It would save quite a few ocean going ships if this was more appreciated. Those once-in-a-thousand-year-rogue-waves are much more frequent than once in a thousand years.

https://www.theguardian.com/science/2006/jul/26/uknews

But we're still designing ships as though they only happen once every thousand years because the cargo and the ship are insured and the crews are considered expendable.

> Research at Imperial College, London, shows that far from being rare events, rogue waves can emerge frequently, and may be responsible for some of the 200 supertankers and container ships longer than 200m that have sunk in poor weather conditions in the past two decades.

The loss of ten ships a year seems incredible to me. Surely "weather" could be avoided.

This amount of loss is steep. Why isn't this being investigated? We should have ocean swell data from every ship that capsizes.

https://www.statista.com/statistics/1326758/number-of-oil-ta...: There were 8,883 oil tankers worldwide in 2021.

If those are all ocean-going, ten a year is about 1.1%. Depending on how expensive it would be to prevent this, that may not be bad for ship owners, but it is very bad for those seeking a career in working on those ships.

1.1% of what?

Of ocean-going oil tankers.

It's ~0.11% of those

Relative to what other transportation profession, though? Pilots, truck drivers, train operators? All of them have deadly accidents.

plus 5,589 containerships

https://www.statista.com/statistics/198227/forecast-for-glob...

although the parent comment mentioned "longer than 200m" but its difficult to find containership stats by length rather than by TEU.

ignoring that though it would be 0.07% lost per year or 1 out of every 1447.

probably not too surprising, since if you pick 1447 oil tankers/containerships and audit them consider what the worst looking one is going to look like after decades of salt damage and abuse and poor maintenance.

'Surely "weather", if it were to blame, could be avoided.'

AFAIK rogue waves are called 'rogue' due to their unpredictability.

Nomenclature aside, they're somewhat predictable in the sense that if the most dangerous rogue waves will typically happen in or near low preasure systems, and especially around strong currents, where you have wind-over-tide type nastiness.

So in the sense of trying not to die, avoid the Gulf and Agulhas currents, especially when a low blows through.

This coming from an amateur sailor. I believe freight shipping etc just plows through most weather systems in order to avoid delays.

There are also source of loss, nothing can be done against. One can not even predict them.

Ocean floor methanhydrate and a volcanic shift, results in tremendous bubble columns, which just open up a shaft that can carry nothing. Water begins to boil, ship falls in, breaks apart... End of story.

Surely the gas-producing seafloor distribution can also be studied and modeled.

How many incidents of this have there been? I've heard of the theory, but no actual losses attributed to it.

Because a few ships a year are expendable for the global economy.

Not worth it to try to make every ship less sinkable - e.g. it may decrease the number of annual losses by 50% at the cost of making an average ship 10% more expensive - obviously a bad trade. Some ships will still sink, because of stuff like human errors.

You know that ships have crew in it?

Yeah, and?

How much do you think the humankind is willing to pay to save a human live - a million dollars, a billion, a trillion?

A lot more than it does at the moment, certainly. That's no bad thing either, even if insurance premiums would go up (and construction quality would then improve to compensate and bring them somewhat back down, and finally fewer incidents would result in loss of life.)

The crew probably has a good idea of the risks

I'm not sure they do - are they oceanographers? These are very rare events; very few crewmembers have experienced such waves.

Also, even if they do, it doesn't stop them from being exploited.

If you have an idea of the risks involved, then certainly people who are embedded in sea-going culture do as well.

Occupational safety doesn't happen just because you think it should. Recognizing when financial incentives are insufficient is important to designing a realistic solution.

This trade-off is everywhere. As a society, we choose to have some people die due to speeding instead of making speed limits 15mph, which would absolutely make roads safer. Human life just doesn't trump all else - we factor in other things.

And a company would go broke trying to secure itself against every conceivably IT security threat, so IT security aims to keep breaches out of the headlines rather than eliminate them completely.

Your own skeletal system is a trade-off of different factors.

Is there a statistics about ships sinking because of rogue waves?

