How Many Tickets Would You Need?
Every lottery, no matter how it's dressed up, comes down to the same three numbers. There's the total pool of possible outcomes — every combination the machine could possibly spit out. There's the number of those outcomes that actually pay out — the winning tickets, however many of them there are. And there's the number of tickets that actually get sold — how many people, or in our case how many computers, actually take a swing.
Everything we've done so far has been an exercise in nailing down those same three numbers for a much stranger lottery. The total pool of outcomes is every possible three-hundred-letter string — all four hundred twenty-five drums' worth of them. The winning tickets, up to now, have been a small, narrow target: any Shakespeare sonnet, one of a mere hundred fifty-some. And the tickets sold have been however many random strings our swarm of computers can crank out.
A hundred fifty winning tickets, out of a four-hundred-twenty-five-drum pool, is about as narrow a target as you could ask for. So let's widen it. Dramatically.
Instead of asking for one of Shakespeare's actual sonnets, let's ask for something much less exacting: any three-hundred-letter string that would make sense — in any language, living or dead. Not eloquent. Not poetic. Just sensible. A real fragment of real language, rather than anything containing even a bit of gibberish.
Here's the trouble: nobody can actually calculate how many three-hundred-letter strings satisfy that description. It depends on the language, on grammar, on which letter can follow which, and the honest answer is that at some points in building a string almost any letter will do, and at other points only one specific letter will keep the thing alive. There's no clean formula for that.
So let's do what we've done before when the real number is out of reach: make a generous, clearly stated assumption, and follow the arithmetic wherever it leads.
Here's the assumption. Imagine that at every single position while building our three-hundred-letter string, exactly half the alphabet — thirteen letters out of twenty-six — would keep the string sensible, and the other half would break it. It's a clean, fair, defensible starting point, and — importantly — it doesn't rig the game against the very idea we're testing.
It's worth pausing on just how generous that fifty-percent assumption really is, because the true number is nowhere close. Take any real sentence, in any language — pick one at random, out of a book, off a sign, out of your own head. Now change a single letter in it, anywhere. In the overwhelming majority of cases, that one small change doesn't produce a slightly different sentence — it produces nonsense. A misspelling, a non-word, a string that means nothing at all.
That's the real asymmetry hiding underneath our assumption: there are vastly more ways to break a sentence than there are ways to build one. Every position in a working string is surrounded, on nearly every side, by letters that would ruin it, and only a small handful that would keep it standing. Fifty percent — thirteen workable letters out of twenty-six, at every single position — isn't a realistic estimate of that ratio. It's an extravagantly generous one, offered deliberately.
Work out how many three-hundred-letter strings survive under that assumption, and you get thirteen multiplied by itself three hundred times. In drums, that number takes three hundred thirty-five to write out. That is the number of sensible three-hundred-letter strings under our generous assumption — an astronomically large number of winning tickets.
How does three hundred thirty-five compare to the size of the whole pool, four hundred twenty-five drums' worth of possible strings? Subtract one from the other, and you get ninety. Ninety drums.
So even under our generous assumption, sensible strings make up only about one out of every ten-to-the-ninetieth-power possible strings. The winning tickets are enormously numerous — three hundred thirty-five drums' worth — but the losing tickets outnumber them so overwhelmingly that buying one ticket is like playing a lottery with ninety drums' worth of numbers and a single winner: one chance in a pool ten billion times larger than the number of atoms in the observable universe. (The arithmetic: Sensible Strings.)
To some, this may seem completely backwards. If half the alphabet keeps the string alive at every single point, fifty percent per spot sounds like a pretty good shot at winning. So let's see why that intuition is dead wrong.
Turn the question into coin flips. Saying that half the alphabet works at every position is like saying that heads wins at that spot and tails loses — and that holds for every one of the three hundred positions. Heads at every spot wins; tails at any spot loses. Do you see the difference? There is only one way to win: heads, every time, three hundred times in a row. If even one flip lands tails, you lose. One tail anywhere, two tails anywhere, any combination at all that isn't all heads — it loses. The losers swamp the single way of winning.
