A friend told me something I have been thinking about: “Pessimists sound smart but optimists make money.”
This reminded me of something I observed a lot as kid: In social scenarios, it felt way more socially acceptable to be bearish than to be bullish. Say you recommend someone a flavour of ice cream and they try it because of your recommendation but don't like it. This disapproval, although subconcious, is a more negative reinforcement than if you had told them some flavour of ice cream that suck but they liked it.
A few days earlier, I was thinking of the difference in how people form bullish or bearish opinions:
My working theory is that being correctly bearish is way easier than being correctly bullish. If i try to put the other two options in the same spectrum, ig it'd be:
correctly bearish > incorrectly bullish > incorrectly bearish > correctly bullish
This stems from the fact that most things fail.
In my [research project](/projects#CSI), I learned that F1 scores are sometimes a better measure for an ML model than accuracy.
F1 score = 2 × (precision × recall)/(precision + recall)
precision = true positives / (true positives + false positives)
ie, how often you were right when you said “yes”
recall = true positives / (true positives + false negatives)
ie, how often you caught all the actual yeses
F1 score balances both when you care about not spamming yes and not missing yes.