Long Memory, from the Nile to the Marketplace

Flip a coin ten times. Heads, heads, tails, heads, tails, tails... Now here's a question that sounds silly but isn't: does the coin remember what it just did? If the last three flips were all heads, is the next one more likely to be tails, "to make things even"? Of course not. A coin has no memory at all. Each flip starts fresh, knowing nothing about the flip before it. Tomorrow's flip does not care one bit what happened yesterday.

For a long time, clever people assumed the stock market was exactly like that coin - a fresh flip every single day. Today's rise or fall, they said, is its own little event, unconnected to yesterday's, with no memory carried forward. If that were true, good days and bad days would be sprinkled about at random, like salt shaken over a plate, with no clumps, no streaks, no long runs of one thing.

But that turns out to be quietly, importantly wrong. Markets are not a memoryless coin. They are much more like weather, or a mood, or the water in a great river - things that carry the past forward, that lean the way they were already leaning, that get stuck in a spell and stay stuck far longer than pure chance would ever allow. Good years tend to arrive in clusters, bunched together like a run of sunny days. Bad years cluster too, one grim season leaning into the next. This stickiness - this refusal to forget - is what this chapter calls long memory. And once you truly believe markets have it, you plan your money in a completely different, sturdier way. A market that is already moving one way tends to keep moving that way for far longer than a fair-coin world would predict - so respect the run you are in rather than assuming each period starts fresh.

Let me show you why this matters with something older than any stock market: rain.

Imagine a village that keeps its water in one big pond, filled by the monsoon each year. The elders need to decide how deep to dig it. So they look back at the rainfall records. On average, they find, the monsoon brings enough. Some years a little more, some years a little less, but on average, fine. So a lazy planner would dig a pond just big enough for an average year and call it done.

Now, if rainfall were a memoryless coin - each year fresh, good and bad years sprinkled at random - that lazy pond would mostly work. A dry year would almost always be followed by a wet one that refilled the pond, so you'd never run empty for long. But rainfall is not a memoryless coin. Droughts come in runs. A dry year makes the ground and the weather patterns lean toward another dry year, and then another. The village can face three, four, five thin monsoons stacked back to back - a long dry spell far longer than "average" ever suggested. A pond sized for an average year runs bone dry in year two of a drought, and then people go thirsty for three more years waiting for the average to "come back."

This is the whole lesson, and it has almost nothing to do with the average rainfall. It is about the fact that bad years cluster. Because they cluster, you cannot plan for the average - you must plan for the long bad run. The elders who survive dig a pond deep enough to carry the village through five dry years in a row, not one. They keep staying power. They spend money on a bigger, more boring pond precisely so that a long clustered drought cannot finish them.

Your savings are that pond. The monsoon is the market's good and bad years. And the single most expensive mistake you can make is to plan for the average when reality arrives in long, sticky, clustered runs.

Let's make the difference between "memoryless" and "long memory" something you can see with your eyes.

Picture a strip of years. Colour each good year one shade and each bad year another. If markets were a fair, memoryless coin, the strip would look scattered - a good year, then a bad, then a good, mixed up like a chessboard, because each year is decided fresh with no leaning. You'd almost never see five bad years in a row, the same way you'd almost never flip five tails in a row: possible, but rare.

Now look at how markets actually tend to lay themselves out. The colours clump. You get a run of good years leaning into each other, then a turn, then a run of bad years leaning into each other. The strip looks less like a scattered chessboard and more like wide bands of weather - a long fair spell, a long foul spell. Same number of good and bad years overall, perhaps, but arranged in runs, not sprinkles.

Why do the colours clump? Because the things that move markets carry forward. When a good run is on, people who made money feel bolder and buy more, which pushes prices up, which makes more people feel bold - the mood feeds on itself, so one good year leans into the next. The same machinery runs in reverse: fear in a bad year makes people sell, which pushes prices down, which spreads more fear. A coin has no such feedback; a crowd of copying humans is nothing but feedback. That is the engine of long memory. It is not a glitch. It is what you should expect from any system where today's mood is shaped by yesterday's.

You can feel the difference in a small number. Suppose good and bad years are equally common. In a memoryless coin-world, the chance of hitting five bad years in a row is like flipping five tails - about one in thirty-two, genuinely rare, something you could half-ignore. But in a sticky, long-memory world where a bad year makes the next year lean bad, that same five-in-a-row run stops being a freak event and becomes something you should actively plan to meet at least once in an investing lifetime. The rare thing in the fantasy world is the ordinary thing in the real one. That single shift - from "long bad runs are freak accidents" to "long bad runs are to be expected" - changes every sensible decision you make with your money.

Here is the part that catches everyone out. The bottom strip is not broken. Nothing has gone wrong with it. This clumping is the normal behaviour of markets, weather, river floods, crowd moods - of almost anything driven by people copying each other and conditions that carry forward. The scattered top strip is the fantasy; the clumped bottom strip is the reality. Yet nearly every simple calculator, every "expected return" someone quotes you, quietly assumes the top strip. That gap - between the neat scattered world people plan for and the clumpy sticky world they actually live in - is where money gets lost.

