A Mathematical Meditation on History

Imagine a giant machine in a hall, taller than your house, full of spinning wheels and slots. Inside it are printed thousands and thousands of different versions of this exact year - the one you are living right now. In most of those versions the year is fairly ordinary: some good days, some dull days, prices drift up a little or down a little. In a few of them something wonderful happens. And in a handful of them something horrible happens - a crash, a flood, a company you trusted quietly falling apart. All of these versions are printed and waiting inside the machine.

Now the machine does one thing, and only one thing. It reaches in, pulls out a single version, and hands it to the world. That version becomes what actually happens. The rest - the thousands of years that were just as possible, that almost got picked - vanish. Nobody ever sees them. We live inside the one slot the machine happened to land on, and then, being human, we stand there staring at that one slip of paper and say, "So this is how the world works."

That is the mistake this whole chapter is about, and once you see it you cannot unsee it. The past that actually happened is not the truth about how things must go. It is just one draw out of a huge box of draws that could have come out instead. The record we have - the price chart, the story of who got rich and who got wiped out - is a single sample. And a single sample can fool you badly, because it looks so solid, so final, so obviously the way it was.

Here is the trap in one line. When you learn a lesson from what happened, you are quietly assuming that what happened was bound to happen - that if the machine ran again it would hand you the same slip. But it wouldn't. Run this year again from the start and you'd very likely get a different slip, sometimes wildly different. So the honest question is never just "what happened?" It's "what could have happened, and how many of those other versions would have gone badly for me?" Judge a choice by all the versions of the story it could have landed you in, not by the single version that happened to come out of the machine.

You might think, fair enough, but surely the one year we did get still teaches us something? It does - a little. The problem is that we treat it as teaching us almost everything, and that is where the damage begins.

Think about how a class-5 mind, or a grown-up mind, actually learns. You do a thing, you see what happens, and you decide the thing was smart or silly based on the result. Touch a hot pan, it hurts, you learn "don't touch hot pans." That works beautifully in a world where the same cause always brings the same result. Hot pans are honest teachers; they burn you every single time.

But money is not a hot pan. Money lives in a world full of dice. In a dice-world, the very same choice can bring a lovely result one time and a ruinous result the next, and you often can't tell from the result alone whether the choice was wise. A reckless bet can win. A careful, sensible bet can lose. If you learn the ordinary way - "it worked, so it was clever; it failed, so it was foolish" - the dice will teach you exactly the wrong lessons. You will end up admiring the lucky gambler and scolding the careful planner, because you can only see the one slip the machine handed out, and on that slip the gambler happened to win.

This matters enormously for your money, because the loudest, most confident-sounding advice in the world usually comes from someone waving a single lucky slip. "I did this and I doubled my money - so you should too." What you are being shown is one draw from the machine. What you are not being shown are all the other versions of that person's year, the ones where the exact same move left them poorer. And you cannot judge their choice honestly until you imagine those missing versions - the alternative histories that quietly didn't happen but very nearly did. The single true story you can see is forever bullying you into ignoring the thousand invisible stories that would have taught you the opposite.

Let's slow this down and look at the shape of the idea, because a picture makes it click.

Picture a single dot on the left - that's today, the starting point, the same for everybody. Now, from that dot, the future doesn't run forward as one straight line. It runs forward as a huge fan of lines, spreading out like the sticks of an open hand-fan or the branches of a tree. Each line is one possible version of the year: this line is the calm one, that line is the boom, this thin one down at the bottom is the crash. All of them start from the same today. All of them are genuinely possible.

Then the year is lived. The machine picks. And exactly one of those fan-lines lights up bold and real - the path that actually happened. Every other line in the fan goes grey and fades, as if it never existed. When we look back, we only ever see the one bold line. That is why the past feels so certain and single: the fan has collapsed into one path, and the mind kindly deletes the memory that a fan was ever there at all.

