On Jan 19 2021, I joined Metaculus and made my first predictions. I was prompted after successfully forecasting how long my brother will stand in a queue (my median was 55 minutes, he took 53), and then after going to the website, they had a bright mode that was easy on the eyes.
One of my first predictions was that the inauguration of Biden would be on the 20th of Jan. I looked up Wikipedia and in the last 20 inaugurations almost all were on that date, and the edge cases didn't seem to apply. I thought 95% was a good estimate.
The community forecast was also 95%, but before choosing it I saw it came with +50 and -150. That was too high of a loss so I played with the numbers for a while
A 99% confidence came with +55 and a massive -323.
Going to 90% decreased my gains to +44, and reduced the potential loss to -70.
I choose 90%
I realized that what I was doing was making decisions based on Karma, rather than what I thought about the world. I heard that before I thought, and then I remembered it was Goodhart's Law I was thinking about. I continued doing this for a few questions, I went to questions that were resolving soon and choose forecasts that provided good Karma. (less than -5 and more than +15).
I calculated that if I answer ~50 questions I would get 1000 points easily and get ranked.
I thought that was a bad thing, mainly for two reasons.
1. Because of honesty/integrity stuff.
2. Because I liked forecasting as a concept and I wouldn't want one of the major hubs filled with blind-karma optimizers.
I'm not sure if the existence of blind-karma optimizers would distort the forecasts, they might create too much noise or something, but I'm very unsure what the effects are.
I'm not sure what could be done, but if anyone can easily get many points by sliding a scale, it seems that something should be done.