Recent news story:

>One passenger was killed and four more injured after a "rogue wave" hit a cruise ship bound for Antarctica, travel company Viking has said.

https://www.bbc.com/news/world-latin-america-63846157 (Dec 2022)

This is so true. Heavy-tailed distributions and extreme value theory [1,2] should be taught more widely.

The normal distribution is just a very particular case of the generalized CLT.

For example, some drug effects are heavy-tailed, but safety trials are often ignoring this.

[1] https://adamwierman.com/wp-content/uploads/2020/09/2013-SIGM...

[2] https://adamwierman.com/wp-content/uploads/2021/05/book-05-1...

> some drug effects are heavy-tailed

where can i read about this? curious

Got any sources which discuss the heavy-tailed effects of drugs and how they're ignored by safety trials? I'd love to read more!

I don't have references with me right now, but there are quite a few reputable studies that have found effects to be log-normal.

Log-normal is a quite nasty heavy-tailed distribution as it's easy to take it for normal if you have small sample sizes.

This has tricked the safety calculations of a few trials. I don't think it's too worrying or generalized, just some area to improve.

Heavy-tailed modeling doesn't reallly matter. There are very few enterprices where preparing for extremely rare events is worth the effort - nuclear power plants, weapons of mass destruction, bio labs working with dangerous organisms, big chemical plants, maybe AI research and big particle accelerators.

the most important of all - your 401k

What does a 401k have to do with extremely rare events?

Preparing for stuff like a nuclear war would mean buying land in South America, Africa or Australia, building an underground bunker, stacking on water, food, medicine and weapons, etc.

so you buy an index fund or ETF. capture some of the positives if/when they happen, and help mitigate the negatives

some would argue performance of the 60/40 portfolio last year was an event from a tail of a very non-normal distribution

If you tuned out of current events, financial and otherwise, five years ago - in early 2018 - and your 401k was entirely on autopilot in an S&P 500 fund, it looks like your average annual return would've been about 11.5%, annual US inflation was 3.7%, and return after inflation 7.5%. Per year. You can't reasonably expect better.

And practically speaking, the tax avoidance means that's considerably better than 11% in a non-retirement account, but it was too much effort to do the math.

Despite all the pestilence, war, earthquakes, and stuff like that, doing nothing was the smart move again, with $1 or $1 million.

I don't think it makes sense to prepare for a global collapse of civilization because in the event whatever you will reap won't make hardly any difference in how much things suck. You can't eat money, much less use it as a shelter from a global rain of molten rock.

So on principle, regardless of the probability of Armageddon, I don't think one should give up much to try to hedge the end of the world.

I'm no expert on stats, this discussion is a bit over my head, but it reminds me that not all distributions have a mean. A Cauchy distribution has no mean; as you average successive samples, it never settles down to a consistent average. I was surprised when I first encountered Cauchy distributions and up until then had considered that all distributions would have a mean. I suppose it is one of those "heavy-tailed" distributions you're talking about?

It’s an example but long tailed distributions can have a mean, and short tailed distributions can have no mean. It’s not indicative.

> short tailed distributions can have no mean

That’s a surprise to me. Now I’m curious: With what definition of “short tailed” can that be true?

There’s no need to get mathy tbh. Distributions are not real, and the ones we create only describe parts of what actually happens. This is obvious and always has been.

How much water does your faucet leak per day? That might be a good case for a normal distribution. But today I smack it with a hammer and it can’t stop flowing. We now have a ton of water far more than should ever appear on that original normal distribution. Is that because the model is wrong? No. It’s because the model is not designed for these circumstances.

We can’t build a model that estimates flow distribution including the possibility that I smack it with a hammer. That doesn’t mean we don’t believe it won’t ever happen or even that we think it is unlikely to happen. It’s just a weird edge case.

A lot of people (taleb) imply statisticians are idiots because they insist the distributions are comprehensive. But statisticians don’t insist this. Black swan, fat tails, and all that are just terms for things outside modeled conditions. And yeah. They happen.

If you work in finance and your competitors attended those universities you have an edge.

The central limit theorem applies to sample means, not individual events. I’m not sure it has any relevance to predicting the probability of a single large outlier tsunami