Let's put real rupees on the table and watch long memory do its work. illustrative

Meet Rohan. He is sensible - no lottery tickets, no hot tips. He starts a monthly SIP of ₹20,000 into a broad Indian equity fund, the boring diversified kind. Before he begins, a friend runs him a tidy calculation on a spreadsheet. It assumes each year is a fresh, independent coin flip around a nice long-term average. On that neat assumption, the spreadsheet promises that even in a rough patch, markets "bounce back within a year or so," and his savings should climb in a fairly smooth line. Rohan feels reassured and begins.

For three years it works beautifully. His pot grows to about ₹9,00,000, ahead of even the cheerful spreadsheet. Then the weather turns. A long foul spell begins - not a crash-and-recover, but a clustered run of bad years, each grim season leaning into the next. Year one of the spell, his pot drops. Fine, he thinks, this is the "bounce back within a year" the spreadsheet promised. But there is no bounce. Year two, it drops again and drifts. Year three, still heavy, still sideways-and-down. His holdings, which the neat model said should recover in about a year, are now underwater for the better part of four years.

Here is what the tidy spreadsheet got wrong. It assumed a bad year forgets itself and the next year starts fresh, so a long deep valley was treated as almost impossible. But real markets remember. A bad spell leans into another bad spell, and drawdowns run far longer than a fair-coin world predicts. Rohan didn't do anything foolish. His model was foolish - it planned for the scattered strip while he was living in the clumped one. The danger wasn't a single bad year; it was that the bad years came stuck together. A downtrend, once it has set in, tends to keep running until something clearly turns it, so a drawdown can stretch for years, not months - plan for the long valley, not the quick dip.

And notice what saved him, or rather what would save him: staying power. Because his SIP was small relative to his income and he had a separate cash cushion, he could keep buying quietly all through the long valley - even buying more units cheaply - and simply outlast the foul spell. The investor sitting next to him, who had planned for the average and had no cushion, panicked in year two and sold at the bottom. Same market, same memory, opposite outcome. The difference was entirely staying power.

Long memory does something sneaky that a fair-coin world could never do: it makes the order of your good and bad years matter enormously, even when the average is identical. Let's watch. illustrative

Meet two sisters, Aayra and Haridya. They each invest the same lump sum, ₹10,00,000, into the same kind of fund, at the same skill, for the same fifteen years. Over those fifteen years, both live through exactly the same set of yearly returns - the same good years and the same bad years, the same average. If markets were memoryless and only the average mattered, they should end up in exactly the same place.

But the years don't arrive in the same order, because each began at a different moment in a sticky, clustered market. Aayra happened to begin just as a long good run was starting. Her early years compounded upward, so that by the time the long bad spell arrived, her pot was already large and could absorb the fall from a position of strength. Haridya began just before a long bad spell. Her early years dug a hole, and she spent years climbing out of it before the good run finally reached her - and by then she was compounding a much smaller base. Fifteen years later, with the identical set of returns, Aayra ends near ₹34,00,000 and Haridya near ₹22,00,000. The gap - more than ₹12,00,000 - is not skill, not the average, not who picked better. It is purely the path.

This is the quiet cruelty of a market with memory: because good and bad years cluster into runs, where you happen to stand in the run shapes your whole result. Two people with the same average returns can end up far apart because the order in which the good and bad years arrive compounds differently - the path you travel, not just the destination's average, decides the outcome.

Sit with how strange this is, because it breaks a rule most people carry in their heads. We're taught that if two things are equal on average, they should end up roughly equal. Add up the same numbers in a different order and you still get the same total, after all. But compounding is not adding - it is multiplying, year on year, and multiplying cares deeply about order. A bad year early shrinks the base that every later good year has to grow from, while a good year early hands every later year a bigger base to work on. So in a compounding, clustered world, the same set of returns in a different order really can hand one sister a life-changingly larger pot than the other. Averages hide this completely; only the path reveals it.

The honest takeaway isn't "be born lucky." It's humbler and more useful. First, do not confuse Aayra's larger pot with Aayra being cleverer - she mostly caught a kinder path, and next time the paths could swap. Second, and this is the practical bit: because you can never know which path you'll get, you should invest in a way that survives the ugly path, not just the pretty one. If your plan only works when the good run comes first, it is not a plan - it is a wish. Haridya's rescue was, again, staying power: enough of a cushion and enough patience to survive the early bad run without selling, so that she was still standing when the good years finally arrived.

By now the practical shape of the idea should be forming. If bad years cluster into long runs, and if the order of runs can put you in a deep hole early, then the one thing that matters above cleverness is: can you survive the long bad run without being forced to sell? That survival capacity is your pond. Let's size it properly. illustrative

Meet Arjun, who plans carefully. He has ₹40,00,000 invested for the long term and lives partly off it. A memoryless model whispers that a bad patch lasts about a year, so he need only keep, say, one year of expenses - ₹6,00,000 - in safe cash as a buffer. That feels sensible. But Arjun has learned that markets remember, and that a clustered bad run can keep his invested pot underwater for four or five years, not one. So he does the boring, sturdy thing: he sizes his cash pond for a long dry spell, keeping closer to four years of expenses - around ₹24,00,000 - in safe, steady holdings, and only the rest in equities he can leave completely alone.

