The Problem of Induction

Picture a boy named Rohan who crosses a small stream on his way to school. There is one flat stepping stone in the middle, and every single morning he jumps onto it, it holds, and he carries on. Day one, he lands carefully, a little nervous. Day thirty, he barely looks - he just leaps. Day one hundred, he is jumping with his school bag on one shoulder and a snack in his hand, completely sure of the stone, because it has never once let him down.

Now here is the question that this whole chapter is really about. After a hundred safe jumps, has Rohan proved that the stone is safe?

It feels like he has. A hundred times is a lot. Our minds treat "it has always worked" as the same thing as "it will always work." But look closer. Under the water, where Rohan cannot see, the mud around the stone has been slowly washing away, monsoon after monsoon. The stone is not fixed at all - it is loosening. Every safe jump made Rohan more confident while the stone got less safe. The hundred safe mornings were not building proof. They were building a false feeling of proof, right up to the morning the stone finally rolls under his foot and he goes into the water.

This is one of the deepest and most uncomfortable ideas in all of investing, and it has a plain name: the problem of induction. "Induction" is just the grown-up word for the way we reason from what has happened to what will happen - "the sun rose every day, so it will rise tomorrow," "this stone held every day, so it will hold today." Most of the time this reasoning is useful. But it hides a trap, and the trap is exactly the kind that empties people's savings. No pile of "it worked before" ever proves that a rule is true; a long record of good behaviour is a feeling of safety, not a fact of safety.

The lesson isn't that Rohan should never cross the stream. It's that a long streak of good days should make him check the stone, not stop checking it. This chapter is about learning to feel a small, useful shiver of doubt at exactly the moment everyone else is feeling most sure.

Here is the strange, lopsided heart of the idea, and it's worth going slowly, because once you see it you cannot un-see it.

Imagine you make a rule: "Every stone in this stream holds." How many safe jumps would it take to prove that rule completely true? The honest answer is: no number is ever enough. A thousand safe jumps, a million - you still have not proved it, because tomorrow's jump is not in your thousand. The very next stone could be the loose one. Confirming evidence piles up and up and never quite reaches "proved." It's like trying to fill a bucket that has no bottom.

But now flip it around. How many falls would it take to prove the rule false - to prove that "every stone holds" is simply wrong? Exactly one. A single wobble, one stone that gives way, and the rule is finished forever. You do not need a thousand falls. You need one.

Sit with how uneven that is. To build the rule up, no amount of good news is ever enough. To knock the rule down, one piece of bad news is plenty. Proving a rule true is almost impossible; proving it false is almost easy. This is not a small quirk - it is the shape of how the world actually works, and most people have it backwards. They think a long good record slowly turns "probably safe" into "proved safe." It never does. The record can only ever fail to have broken yet.

Why does this matter so much for money? Because nearly everything that goes badly wrong with savings wears the disguise of a long good record. The strategy that "has never had a losing month." The company that "has grown every single year." The fund that "has beaten the market for a decade." Each of these is a bucket with no bottom that people mistake for a full bucket. The clean record is doing something sneaky: it is quietly turning off the very caution you would need on the day the stone finally rolls. The better the record looks, the more relaxed you become - which is precisely when a hidden danger has the most room to hurt you. A track record cannot promise you the future. At best it tells you the past has not broken so far.

Before we go further, it's worth asking a fair question: if this trap is so simple, why do clever, careful, grown-up people fall into it again and again? The answer is that our minds were built to fall for it - and understanding why is half the cure.

For almost all of human history, "it happened before, so it'll happen again" was brilliant advice. The berries that were safe last year are safe this year. The path that led home yesterday leads home today. A brain that trusted patterns and stopped worrying once something had worked a few times saved enormous energy and usually survived. So evolution wired us to feel comforted by repetition. Every time something works, a little voice inside relaxes and says, "settled - stop checking." That instinct kept our ancestors alive, and most days it still serves us well.