Hold that fan in your head, because it is the master key. Everything else in this chapter is just what goes wrong when we forget the fan and stare only at the one bold line. A wise investor does the strange, effortful thing of keeping the whole fan visible - refusing to let it collapse - so that when they judge a decision, they judge it against the full spread of years it might have produced, not the single year it happened to produce this time.

Let's put real rupees down and watch the fan fool a whole group of people at once. illustrative

Two friends, Rohan and Aman, each start the year with ₹5,00,000 saved up. They take completely different approaches.

Rohan hears a thrilling story about one small company and puts his entire ₹5,00,000 into that single stock. No spreading out, no cushion - everything on one square. Aman does the dull thing: he spreads his ₹5,00,000 across a broad, boring basket of many companies through a simple monthly SIP, so that no single company can hurt him much.

The year is lived. The machine picks its slip. And on this slip, Rohan's one company soars - his ₹5,00,000 becomes ₹10,00,000. Aman's boring basket drifts up gently to about ₹5,50,000. Now watch what happens in the group chat. Rohan is a hero. People ask him for tips. He starts to believe his own legend - clearly he has the eye, a gift for picking winners. Aman, meanwhile, feels a bit foolish and slow, up a mere ₹50,000 while his friend doubled.

But we have the machine. Let's do the thing nobody does: let's run Rohan's exact same choice across the other slips it could have drawn. Rohan bet everything on one small company. In the fan of possible years, a small company can soar - but it can just as easily halve on a bad quarter, or fall 80% on a scandal, or quietly go nowhere. If we honestly picture, say, the ten most likely versions of Rohan's year, in only two or three does he end up richer, in several he's flat or bruised, and in a couple he is seriously hurt - down to ₹1,50,000 with no cushion to recover. Rohan didn't make a good bet that happened to win. He made a dangerous bet that happened to win. Those are not the same thing, even though the one visible slip makes them look identical.

Aman's dull choice, run across the same fan, almost never doubles - but it almost never craters either. In nearly every version of the year he ends up somewhere between "a little down" and "nicely up," and in none of them is he wiped out. His choice was sound across the whole fan; Rohan's was reckless across the whole fan. The single result we can see rewarded the reckless one and made the sound one look silly. A win doesn't make a bet wise: to grade the choice honestly you have to replay it across the many years that could have happened, and ask how you'd have done in most of them, not the lucky one.

The lesson isn't "Aman is right and Rohan is wrong because of how the year turned out" - that would just be the same outcome-worship in reverse. The lesson is that the result cannot tell you who chose well. To know that, you have to leave the one slip behind and imagine the whole box.

There's a nastier way the single sample fools you, and it's worth its own worked example, because it's the one that actually ruins people. illustrative

Meet Arjun. Arjun has ₹1,00,000 of his own, but he's impatient, so he borrows another ₹3,00,000 and puts the whole ₹4,00,000 into the market. His own money is only a quarter of the pot; the rest is the lender's, and - this is the cruel part - if the pot shrinks, the loss eats his quarter first, because the lender must be paid back in full no matter what.

For three years, the machine hands Arjun calm, friendly slips. The market drifts up a gentle 10% each year. On ₹4,00,000 that's a tidy gain, and after paying the small interest on his loan, Arjun is pocketing far more than boring Aman ever could on ₹1,00,000. Three years running. Arjun is delighted, and more than delighted - he's convinced. He looks back at his three-year record, sees nothing but green, and concludes that borrowing to invest is simply the clever thing to do. The past, he thinks, has proven it safe. Three years of evidence!

Here is the whole problem in one breath: three calm years is not proof the storm is gone. It's just three slips where the storm didn't come out of the machine. The fan of possible years always contained a crash-line - a bad month where the market falls 30%. Arjun never saw it, so he began to believe it didn't exist. But absence from his little record was never absence from the box. And in year four, the machine hands him the crash slip. The market falls 30%. On ₹4,00,000 that's a ₹1,20,000 loss - and since the loss eats his quarter first, Arjun's entire ₹1,00,000 is gone, wiped clean, while he still owes the lender every rupee of the ₹3,00,000. Three years of "proof" undone in three weeks.

