5 ms·
I'm pretty impressed that it took less than 1000 attempts. It would still be feasible even it were several orders of magnitude harder. This twitter project I ma
by jere 6y ago
I'm pretty impressed that it took less than 1000 attempts. It would still be feasible even it were several orders of magnitude harder. This twitter project I made (predicting celebrity deaths) took about 250k tweets over the course of a few months: https://twitter.com/ghastly_omens https://twitter.com/ghastly_omens
- OisinMoran 6y agoThis is great! That's not 250k tweets per death, right? Did you do the classic make a lot of predictions then delete the wrong ones? How did you select the celebrities? Would be interesting to do something with actuarial tables here.
- jere 6y agoYes, delete the misses. 250k total. About 700 celebrities. I was predicted the day of death too, so each required 365 tweets. I did a combination of scraping IMBD + manual selection plus filling in the details. The whole process could probably be automated and I regret not doing that. It would be wild to see this in 2020... I did consult actuarial tables mainly to decide how many of each age to include. With that few candidates, it wasn't wise to focus on people in their 20s and 30s obviously because it's very likely all of them would have survived. A few more details here: http://jere.in/i-predicted-23-celebrity-deaths-in-2017-then-my-predictions-came-true http://jere.in/i-predicted-23-celebrity-deaths-in-2017-then-...