The trouble is that money markets are not a berry bush. In a world where rare, violent events are possible - a crash, a default, a sudden freeze - the very instinct that says "it worked, relax" is aimed at the wrong target. It quietens your caution based on the calm days, which are exactly the days that tell you nothing about the storm. Worse, there's a second wiring problem: we notice and remember the many small green days vividly, one after another, while the single red day is invisible because it hasn't happened yet. Our minds fill up with evidence we can see (all those good days) and stay empty of the evidence we can't (the rare bad day that isn't in the picture). So we end up feeling certain about a story that is missing its most important chapter.

Put those two together - a brain that relaxes at repetition, and a brain that ignores what it can't yet see - and you have a creature almost perfectly designed to trust a clean record right up until it breaks. Knowing this doesn't switch the instinct off. But it lets you catch yourself. When you feel that warm "it's fine, it always works" feeling, you can now recognise it for what it is: an ancient reflex firing in a place it doesn't belong. That flash of recognition is the moment you get to override it and go looking for the red day on purpose.

Let's turn the idea into a picture you can carry in your head, because the whole thing rests on one simple, unfair balance.

On one side, imagine a long row of green ticks - every good day, every safe jump, every month the strategy made money. It can be as long as you like: thirty ticks, a hundred, a thousand. On the other side, a single red cross - the one day it breaks. The problem of induction says these two sides are not equal, even though the green side looks so much bigger. The thousand green ticks cannot add up to "proved safe." The one red cross can add up to "proved dangerous." The tiny red side outweighs the giant green side, because they are answering two different questions. Green ticks answer "has it worked?" Only the red cross answers "can it break?" - and that second question is the one your savings actually live or die by.

So the practical move is almost a mental flip. When someone shows you a beautiful run of green ticks, the useful reaction is not "wonderful, it's safe." The useful reaction is "interesting - now, what would the one red cross look like, and could I survive it?" The record tells you the machine has not broken. It cannot tell you the machine cannot break. Only by hunting for the hidden red cross do you learn anything that actually protects you.

Let's put real rupees on the table and watch the problem of induction do its quiet damage. illustrative

Meet Arjun, a careful, hard-working man who found what looked like a small money-making machine. Every month, using a stock-market strategy he'd read about, he could collect a steady little payment - he called it his "rent." In a calm month the strategy handed him about ₹9,000. Nothing dramatic, just a reliable trickle. He started with a pot of ₹5,00,000.

Month one: ₹9,000 in. Month two: ₹8,500. Month three: ₹9,200. Month after month, the green ticks lined up. By the end of the second year, Arjun had collected steady payments twenty-three months out of twenty-four - one month was flat, but he never had a real loss. His ledger was a wall of green. And exactly as the problem of induction predicts, each good month made him more sure and the strategy quietly more dangerous, because Arjun did the natural human thing: he grew his bet. If the machine works, why feed it only ₹5,00,000? He built the pot up to ₹9,00,000, reasoning - out loud, to his family - that the strategy had "never had a bad month, so it's basically safe."

Notice the exact sentence. Never had a bad month, so it's safe. That is the loose-stone reasoning, word for word. Twenty-three green ticks felt like proof. They were not proof; they were only the absence, so far, of the one red cross. And the strategy he was using had a nasty shape hidden inside it - it collected small sips of "rent" in calm weather but was silently exposed to a rare, violent market drop, the kind that comes once every few years without an invitation.

Then the market had one of its ugly mornings. A sudden global scare, prices gapping down before anyone could act. In that single stretch, Arjun's strategy didn't just give back a month of rent - it lost ₹6,20,000. Two years of patient green ticks, roughly ₹2,00,000 collected in total, wiped out several times over in a few days. His ₹9,00,000 pot became about ₹2,80,000. And here is the cruel part that the clean record hid from him: the more he had trusted the streak, the bigger his bet was on the day the streak finally lied. The record didn't just fail to protect him - it lured him into betting more right before the fall.

Arjun's mistake was not greed and it was not stupidity. It was trusting a bottomless bucket. He read a long good record as a promise, sized his money to that promise, and met the one event the record could never warn him about. A steady record can talk you into betting so big that the first real shock takes money you can never grow back - and there is no recovering from a wipeout.