Notice what fooled him. It wasn't stupidity. It was a clean track record. The very smoothness of those three calm years is what lulled him - the longer the calm lasts, the safer the dangerous thing feels, right up until the shock arrives. The rare event that was missing from his sample didn't just dent his result; it swamped everything the three good years had built and then took his savings on top. A rare, ruinous event that hasn't shown up in your record yet is not gone - its absence is not proof of safety, and one arrival can wipe out every calm year that came before, so size your bets so no single shock can end the game.

The fix is not to fear everything. It's to always ask the second question - not "has this ever gone wrong for me?" but "what would one genuinely bad day do to me?" Arjun only ever asked the first, and the first question is answered by his tiny sample, which had simply never drawn the bad day. Had he asked the second, he'd have seen in an instant that a single ordinary crash - the kind that comes every few years - ends him completely. He'd built a house that stood beautifully in three summers of gentle weather, on a riverbank that floods once a decade.

Let's go one layer deeper, because there's a piece of arithmetic here that most people never see, and it explains why the missing rare event is so dangerous rather than just unlucky.

Imagine we could line up all the versions of a year that the machine might hand you, and sort them by how they'd end for your money - worst on the left, best on the right - and stack them into bars, like a crowd of possible years standing shoulder to shoulder. For a sensible investment, most of the crowd bunches up in the middle: lots and lots of "ordinary" years, a little up or a little down. Out on the right, a few tall "great" years. And out on the far left, standing alone, one or two rare "disaster" years - a crash, a default, a wipeout. There aren't many of them. They're rare. That's exactly why they hide.

But here's the twist. Rareness tells you how often something happens. It says nothing about how much it matters when it does. And the disaster-year on the far left is so deep - it takes away so much - that even though it almost never comes out of the machine, it can drag down your true average outcome more than all the pleasant ordinary years lift it. One year of losing everything cannot be repaired by ten years of gentle gains, because you cannot grow a number that has become zero. So a fair way to judge an investment is not "how does it do in a typical year?" (the fat middle of the crowd) but "how does it do once you include the rare monster on the left, weighted by how much it destroys?"

This is the quiet reason a short track record lies to you. A record is just a few draws from that crowd - three years, five years, even ten. If the rare disaster on the far left comes only once every fifteen or twenty years, a short record will usually skip it entirely, purely by chance. So the record shows you the fat, cheerful middle and hides the monster on the left. It reports the "typical year" and stays silent about the honest average. And you, reading the record, feel safe - not because the danger is small, but because your little sample was too short to have met it yet. The comfort and the risk are growing at the same time, hand in hand, which is the most dangerous combination there is.

Now we come to the slickest, most modern version of this whole mistake - the one that gets sold to beginners in shiny courses and videos. It's called the backtest, and it's the single-sample trap wearing a suit and tie. illustrative

Meet Aarohi. She's clever with a computer and wants a "system" - a fixed rule that tells her exactly when to buy and sell, so she never has to think or feel. So she takes the past ten years of a stock's prices and starts hunting for a rule that would have made money on it. She tries thousands of versions. Buy when the price crosses this line; no, that line; sell after five days; no, seven days; only on Tuesdays; only when it rose three days running. She lets the computer test every combination against the past ten years, and eventually - of course - it finds a winner. One particular rule, with one particular set of settings, would have turned ₹1,00,000 into ₹6,00,000 over that decade. The chart of it is beautiful, a smooth line climbing to the sky. She's thrilled. She's found the magic key.

Then she puts real money on it. And it doesn't work. It limps, it loses, the beautiful line falls apart the moment live rupees follow it. What went wrong?