Watch what this buys him when a long foul spell actually arrives and his equity pot falls hard and stays down for four years. Arjun never has to sell a single falling share to eat, because his deep pond covers his needs right through the drought. He waits out the whole clustered run in comfort and is fully invested when the good years finally cluster back. His neighbour, who kept only the one-year buffer the memoryless model recommended, ran dry in year two and had to sell shares at their lowest to pay for ordinary life - locking in the loss forever. Same market, same memory. One had a pond sized for the real world; one didn't.

So staying power isn't a side detail; in a market with memory, it is the main event. Cleverness picks the flowers; staying power keeps you alive through the long winter so you're still there when spring clusters back. And here's the beauty of it - you do not need to predict the long bad run to survive it. You only need to have prepared for it, by keeping a pond deep enough that no clustered drought can ever force your hand. Preparation beats prediction every time, because you cannot forecast the weather, but you can always dig a bigger pond.

There is a second, sneakier danger inside long memory, and it hides not in the bad runs but in the good ones.

Because good years cluster too, markets can hand you a long, uninterrupted spell of things going right - five, six, seven years where the line just climbs, nothing breaks, and every worry you had looks foolish in hindsight. This feels wonderful. It also does something dangerous to human minds: the longer the calm lasts, the safer everyone feels, right at the moment they are becoming less safe. People forget what a bad run even looks like. They shrink their cushions. They borrow to buy more, because "the market always comes back quickly" - and in a long good run, it has, so the belief seems proven. The very length of the good cluster is what lulls people into throwing away their staying power.

This is the trap. Long memory means a good run can stretch on and on - but it also means that when the turn finally comes, the following bad run can stretch on and on too. The people who used the long calm to load up on debt and cut their cushions to the bone are precisely the people a long clustered drought destroys, because they have no pond left when it stops raining. A long calm run of good years quietly tempts people to borrow more and hold less cushion, so the stretch that feels safest is the stretch that is secretly building the fragility that the next bad run will break.

Picture Aarvi in year six of a glorious good run. Her ₹15,00,000 has grown to ₹28,00,000, and it has been years since anything went wrong. A voice says: why keep a boring cash pond at all? Markets bounce back fast - look, they always have! So she spends her cushion, and even borrows ₹5,00,000 against her holdings to buy more, confident the good run will pay the loan. Then the weather turns into a long clustered bad spell. Her holdings fall and stay fallen for years, but her loan does not wait - it demands payment through the whole valley. With no cushion and a loan to feed, she is forced to sell into the drought at the worst possible time. The long calm didn't protect her. It disarmed her, right before the storm. The safest-feeling years were the ones that set the trap.

Almost every slip around long memory comes from one wrong belief: "it must turn around soon."

It shows up on both sides. In a bad run, after the market has fallen for a year, people whisper "surely it's due to bounce" - and so they spend their cushion early, or borrow to "buy the dip," expecting a quick recovery. But a market with memory owes you no quick bounce; the down-run can lean into another down-run and grind on for years, and the person who bet on a fast turn runs dry long before the turn arrives. In a good run, the same wrong belief flips: "surely it's due for a fall," so people sell out of a rising market far too early and miss years of clustered gains. Either way, the mistake is treating the market like a coin that owes the other side a turn. It doesn't. It has no sense of "due."

Now the honest part, because long memory is powerful enough to be dangerous when it's misused.

The first trap is to hear "trends persist" and turn it into "so I can ride the trend and hop off right before it ends." You can't. Long memory tells you that runs tend to last - it does not tell you when they stop. The turn is only ever obvious after it has happened. Anyone who claims they can feel the exact top of a good run or the exact bottom of a bad one is fooling you or themselves. The lesson of long memory is not "time the turns"; it is the opposite - because you cannot time the turns, build a plan sturdy enough that you never need to. Preparation, not prediction.

The second trap is to swing so far that you become frozen with fear of the long bad run and never invest at all, hiding everything in cash forever. But a pond that is never filled is just as useless as one that runs dry. Sitting entirely out means a slow, certain loss to rising prices over the years - a different way of drowning, only quieter. The goal was never to avoid every bad run. It was to survive the bad runs so you're still holding your units when the good runs cluster back, because over long stretches those good clusters are what actually grow your money. Staying power is meant to keep you in the game, not to keep you out of it.

The third caution is subtler. Long memory is a strong tendency, not an iron law. Runs cluster more than a coin would allow - but not every year belongs to a neat run, and sometimes a spell really does turn quickly and surprise everyone. So don't use "trends persist" as an excuse to ignore a real, clear change in a business or the world just because the old run was pleasant. The idea is a lens for humility and preparation - expect long spells, size your pond for them, stop betting on quick reversals - not a crystal ball that lets you predict the shape of the next ten years. Use it to become sturdier, never to become certain.