The same trap wears a much more respectable suit when it walks into the world of companies, so let's watch it there too. illustrative

Meet Aayra, a thoughtful young investor who prided herself on doing homework. She found a lending company - the kind that borrows money cheaply and lends it out at a higher rate, pocketing the gap. What thrilled her was its record: for eight years in a row, its profits had grown, smoothly, like a staircase. No down years. No nasty surprises. Every annual report was a fresh green tick. To Aayra, eight straight years of growth felt like the closest thing to proof a company could offer. She put in ₹3,00,000, a big slice of her savings, and told herself she'd found something rare - a business that simply does not have bad years.

But think about what a lending company actually is, and where its spider hides. Its smooth, growing profits came partly from lending more and more money each year, borrowed from others. The very smoothness Aayra loved was the symptom of the danger: to keep the staircase rising, the company had to keep borrowing and lending faster, and the loans it was handing out were getting riskier in ways that would only show up if the economy stumbled. The clean eight-year record wasn't proof of strength. It was the calm surface over rising water - the mud washing away under the stone.

In year nine, the economy wobbled. Some of the company's borrowers couldn't pay back. Because the company itself had borrowed heavily to grow, even a modest wave of bad loans hit it hard, and the trust that let it keep borrowing evaporated almost overnight. The share price fell about 70%. Aayra's ₹3,00,000 became roughly ₹90,000. And the arithmetic is unforgiving: to climb back to ₹3,00,000, that ₹90,000 would need to more than triple. The eight green ticks had bought her exactly nothing except a larger, more confident bet on the year the record broke.

Here's the sentence to remember from Aayra's story: the unbroken record was not evidence the danger was gone - it was the thing hiding the danger. A rising staircase of profits can mean a business is genuinely wonderful, or it can mean a business is taking a growing hidden risk that simply hasn't been called yet. The record alone cannot tell you which. You only learn the difference by putting the good record down and asking the harder question - what is this smoothness costing, and what would one bad year do? A rare bad year missing from a company's record is not proof it can't happen; the calm run is exactly what stops you asking what a single shock would do.

Now let's go one level deeper, into why these clean records are so good at fooling careful, intelligent people. It comes down to the shape of the wins and losses. illustrative

Picture Aman keeping a simple ledger of a strategy over four years - one line per month, green for a gain, red for a loss. When you look at it, almost every line is green, and each green number is small and friendly: +₹5,000, +₹6,500, +₹5,200, on and on. Forty-five green months in a row. The strategy feels like a gentle, dependable friend. If you drew its progress as a line, it would climb smoothly, month after month, a picture of calm reliability. Anyone glancing at it would say, plainly, "this works."

But hidden inside the type of strategy is one enormous red month waiting somewhere in the future - not scheduled, not predictable, just lurking. When it finally lands, it is not a small friendly red like the greens were small friendly greens. It is -₹4,10,000 in a single month. Add up the whole ledger honestly: forty-five green months of about +₹5,800 each is around +₹2,60,000 collected over nearly four years. One red month of -₹4,10,000. The strategy that "works" has, across its whole life, lost about ₹1,50,000. The smooth climbing line ended in a cliff, and the cliff was taller than the entire climb.

This shape is the reason smart people get fooled. Our eyes and our feelings judge a strategy by the number of green months - and here the green months are overwhelming. But rupees don't care how many months were green; they care how big each month was. A hundred small wins and one giant loss can add up to a loss, while looking, right up to the final month, like a triumph. The clean record isn't just failing to prove safety - it is actively painting a dangerous thing to look like a safe one. Judge a strategy by what one rare, oversized loss would do to it, not by how long the calm run of small gains has lasted. The moment to ask "where's the cliff?" is exactly when the climb looks smoothest, because a smooth climb is what a cliff-shaped strategy looks like right before the edge.

The slip almost never sounds reckless. It sounds reasonable, even wise. It comes out as a calm, confident sentence: "That has never happened before." A bank that has never missed a payment. A neighbourhood scheme that has paid out every month for years. A price that has never fallen this far. Each time, "it has never happened" is offered as if it were a reason for safety, when it is really just a description of a short, lucky history.

