna ki jagah rehti hai, wajah nahi. Aur rule kabhi dobara nahi padha jaata, kyunki jo usne roka woh kahin likha hi nahi jaata — isliye har rule muft lagta hai. Teen line likho: kis cheez se bacha raha hai (bina purani kahani ke), galat chalne pe kya kharcha hai, aur agli baar kab padhoge — tareekh pe, mood pe nahi.

What to remember

A rule written after one event encodes the surface features of that event, because that is what memory stores — the category, not the cause.
Take the loss apart into what it was made of, and check the rule against each part. Usually it addresses the least important one.
A rule’s benefits are vivid and its costs leave no record, so every rule feels free and none of them is.
Give every rule three lines: the mechanism stated without the story, what a wrong firing costs, and a review date with a falsifier.
Review on a date, never on a feeling — a rule that is right rarely will feel wrong almost always.

Finished this lesson?

Mark it done to track your progress through the curriculum.