Count it with an odometer whose drums show only two symbols, 0 and 1. Three hundred such drums give 2 multiplied by itself three hundred times possible settings — about ten to the ninetieth, the same number we got a moment ago. All heads is just one of those settings. Basically impossible to win.
So if the losers outnumber the winners by ninety drums, how many tickets would you actually need to buy for a real, fighting chance of winning? Not a guarantee. A real shot. The answer, it turns out, tracks the size of that gap almost exactly: you need something in the neighborhood of ninety drums' worth of tickets. That's ten billion tickets for every atom in the observable universe, just to have a reasonable chance of winning.
Think about that for a moment. We started with a small target: a hundred fifty-some sonnets. We widened it enormously, to any three-hundred-letter string that makes sense in any language — a staggering number of winning tickets. And still, there are so many more losing tickets out there that to have a chance of winning, we'd have to buy ten billion tickets for every atom in the universe. That's ridiculous. This lottery is a cheat.
Can anything, anywhere, buy ninety drums' worth of tickets?
Let's find out using actual matter instead of the impossible blown-up-universe swarm from earlier in this chapter. This time, no blowing atoms up into universes of their own. Just the atoms we actually have.
Start with the whole observable universe. It holds, as we've established, about ten to the eightieth power atoms. Most of that matter, though, is out of reach: either locked away inside stars — crushed, fused, and about as useful for building anything as the inside of a furnace — or spread thinly as gas across the vast spaces between the galaxies. Let's generously allow one-tenth of the total as usable: about ten to the seventy-ninth atoms, give or take.
Take those atoms and build ordinary computers out of them — real ones, not hypothetical universe-sized swarms. A small modern computer, about the size of a phone, is made of somewhere around ten to the twenty-fifth atoms. So the universe's usable matter could build somewhere around ten to the fifty-fourth computers. Set every one of them running at our familiar rate — a million random letters a second — and let them run for the entire age of the universe, every second of it.
The whole exercise produces about seventy-eight drums' worth of trials. (The arithmetic: Computers from Real Matter.)
Seventy-eight, against the ninety we needed. Short. Not by a little. By twelve full drums — which, remember, is a factor of a trillion. Every usable atom in the entire universe, turned into computers, run for the full age of everything, and it still isn't close.
Now bring the question down to Earth, literally.
Earth holds around ten to the fiftieth atoms. Most of that, too, is off the table: the planet's core and mantle together make up nearly everything, buried under pressure and heat that would crush any device we could imagine. What's left is the crust — a thin shell, well under one percent of Earth's total mass.
Let's not just count the crust matter that could realistically be shaped into a computer. Let's count all of it — every atom in every ocean, every breath of atmosphere, every living thing, every grain of sand and seam of rock in the crust, whether or not it could ever plausibly become a working device. None of that generosity is necessary. It's a gift to the other side of the argument, freely given, so that nobody can later object that we shortchanged the calculation.
Even with that entire, maximally generous allowance, the crust, oceans and atmosphere together hold only about ten to the forty-eighth atoms — a sliver of Earth's total, and a vanishingly small sliver of the universe's. Build computers out of all of it, run them at the same million letters a second, for the same full age of the universe, and the total output comes to about forty-seven drums' worth of trials.
Forty-seven, against the ninety we needed. Not just short — short by an entirely different order of failure than the whole-universe figure. Forty-three drums short, this time. The gap between "not enough" and "nowhere close to enough" is itself enormous, and Earth, on its own, lands firmly in the second category.
So even granting every usable atom in the observable universe, run for the full span of cosmic time, randomness alone falls short of a fair shot at our broadened, generously defined target — not one narrow set of sonnets, but any sensible three-hundred-letter string in any language. And Earth alone, which is of course the only place any of this actually has to happen, isn't merely short. It's short by a margin so large that "short" barely captures it.
Remember ninety drums. We're going to need it again, and soon.
Last revised 2026-11-16