Hacker News Reader

Comment by _xnmw | Hacker News Reader

Reminds me of when the full-scale invasion started in Ukraine, I was in Kharkiv, and I met several people for whom "this was not their first war". The neighbour's family was from Nagorno-Karabakh, and had just fled the Armenia-Azerbaijan war that started a year earlier, the taxi driver was a Syrian refugee, and Kharkiv has many people from the Donbass region where the war started years before.

People who have experienced one extremely unlikely event are more likely to experience it again, see Poisson clumping (although I know this type of event is not independently random). Ernest Hemingway and his wife survived two plane crashes a day apart.

View on HN

The similarities in repetitions are indeed present, but not only in wars. No one forced this person to go to work after such an event, but apparently, he is truly an exceptional worker, and the Mitsubishi managers were incredibly fortunate. He was already on his 9th day back at work, covered in bandages, after a nuclear explosion. (Of course, I think he wanted to share his experience as soon as possible to help others, given it was wartime.)

After the start of the war in Ukraine, companies divided into three categories:

Good ones: They paid 3+ months' salaries in advance, covered the relocation costs for the entire family, and other expenses. For startups, these were donation-investments of $100k like in Awesonic (YC).

Ordinary ones: They paid nothing extra but provided some money for relocation until the next salary.

Bad ones: They paid nothing, no advance payment and demanded reports for working hours from bomb shelters during bombings.

I think such coincidences are not accidental, and this case with Tsutomu falls into either the 2nd or 3rd category. It's unclear how these coincidences would play out in peaceful life without war, but no one knew them except the management of the company where he worked.

Are you still in Kharkiv?

I'm not in Kharkiv anymore, but I'm in a relatively calmer part of Western Ukraine.

Stay safe, I hope humanity will heal from the virus of war in the near future.

That isn’t true, and isn’t what poisson clumping is caused by. Totally random independent events clump together entirely by chance: having experienced one doesn’t increase or reduce the probability of it occurring again.

I've had this discussion with a professional mathematician. As unintuitive as it does sound, actually random independent events are more likely to follow after it has happened.

https://stats.stackexchange.com/questions/411018/poisson-dis...

You as misinterpreting their reply. The fact that the Poisson distribution falls off exponentially is because the longer you wait, the more likely that the very next event has already happened. Once an event happens, that specific event cannot happen again so the probability is low after long distances and/or time intervals.

This effect only shows up when you are looking at the probabilities of individual events- when you look at the aggregate results after a long time, they are totally random, and no more likely to occur next to one another. The exponential nature of the Poisson distribution does not cause any clumping.

If you survive a plane crash, you are still no more likely to survive another plane crash than a regular person, unless you do something else that makes it no longer independent, and therefore no longer following the same Poisson distribution as everyone else (e.g. you fly a lot).

However, assuming standard behavior, if you do get in a plane crash, you are more likely to get in your very next plane crash in the next year, than in year 10. However, a person never in a plane crash is also equally more likely to get in their first plane crash in the next year than 10 years out. And you are still just as likely to get in a plane crash in year 10, just not your next one.

Poisson clumping is just a result of human bias-sort of like an optical illusion. We incorrectly expect random things to be evenly spaced. Music players with random shuffle for example, will often seem to play the same song over and over, and will only feel 'random' when they are modified to be non-random such that they avoid playing the same song within some time interval.

Reply on news.ycombinator.com