Early last week, just a few days before the inauguration of Donald Trump, I made a bet with my podcast co-host Andrew Barber. It’s one that I hope desperately to lose.

I’ve been doing a series of Twitter polls out of curiosity about how people expect his presidency to go. I’ve asked, for instance, how many protesters will be killed in the US in the next four years. That one turned out to be polarized between people who thought it would be less than ten, and people who thought it would be more than one thousand.

When I asked about the odds that a nuclear weapon would be detonated in the same time period (tests not included), I expected similar results. But on that front, most people who answered were optimistic, setting the line at under 0.5%. On Facebook, Andrew guessed 50-1 (~2%). I thought even this was too optimistic, and said I’d guess closer to 15-1 (~6.5%). After a bit of negotiation, we ended up betting at 20-1, my $100 against his $2000. You can see why I’m not eager to win this bet.

20-1 is long odds, but 200-1 is insane

I may indeed be getting the worst of it at 20-1, but I’m pretty sure that Andrew and I are on the right order of magnitude. Just looking at the baseline odds, if you picked a random year between the invention of nuclear weapons and today, the odds are greater than 1% that you’ve picked a year in which they were used. Yes, there have only been two instances, and they happened within the same year, but the weapons themselves have existed for fewer than 100 years.

Now, there’s truth to the “mutually assured destruction” hypothesis, that the proliferation of nuclear weapons presents a big deterrent to their use. At the same time, it directly increases the number of entities that could choose to use them, and the number of scenarios under which that could happen.

It wasn’t very long ago, just a couple of years, that the idea of a Trump presidency was not just impossible, but comically impossible. Right up until the ballots were counted, it was still regarded by most as highly improbable. Andrew himself made multiple bets against it happening. There are patently not normal times and we should recognize that our lived experience is likely to mislead us when it comes to the likelihood of things happening which we have not experienced.

More to the point, though, even ignoring that the current world situation is uncharted territory, consider what it actually means when you assign a very small probability to something happening in the next four years.

The eight-hundred year storm

Consider bridges. A seeming non-sequitur, I know. But I have friends who are civil engineers, and I know from talking to them that when it comes to designing structures, you have to decide just how improbable an event you want them to withstand.

Non-engineers tend to be shocked when a structure collapses, as if it were a reasonable expectation of engineers to design things so as to be entirely, permanently indestructible. Although there’s no real hard upper limit on how sturdy you can make something, the costs grow exponentially, since you end up needing supports to support the supports, and so forth. It would be nice if we lived in a world where we didn’t have to assign dollar figures to human life, but if we didn’t, nothing could ever get built, which brings us to storms.

There are all sorts of things that could make a structure collapse, but storms are by and large the most likely cause. And so, structures are built with various degrees of storm in mind, described as “the 20-year storm,” “the 50-year storm,” “the hundred-year storm” and so forth. In other words, based on weather records for the place where the structure is to be built, it’s designed to withstand the most powerful storm that place is likely to see in a given period of time. The more important the structure and more dire the consequences of its failure, the longer that period of time is.

When I did my poll about a nuclear weapon being used, the lowest probability I listed was “less than 0.5%.” Keep in mind that this is over four years, i.e. the first term of President Trump, assuming he makes it that far. Claiming it is that unlikely, then, is equivalent to saying that given the current world situation, the average interval between instances of nuclear weapons being used is greater than 800 years.

A lot happens in 800 years. Ignoring the 20th century for the moment, we’ve seen the tail end of the Crusades and Genghis Khan’s campaigns, the Hundred Years’ War, the Black Plague, the European colonization of the “New World” and genocide of most of its populace, the American Revolution, War of 1812 and Civil War, and countless other wars, atrocities and disasters far too numerous to mention.

Considering the prospect from that perspective, is the use of nuclear weapons an 800-year event, given that they already exist? Probably not.

The eighty-year storm

If an event with a 0.5% chance of occurring in a four-year period is an 800-year storm, then one with a 5% chance of occurring in that window is an 80-year storm. It just so happens that 80 years ago works out to be 1937, just after Hitler and Mussolini struck their alliance and just before World War Two began in earnest.

Naturally, we don’t expect an 80-year event to occur once every 80 years like clockwork; that’s the same fallacious reasoning as expecting a your number to come up at roulette because it hasn’t for the last 35 spins. However, in the absence of any more solid evidence, you can do worse in estimating the frequency of an event by using the last time it happened.

