I Predicted 23 Celebrity Deaths in 2017 - Then My Predictions Came True

Let's make 3 things crystal clear before we start:

Twenty sixteen was a terrible year for celebrity deaths. Or... a really good year depending on your perspective. Famous people were dropping like flies. You could feel it in the air. Musicians, actors, novelists, astronauts, dictators; 2016 claimed all. David Bowie, Harper Lee, Prince, Muhammad Ali, Gene Wilder, Leonard Cohen, Fidel Castro, and John Glenn to name a few. Later analyses would show that 2016 actually was a statistical anomaly in regards to dying celebs. Things were so bad that a GoFundMe was launched called Help protect Betty White from 2016.

Betty White

A very important squirrel

My project to predict celebrity deaths started in the fall of 2016, as the United States was preparing to elect a reality TV show host as president. It was clear that people couldn't get enough celebrity news, especially when it was an obituary. It was the first discussion in the office. It was on the front page of every magazine and newspaper. Scrolling across every ticker on every 24 hour news show. It's in your face all the time. Things happen to people we've never met and never knew we existed that have no material impact on us and yet we can't look away. Never mind the 150,000+ "normal" people that die every day on this planet.

Zuckerberg drinking water

In the timeless words of our next celebrity president Mark Zuckerberg:

"A squirrel dying in front of your house may be more relevant to your interests right now than people dying in Africa."

And I thought, "If people saw that these celebrity deaths were prophesied, they would lose their damn minds."

I made my predictions on a twitter account called Ghastly Omens. Not only did I predict 23 celebrities would die in 2017, but I predicted the exact day each of them would die. You can go see for yourself. Tweets made in 2016. Deaths happening in 2017. Every single tweet spot on. It's all very spooky!

So what's the trick?

A lie of omission

It's all rather simple. In addition to 23 accurate predictions, I also made about 250,000 that were false. I kept the hits and deleted the misses. Prophecy and prediction is easy if you can sweep your misses under the rug.

I later found out that mine wasn't the first project to attempt such predictions. The most well known was beyoncefan666, who predicted Trump getting elected, Britain leaving the EU, and Beyonce having twins. The whole idea is explained very well by How to Flawlessly Predict Anything on the Internet.

Believe it or not, I wasn't aware of any of this when I started Ghastly Omens. BeyonceFan666 hadn't been made public yet, though when it did it instantly went viral gaining over 30,000 followers. Having seen that, with my plans already in motion, I figured my project would scare the pants off of people. Trump? Brexit? That's small potatoes compared to the exact day on which people die. As it turns out, nobody would end up caring. More on that later.


It's not easy making or deleting 250,000 tweets, so let me explain how it all worked. First, it helps to know a few important facts about Twitter that make this all possible:

I would make my account private, accepting no followers. With no one watching, I could post tweets that no one would ever see and would not be cached by search engines. When the project was wrapped up, I could delete all but the 23 hits and then turn my account public again.

But there is also a practical limitation. Access to the Twitter API is rate limited, equating to a maximum of roughly 100 tweets per hour. I would have to spread my posts over 4 months, September to December 2016. Some of my predictions were already proving false during this time period. Carrier Fisher passed away on December, 27 2016. Her mother, Debbie Reynolds, died the next day.

XCKD Actuarial comic


Have you ever wondered: how many famous people are in your head? While you might personally know a few hundred acquaintances, the number of actors, musicians, writers, directors, athletes, and politicians you are aware of probably totals in the thousands.

With time for only 250,000 predictions and 365 required per celebrity, I was restricted to a list of about 700 people. My list would be far from exhaustive. I compiled it mostly by trawling IMDB's celebrities ranked by their "STARmeter", whatever the hell that means, then tacking on names from a few other random lists of famous musicians, novelists, etc.

Only the popular die young

Inclusion onto the list was a tradeoff between popularity and age. I obviously preferred popular celebrities, but I also needed a good number of accurate predictions. This is where the cold hard math of actuarial tables comes into play. At age 20, your chance of dying in the next year is about 1 in 1000. At age 90, it's 1 in 6.

