Being on the right side of maybe has never felt so depressing.
Conversation
I plan to write up a forecast retrospective for soon™, but overall I performed well—especially towards the end.
Me vs. Metaculus Community during the final days with probabilities recorded when I updated or at 1pm GMT. (2/6)
1
1
7
Stats:
—Time: 64.5 days
—Updates: 80
—Median daily prediction: 78%
—Average daily prediction: 76%
—Lowest daily prediction: 45%
—Days spent >=60%: 62.5
—Days spent >=80%: 31.5
—Days spent >=90%: 12.5
—Days spent >=95%: 5.5
—Days spent >=98%: 4.5
—Days spent >=99%: 3.5
(3/6)
1
3
My forecast definitely benefited from a strong prior model (something I believe many lacked) and a lot of hours on question, but equally if not more important to my success were the two Twitter lists l curated.
Main: twitter.com/i/lists/148203
Extra: twitter.com/i/lists/149228 (4/6)
1
1
3
Conversely, I'm disappointed with how many forecasters performed. Slow to update, noisy and volatile, or a combo of the two.
described my sentiment well: "This crisis has not been the forecasting community’s finest hour". inews.co.uk/news/world/rus (5/6)
3
3
19
Finally, I’ve added my #RussiaUkraine forecasts to the article which you can view here: globalguessing.com/russia-ukraine.
I've also reworked my main Twitter list for the invasion. If you want more accounts, check the Extra list above. (6/6)
1
5
That's my hypothesis. The ratio of forecasts per day : total forecasters was low everyday I updated the numbers.
2
2
Show replies
Rarely do you see similar questions across all the platforms. However, when it comes to big questions I plan to do more of this aggregation on . I did something similar for Powell's renomination:
Quote Tweet
Bravo to the #superforecasters at Good Judgment Inc. (@superforecaster) who held the most steadfast on #Powell's renomination over the past few weeks! The lowest briers of the bunch!
Nevertheless, the PMs deserve a round of applause for giving better info than the #MSM. twitter.com/ClayGraubard/s…
Show this thread
1
2
Show replies