As it turns out, though, we actually do have a second data point; not of an actual nuclear event, but an extremely near miss, on October 27, 1962, during the Cuban Missile crisis. Under the mistaken belief that war had broken out, the Soviet submarine B-59 prepared to launch a nuclear torpedo. All three senior officers needed to agree, and the lone holdout was the flotilla commander Vasili Arkhipov. For someone on board that submarine at that time, the odds of the torpedo being launched would surely have exceeded even money prior to Arkhipov’s intervention.

Thus, we have what you might call “one and a half” nuclear attacks in the past 80 years, which reinforces the notion that it’s closer to an 80-year event than an 800-year one, although the error bars on that estimate are huge.

Once in a lifetime

Eighty years also happens to be roughly equal to human life expectancy in developed countries. Thus, an 80-year event is the most literal way you could interpret the expression “once in a lifetime.”

Looking at some of the things which have happened in the past 80 years, we can get a feel for just how exceptional events on this timescale can be. To name just a few: the Second World War; the Vietnam War; the 9/11 attacks; the meltdown at Chernobyl; the Bangladesh Cyclone of 1991, which killed an estimated 135,000; the eradication of smallpox; the appearance of HIV; the creation of the Israeli state; the collapse of the USSR; the moon landings; and the dawn of the Internet era.

When it comes to human-caused (or human-preventable) catastrophes, there’s also a non-coincidental reason why we might expect specific scenarios to be once-in-a-lifetime. Moreover, this reason is intimately connected to the fact that so many people consider the use of nuclear weapons to be so unlikely. This is that the people’s assessment of the risk of a given scenario depends heavily on whether they’ve experienced it themselves, and if so, how well they remember it. At a certain point, the conviction that something mustn’t be allowed to happen again begins to replaced by the belief that it won’t.

Now, 72 years after the bombings of Hiroshima and Nagasaki, the percentage of the population which was old enough at the time to remember them is declining sharply. The same goes for the Second World War. That being the case, it’s not unreasonable to fear that the odds of repeating those mistakes are growing.

The broken symmetry

One final reason that I suspect people assign such a low probability to the use of nuclear weapons is that they’re imagining a Dr. Strangelove-style all-out, end-of-the-world deployment of strategic nuclear bombs. The odds that two or more of the world’s major nuclear powers begin bombing one another are probably considerably less than 0.5%, or at least so one would hope.

But that’s a fear left over from the Cold War era, when the United States and USSR were relatively even and opposing forces. That’s a symmetry that no longer exists; Russia is still powerful, of course, and still has nuclear weapons, but the two are no longer at war and may even become allies under a Trump presidency. Meanwhile, the most likely targets for a nuclear strike by these countries are not themselves nuclear powers and cannot retaliate in kind, while the rest of the world would likely respond diplomatically, but not risk escalating the situation by involving themselves militarily.

Nor is the use of weapons by a state the only plausible scenario. Several states which either have or are developing nuclear capabilities have cause to want to see the western world destabilized. Many are themselves unstable. Even if it would be suicidal for these governments to deploy their weapons directly, there are any number of ways one could imagine the weapons being supplied to a terrorist organization or taken over by a rogue military. Most of these seem fanciful when considered individually, but a sufficient number of individually unlikely scenarios can begin to add up to something worth worrying about.

Why bet on it?

At this point, many readers are likely wondering how I could be cynical enough to bet on something like this. I was hesitant to, but if you’ve listened to the podcast, you’ve likely heard Andrew and me talking about the value of betting as a cognitive tool.

When you’re thinking about things instinctively, emotions tend to involve themselves. This can distort probabilities and make certain scenarios either loom larger or seem impossibly remote depending on your hopes and fears, your self-image, your political beliefs, and so forth.

Betting on an event forces you to think rationally and quantitatively, and to try to find numbers to either support or refute your instinctive guesses. Quite often, it turns out that your opinion changes as a result.

Let me leave you with one further thought: Politicians all around the world are making bets like these all the time, just not necessarily with money. In some cases, the chips being wagered are human lives. Having walked through my logic with me on this, would you guess that a majority of them are overestimating, or underestimating the likelihood of such a scenario? And which of these possibilities is more consequential?

Alex Weldon (@benefactumgames) is a freelance writer, game designer and semipro poker player from Montreal, Quebec, Canada.