Nothing went wrong with the computer. Everything went wrong with the thinking. Remember: the past ten years of that stock is one slip the machine handed out - one branch of the fan. When Aarohi tried thousands of rules against that one slip and kept only the rule that fit it best, she wasn't discovering how markets work. She was discovering the exact shape of that one particular past, wiggles and flukes and all. Try enough rules against any single history and one of them will match it perfectly - not because it's wise, but because with thousands of tries, some rule is bound to line up with the random bumps by pure chance. She memorised the noise of one slip and mistook it for a law of nature. The moment the machine drew a new slip - the live future, a different branch of the fan - her rule had nothing to say, because it had never learned anything real. It had only learned yesterday by heart.

So when someone shows you a strategy that "would have returned 400% in the past," your first feeling should not be excitement - it should be suspicion. Ask them the deadly question: how many different rules and settings did you try before you found this one? If the answer is "thousands," then the dazzling result is almost certainly just the best fit to one slip's flukes, and it tells you about that slip and nothing about tomorrow. A rule tuned until it perfectly fits the one past you happen to have tells you about that past, not the future - torture a single history with enough different settings and it will always confess to some magic rule that promptly fails on real money.

The honest version of testing a rule is almost the opposite of the exciting version. You'd want a reason the rule should work before you ever look at the numbers, and you'd want to test it on histories it was not built from - other years, other markets, the slips it never got to peek at. A rule that only shines on the exact past it was carved to fit has proven nothing at all.

The slip is almost never "I ignored the evidence." It's the opposite - people trip precisely because they respect the evidence, without noticing how thin and one-sided the evidence secretly is.

Two habits do most of the damage. The first is worshipping the outcome: seeing that something worked and concluding the choice behind it was wise, without ever picturing the other years it could have produced. This is how lucky gamblers become gurus and careful planners get mocked. The second is survivor-counting: you only ever hear from the people the machine treated kindly. The friend who doubled his money tells everyone; the ten friends who tried the identical move and got hurt go quiet. So the stories that reach your ears are pre-filtered to be the lucky slips, and the whole crowd of unlucky branches is invisible. Both habits share one root - mistaking the single sample you can see for the full box you cannot.

Now the honest part, because this idea, like every good idea, can be pushed until it breaks.

The first trap is turning into a person who calls everything luck and learns nothing. If every winner was "just a lucky slip" and every loss was "just an unlucky slip," then no choice is ever better or worse than any other, and you might as well flip a coin. That's not the lesson. The lesson is to grade the process, not to erase the result. A careful, well-reasoned bet that happened to lose still taught you good habits; a reckless bet that happened to win still taught you bad ones. You separate the quality of the thinking from the roll of the dice - and then you keep the good thinking regardless of how the dice fell. The point of imagining the fan is to judge choices better, not to give up on judging them at all.

The second trap is hiding in a corner forever. If you become so haunted by the rare disaster on the far left that you never invest at all - all your money sitting in cash, year after year - you haven't escaped the machine, you've just picked a different slow leak, as prices of everything creep up and your still money quietly buys less each year. Remember what the crowd of possible years actually showed: the fat cheerful middle is real too. Most years are ordinary, and ordinary compounding is how sensible people grow their savings. The goal was never to flee every risk. It was to make sure that when the rare monster does come out of the box - and over a long enough life, it will - no single one of your bets is big enough to end the whole game. Insure against ruin so you can stay in through the ordinary weather, not run from the weather entirely.

And a third, quieter caution: imagining alternative histories is guesswork, and you must hold it humbly. You don't actually know the true shape of the fan - nobody does. You're sketching it, roughly, from reason and from wide experience, and you can sketch it wrong. So use it as a tool for humility and caution, not as a fake crystal ball. It should make you say "I could be wrong, and I've built things so that being wrong doesn't wreck me" - not "I have calculated the future." The whole point of respecting how much you cannot see is that it stops you pretending you can see it.