Chapter 17: What You Have to Believe
The Slot Machine
Suppose you own a casino, and slot machines are its only business. You have a hundred of them. Each one cost you a hundred thousand dollars, and all of them were built to the same specification by the same manufacturer. In any given week, each machine has about a one-in-a-hundred chance of paying a jackpot — so a single machine pays out about once every two years, and across the whole floor, about one jackpot a week. That's just what happens, week after week. The machines behave the way expensive, carefully engineered machines are supposed to behave.
One morning your floor manager hands you a report. One particular machine has paid a jackpot in each of the last three weeks. You weren't watching it happen, and nobody is asking you to wait and see whether it does it a fourth time. It's over, entirely in the past. All you have in front of you is the record, and a decision to make from it, now: keep the machine running, or pull it.
You call the manufacturer. They tell you, quite confidently, that nothing is wrong. Every machine is built identically; there's no way one of them could start paying more than the others. Repair is out of the question, since there's nothing to repair. So you take the report to a statistician. She runs the numbers: the chance of one machine paying a jackpot three weeks running, by chance alone, is one in a million. (The arithmetic: The Slot Machine.) Rare, she admits — but there are a great many slot machines in the world, so somebody's machine was bound to be the one in a million. If you want to be thorough, count every slot machine on every planet that might have casinos, and yours becomes a vanishingly small part of the total. Count every universe there may ever have been, and a machine doing exactly this is practically guaranteed somewhere. You weren't cheated. You just own the one that did it.
Now suppose the report said something different. Not three weeks running, but four: one in a hundred million. Or five: one in ten billion. In every version, the way you're asked to decide stays the same. You never see the machine misbehave. You're handed a finished record of the past, once, and asked to judge it by the number attached to it.
At some point — and for most people, well before one in ten billion — a sensible owner stops listening to the statistician and pulls the machine. Not because anyone has shown what's wrong with it, but because the number leaves no other reasonable choice. And notice what plays no part in that decision: how many other slot machines exist elsewhere, how many planets might have casinos, how many universes might be out there. None of it changes the report on your desk, and no sensible owner would let it. Notice, too, that the decision is a judgment about the past, made with probability — exactly what we were told can't be done.
A statistician might raise one more objection, and it's a fair one. Every result is improbable. Shuffle a deck of cards, and the exact order you get had a chance of about one in ten to the sixty-seventh — yet nobody calls the dealer a cheat. Why is the slot machine different? Because the owner isn't looking at just any pattern. She's looking at the one pattern that matters, picked out in advance: jackpots. A deck in some random order means nothing; a deck that comes out sorted by suit means somebody arranged it. Life belongs with the sorted deck. A working protein isn't just any string of amino acids; it's one of the rare strings that does a job. That's what the winning tickets in Chapter 2 were: not merely an improbable outcome, but an improbable outcome of the kind that counts.
The same three appeals are made when the subject turns from a slot machine to life itself, and they fail there for the same reasons, only more so.
The appeal to other machines works for a casino owner, so far as it works at all, because the other machines are known to exist. She can point to inventories, other casinos, machines running under the same conditions. Nothing like that exists for life. There is no inventory of other planets where life was attempted and failed to arise. There isn't even confirmation that suitable conditions exist anywhere else. Spreading the odds over a vast field of other trials works only if those trials are known to exist. Here, not one of them is.
The multiverse fails more completely still. It isn't merely unconfirmed; it can't be confirmed, by definition. Any other universe we could detect, measure, or signal to would, for that very reason, be part of our own. A truly separate universe is one that can never be checked, even in principle. Reaching for a hypothesis that can never be tested, at just the moment a number becomes uncomfortable, isn't evidence of anything. It only tells you where the discomfort was sent.
And "we're here, so it can't have been that hard" fails for the same underlying reason. It treats the outcome as its own justification, without producing a number, another trial, or anything else that could be checked.
Setting the number aside is not a neutral act. It isn't "no decision." To proceed as though an extremely improbable event needs no explanation is itself a decision, made with no evidence to support it: no confirmed other trials, no confirmed other universes, nothing but the discomfort of the number and a wish for it to go away. A casino owner who reasoned that way, with a hundred thousand dollars at stake, would be called irrational by anyone watching, and rightly. The reasoning doesn't improve when the stakes change from a casino floor to the question of how anyone came to be here to ask.
Last revised 2026-11-19