Here's the mechanism, because it's worth seeing clearly. The danger and the comfort move in opposite directions. As a good record grows longer, your comfort grows - but often the hidden risk grows too, because a long calm lets people borrow more, bet bigger, and lower their guard. So you end up most relaxed at the exact moment you are most exposed. The feeling that should be flashing a warning - "this has been calm for a very long time" - is instead read as an all-clear. That inversion is the whole trap: the record turns off your alarm precisely when the fire risk is highest.

So what do you do with all this? It would be easy to walk away either paralysed ("trust nothing!") or unchanged ("interesting, anyway that fund has a great record..."). The useful path is a small set of plain habits that turn the idea into action.

First, rename every record in your head. When you read "never had a losing year," quietly translate it to "hasn't had a losing year yet." When you hear "this has always paid out," translate to "this has paid out for the length of time I happened to watch." The words feel almost the same, but the second version keeps your alarm switched on. Aayra didn't need a genius insight to avoid her lending company - she only needed to hear "eight years of growth" and reflexively add the word yet.

Second, ask the red-day question before you commit, not after. Before any rupee goes in, force yourself to finish this sentence out loud: "The rare bad event that would hurt this is ______, and if it happened tomorrow, I would lose ______." If you cannot even name the bad event, that is not a sign of safety - it's a sign you haven't looked hard enough. Arjun could have saved himself simply by asking, before growing his pot to ₹9,00,000: "What happens to this strategy on one terrible market morning?" The honest answer would have stopped him cold.

Third, and most important, size every bet so a single break can't finish you. This is the habit that makes all the others safe, because it means you don't have to predict the storm - you just have to survive it whenever it comes. If Arjun had promised himself that no single bad event could ever cost him more than, say, ₹1,00,000, the shape of his whole story changes: the rare shock becomes a painful lesson instead of a wipeout, and he lives to invest another day. The clean record fooled him into growing his bet; a firm rule about survival would have held his bet still. Decide your worst survivable loss first and size to it, because a bet that can be ruined by one break is a bet you have already half-lost.

None of this requires you to predict anything. That's the quiet beauty of it. You are not trying to forecast the crash, spot the loose stone, or time the storm. You are simply refusing to let any good record grow your bet past the point where a single surprise ends the game. You stay humble about the future and you stay in the game - which, it turns out, is most of what winning slowly is really about.

Now the honest part, because the problem of induction is powerful enough to be misused, and a class-5 mind that grabs it too hard can end up in a different ditch.

The first way it misleads: it does not mean "distrust all records and never invest in anything." That would be its own kind of foolishness. If Rohan refuses to cross any stream ever again because no stone can be proved safe, he simply never gets to school - he has swapped one loss for another. Records are still genuinely useful; they just do a different job than people think. A good record cannot bless an idea as safe, but a broken record can still disqualify one as dangerous - and that's real, usable information. Use track records to throw out the clearly-broken things, not to hand a lifetime guarantee to the clearly-unbroken ones. The point is to stop over-trusting good records, not to start ignoring all evidence.

The second way it misleads: the goal was never to be frozen by doubt. Someone who becomes so haunted by the one rare disaster that they keep every rupee in cash forever has not escaped the problem - they've just chosen a slow, certain loss (their savings quietly eaten by rising prices) over a rare, uncertain one. The cure for the induction trap is not paralysis. It is sizing: invest, but never bet so much on any one streak that a single break can wipe you out. You want to stay in the game through the ordinary weather and survive the rare storm - not sit out the weather entirely. Doubt the record enough to keep your bets survivable; don't doubt it so much that you never make one.

And a third, quieter caution: knowing that rare events exist does not tell you when one will strike. This idea makes you humble, not psychic. It will never let you say "the crash comes in March." What it lets you say is "because I can't know when, I will keep every bet small enough to survive whenever it comes." The problem of induction is not a crystal ball. It's a reason to build a boat that floats in any weather, precisely because you can't read the sky.