If I filled my list up with the hottest actors (who tend to be 20 or 30 somethings), I'd risk not having any hits.

If the list was only centenarians (good luck compiling such a large list), I'd have some nice predictions but few would recognize the names.

The end result was a focus mostly on celebrities age 60-80. For this reason, my predictions completely missed younger souls like Charlie Murphy (57) and Chester Bennington (41). The youngest person to die on the list was Bill Paxton (61) and the oldest Irwin Corey (102).

I settled on a simple format for the tweets much like a gravestone: name followed by birth and death date. I wrote a PhantomJS script to iterate over my list and try to grab birthdays off of IMDB, but I had to fill in some gaps by hand.

I loaded up a MySQL database with the full list of predictions, then set up a cron job to run every 5 minutes for about 4 months. This job pulled a handful of random prediction records and posted a tweet for each using the Twitter API.

And then I waited.

Old man

Separating the living from the dead

This project required quite a bit of patience before seeing any payoff. I was hoping to reveal Ghastly Omens at the end of 2017 to coincide with end of year lists being compiled and shared. I couldn't wait until December to start deleting because deletions are also rated limited, so I had to do it in batches throughout the year. Without any automated to way actually find who died on my list, I simply checked IMDB and news sites periodically. Once I knew someone on the list had died, I flagged them in the database so that a subsequent deletion script would avoid deleting the associated tweets. There wasn't much room for error here. Deleting a tweet is nonreversible.

The moderately sized reveal

When I finally switched my Twitter account from private to public, I ran into two big problems.

Even though my deletion script had marked all but 23 tweets as deleted, my total tweet count displayed over 50. Something was wrong. I assumed at first that it was some eventually consistent nonsense, you know, the way reddit upvote counts fluctuate wildly up and down upon refreshing the page. That wasn't it. The Twitter API had failed to delete some of my 250,000 tweets, but I didn't know which ones! Talk about needles in haystacks....

You might assume this was as simple as going onto the Twitter client and manually clicking "delete" a few dozen times. Nope. I couldn't even see these phantom tweets. I soon learned that Twitter only shows your most recent 3000 tweets. Somehow other people were seeing the tweets that I could not see on my own timeline. This was a tad infuriating, but I found a third party app that somehow provided links to all of my old tweets and I deleted them. Luckily those extra tweets hadn't doomed the poject because pretty much no one knew about it yet...

A lot of tweets

I have absolutely no idea how to market anything

The second problem was not as easily resolved. Long story short, no one cared.

I didn't have a great plan for marketing the account, but I assumed that it wouldn't take much to go viral. I followed over 1900 people and, by hand, liked over 700 tweets (mostly to do with celebrity deaths which I had predicted).

But overall, it was radio silence. Fewer than 200 followers, most of which were bots. I'm sort of used to it at this point. I have two Twitter bots with a combined followership of 32.

I do wonder if my project itself is flawed. Was I too subtle with the presentation? I don't go outright and say HEY I MADE THESE TWEETS IN THE PAST. It's more up to the reader to connect the dots and look at the timestamp. Did it fly over their heads? Or maybe Twitter is just too skeptical these days and saw right through the gimmick, though I really doubt it.

Memento mori

My last guess is the most dismal. Twenty seventeen wasn't an outlier for celebrity deaths like 2016 was. Perhaps the celebrities on my list are just too old to get the attention of influencers. Because eventually even celebrities are forgotten. Even Batman. Even James Bond. It happens to everyone.

Memento mori: remember that you will die.

Ghastly Omens hits

Vote on HNDiscuss on Hacker News, you know, if that's like your thing.


A Great Disturbance in the Neighborhood - The Castle Doctrine Lives!
Year in Review 2016
Make America Procedurally Generated Again!


Short: http://jere.in/21
Pretty: http://jere.in/i-predicted-23-celebrity-deaths-in-2017-then-my-predictions-came-true

Oh hi there. I'm Jeremiah. I like to make stuff with code.